JP6451414B2 - Information processing apparatus, summary sentence editing method, and program - Google Patents

Description

  The present invention relates to an information processing apparatus, a summary sentence editing method, and a program.

  In the present day when a large amount of information is provided in the form of a document, the summary sentence is useful for efficiently grasping the contents of the document in a short time and searching for and utilizing the target document. However, it takes time and effort to summarize a large number of documents. Therefore, document summarization techniques that support the creation of summary sentences using computers have been studied. This technology is expected to be applied to, for example, summary summaries created in the medical field, analyst report summaries created in the stock and securities fields, and call report summaries created in call center operations.

  For example, a technology is proposed that creates a summary sentence by performing a morphological analysis / syntactic analysis (dependency structure analysis) on the source document (hereinafter referred to as the original text) and deleting a part that the computer determines is unnecessary from the original text. Has been. In addition, a technology has been proposed in which multiple ratios (summary rates) for deleting character strings from the original text are set in advance, and the user can select a desired summary text from a plurality of summary texts created according to the summary rate. Has been. In addition, a technique for highlighting a word designated by a user, a technique for inserting / removing a word / phrase prepared in advance corresponding to a designated location, and the like have been proposed.

International Publication No. 2010/052764 JP-A-11-25091 Japanese Patent Laid-Open No. 11-219361 JP 2014-56499 A

  Although the above-described technique is applied to support the creation of a summary sentence, additional editing work occurs when words deleted from the original sentence are not appropriate. For example, when a phrase longer than the user desires is deleted, an operation of inserting the deleted phrase into the summary sentence again occurs. On the other hand, when a word that the user wants to delete remains without being deleted, an operation of deleting the word from the summary sentence occurs. Among the techniques described above, a technique that allows words to be inserted / removed only by a specified operation can contribute to reducing the burden of such subsequent editing work.

  However, there is a case where the expression still desired by the user is not obtained even when the prepared phrase is inserted / removed. In this case, the user directly edits the summary sentence. If the degree of freedom of words that can be inserted / removed by the designation operation can be increased, the possibility of completing the editing of the summary sentence only by the designation operation is increased, and a reduction in the workload is expected.

  Therefore, according to one aspect, an object of the present invention is to provide an information processing apparatus, a summary sentence editing method, and a program that can easily edit a summary sentence.

  According to one aspect of the present disclosure, a storage unit that stores an original sentence and a syntax tree in which a dependency structure of a phrase based on a syntactic analysis of the original sentence is expressed by a connection relation of nodes corresponding to the phrase; And a display section that displays a summary sentence that summarizes the original text with the phrase omitted, and the root of the syntax tree connected to the first node corresponding to the phrase at the specified location when a specification operation for the original text is accepted. When the second node in the direction toward is specified, the words corresponding to the first and second nodes are added to the summary sentence, and the designation operation for the summary sentence is accepted, the word corresponding to the designated place An operation unit that identifies a fourth node that is connected to the third node and that is in a direction toward the end of the syntax tree, and that deletes words corresponding to the third and fourth nodes from the summary sentence. An information processing apparatus is provided.

  According to the present invention, it is possible to easily edit a summary sentence.

It is the figure which showed an example of the information processing apparatus which concerns on 1st Embodiment. It is the figure which showed an example of the hardware which can implement | achieve the function of the information processing apparatus which concerns on 2nd Embodiment. It is the block diagram which showed an example of the function which the information processing apparatus which concerns on 2nd Embodiment has. It is the figure which showed an example of the morphological analysis result which concerns on 2nd Embodiment. It is the figure which showed an example of the syntax analysis (dependency analysis) result concerning 2nd Embodiment. It is the figure which showed an example of the syntax tree and summary sentence text which concern on 2nd Embodiment. It is the figure which showed an example of the corresponding data of the original sentence and syntax analysis result which concern on 2nd Embodiment. It is the figure which showed an example of the corresponding data of the summary sentence and syntax analysis result concerning 2nd Embodiment. It is the figure which showed an example of designation | designated operation and a process about addition of the phrase which concerns on 2nd Embodiment. It is the figure which showed an example of designation | designated operation and a process about deletion of the phrase which concern on 2nd Embodiment. It is the 1st flowchart which showed the flow of the process about operation | movement of the information processing apparatus which concerns on 2nd Embodiment. It is the 2nd flowchart which showed the flow of the process about operation | movement of the information processing apparatus which concerns on 2nd Embodiment. It is the figure which showed an example of the co-occurrence probability table which concerns on 3rd Embodiment. It is the figure which showed an example of designation | designated operation and a process about addition of the phrase which concerns on 3rd Embodiment. It is the figure which showed an example of designation | designated operation and a process about deletion of the phrase which concern on 3rd Embodiment. It is the 1st flowchart which showed the flow of the process about the operation | movement of the information processing apparatus which concerns on 3rd Embodiment. It is the 2nd flow figure showing the flow of processing about the operation of the information processor concerning a 3rd embodiment. It is the 3rd flowchart which showed the flow of the process about the operation | movement of the information processing apparatus which concerns on 3rd Embodiment. It is the 4th flow figure showing the flow of processing about the operation of the information processor concerning a 3rd embodiment. It is the figure which showed an example of the syntax analysis (dependency analysis) result which concerns on 4th Embodiment. It is the figure which showed an example of designation | designated operation and a process about addition of the phrase which concerns on 4th Embodiment. It is the 1st flowchart which showed the flow of the process about operation | movement of the information processing apparatus which concerns on 4th Embodiment. It is the 2nd flowchart which showed the flow of the process about operation | movement of the information processing apparatus which concerns on 4th Embodiment. It is the 3rd flow figure showing the flow of processing about the operation of the information processor concerning a 4th embodiment.

  Embodiments of the present invention will be described below with reference to the accompanying drawings. In addition, about the element which has the substantially same function in this specification and drawing, duplication description may be abbreviate | omitted by attaching | subjecting the same code | symbol.

<1. First Embodiment>
The first embodiment will be described with reference to FIG. FIG. 1 is a diagram illustrating an example of an information processing apparatus according to the first embodiment. The first embodiment relates to a document summarization technique for automatically creating a summary sentence by summarizing documents, and an editing support technique for supporting editing of the summary sentence. Hereinafter, the document of the summarization source is referred to as the original text. Further, for convenience of explanation, an explanation will be given taking as an example a case where a summary sentence 32 is created by summarizing the original sentence 31 shown in FIG. 1 and the summary sentence 32 is edited.

As illustrated in FIG. 1, the information processing apparatus 10 includes a storage unit 11, a calculation unit 12, and a display unit 13.
The storage unit 11 is a volatile storage device such as a RAM (Random Access Memory) or a nonvolatile storage device such as an HDD (Hard Disk Drive) or a flash memory. The arithmetic unit 12 is a processor such as a CPU (Central Processing Unit) or a DSP (Digital Signal Processor). However, the arithmetic unit 12 may be an electronic circuit such as an application specific integrated circuit (ASIC) or a field programmable gate array (FPGA).

  For example, the calculation unit 12 executes a program stored in the storage unit 11 or another memory. The display unit 13 is a display device such as CRT (Cathode Ray Tube), LCD (Liquid Crystal Display), PDP (Plasma Display Panel), or ELD (Electro-Luminescence Display).

  The display unit 13 does not have to be formed integrally with the information processing apparatus 10. For example, the content displayed on the display unit 13 is a display device of an information terminal connected to the information processing apparatus 10 via a network. You may make it the structure displayed on. The functions of the information processing apparatus 10 can also be realized by a system that combines a plurality of computers that share the functions of the arithmetic unit 12 and a storage apparatus that has the functions of the storage unit 11.

The storage unit 11 stores the original text 31 and the syntax tree 20.
The syntax tree 20 is information that expresses a dependency structure of a phrase based on the parsing of the original sentence 31 by a connection relation of nodes corresponding to the phrase. The phrase is, for example, a clause or a phrase.

  Parsing is a method of analyzing the structure of a sentence in units of phrases and clauses according to grammatical rules. A phrase refers to a set of words in which two or more words come together and function in the same way as a single part of speech. A phrase is a unit in which Japanese is delimited by a unit whose meaning is understood, and is the smallest unit that is delimited by natural pronunciation when a sentence is read. Any one phrase in the Japanese sentence has a dependency relationship with at least one phrase following the phrase. A phrase structure having such a dependency relationship is called a dependency structure.

  Note that morphological analysis is performed as a premise when performing syntax analysis. Usually, the smallest meaningful character string unit is called a morpheme. The operation of dividing a sentence into words and adding part-of-speech information to each word is called morphological analysis. A system that performs morphological analysis divides a sentence into morphemes using natural language grammar rules and dictionary information, and mechanically gives part of speech information to each word. For example, when the original text 31 shown in FIG. 1 (B1) is analyzed by the above analysis method, a syntax tree 20 as shown in FIG. 1 (A) is obtained.

  In the example of Fig. 1 (A), "Last year", "End of August", "Hot", "Sun", "In the economic platform", "Heavy", "It was", "One person", "", "" "", "" Mid-"", "" the world "", "" "left" "are phrases that are elements of the syntax tree 20. Hereinafter, elements of the syntax tree 20 are referred to as nodes. The syntax tree 20 expresses a dependency relationship between phrases as a connection relationship between nodes (in FIG. 1A, a line connecting nodes). Hereinafter, the connection relationship between nodes may be referred to as a branch.

  A node located at the root of the syntax tree 20 having a tree structure may be referred to as a root node, and a node located at the end may be referred to as a terminal node. In the example of FIG. 1A, the node corresponding to “Leave” is the root node, and “Last year”, “Hot”, “In the economic platform”, “Heavy”, “One person”, “The world” , “Zhi” are terminal nodes. That is, a word having a dependency relationship does not follow the word corresponding to the root node, and a word having a dependency relationship does not exist before the word corresponding to the terminal node. The storage unit 11 stores information related to such a syntax tree 20.

  The display unit 13 displays an original sentence 31 and a summary sentence 32 that summarizes the original sentence 31. The summary sentence 32 is obtained by omitting some words included in the original sentence 31 based on the syntax tree 20. For example, the summary sentence 32 arbitrarily selects a path connecting a series of nodes and branches from the root node to the end node, and arranges words and phrases corresponding to each node in the selected path in the same order as the words and phrases in the original text 31. Can be obtained. FIG. 1B1 illustrates a summary sentence 32 when a path to the terminal node corresponding to “one person”, “will”, and “the world” is selected.

  The display unit 13 displays both the original sentence 31 and the summary sentence 32. Then, the calculation unit 12 receives a user's designation operation for the original sentence 31 and the summary sentence 32. When the operation unit 12 receives a designating operation on the original text 31, the computing unit 12 identifies the second node that is connected to the first node corresponding to the phrase at the designated location and that is in the direction toward the root of the syntax tree 20. , Words corresponding to the first and second nodes are added to the summary sentence 32.

  In the example of FIG. 1 (B1), “heavy” in the original text 31 is designated. In this case, the computing unit 12 identifies the node corresponding to “heavy” as the first node, and identifies the node in the direction from the first node toward the root node as the second node. Note that the designation operation may be an operation of selecting and designating a phrase, or an operation of designating characters. In the case of an operation for designating a character, the calculation unit 12 identifies a word / phrase including the designated character and determines that the specified word / phrase has been designated.

  In this example, the calculation unit 12 detects “was present”, “examiner”, and “exited” as candidates for the second node, and the “examiner” already included in the summary sentence 32, “ “Exist” other than “Leave” is specified as the second node. Then, the computing unit 12 determines the first and second nodes as the additional range 21, and adds “heavy” and “was present” corresponding to the additional range 21 to the summary sentence 32 (FIG. 1 (B2)). See underlined).

  On the other hand, when the designation operation for the summary sentence 32 is received, the arithmetic unit 12 selects the fourth node that is connected to the third node corresponding to the phrase at the designated location and is in the direction toward the end of the syntax tree 20. The words corresponding to the third and fourth nodes are deleted from the summary sentence 32.

  In the example of FIG. 1 (B2), “mid-half” of the summary sentence 32 is designated. In this case, the computing unit 12 identifies the node corresponding to “mid-way” as the third node, and identifies the node in the direction from the third node toward the terminal node as the fourth node.

  In this example, the arithmetic unit 12 detects “zhi” as a fourth node candidate, confirms that “zhi” is already included in the summary sentence 32, and sets “zhi” as the fourth node. Identify. Then, the calculation unit 12 determines the third and fourth nodes as the deletion range 22, and deletes “zhi” and “mid-d” corresponding to the deletion range 22 from the summary sentence 32 (see FIG. 1 (B3)). ).

  According to the first embodiment, a phrase at a specified location is inserted / removed based on a dependency relationship having a phrase or phrase as a unit. When a part of the original sentence 31 is designated, the phrase at the designated place and the phrase having a dependency relation are inserted into the summary sentence 32, and when a part of the summary sentence 32 is designated, the phrase at the designated place. And the phrase having a dependency relationship are deleted from the summary sentence 32.

  Even if nodes are on the same path, if a word corresponding to a different node is selected, a word to be inserted / removed starting from the node corresponding to the word is determined. Therefore, by adjusting the word to be added or deleted while changing the designated portion, the desired summary sentence 32 can be obtained by repeating the designation operation. That is, the degree of freedom of editing by the designation operation is improved as compared with the case where the words to be inserted / removed are fixed, and the summary sentence can be brought closer to a desired expression by a simpler operation. As a result, the summary sentence can be easily edited.

The first embodiment has been described above.
<2. Second Embodiment>
Next, a second embodiment will be described. The second embodiment proposes a method for enabling addition / deletion of words to / from a summary sentence with a simple operation, with respect to the summary sentence editing support method. Hereinafter, the information processing apparatus 100 capable of realizing this method will be described. The information processing apparatus 100 is an example of an information processing apparatus according to the second embodiment.

[2-1. hardware]
Here, the hardware of the information processing apparatus 100 will be described with reference to FIG. FIG. 2 is a diagram illustrating an example of hardware capable of realizing the functions of the information processing apparatus according to the second embodiment. That is, the functions of the information processing apparatus 100 to be described later can be realized using the hardware resources illustrated in FIG. The functions of the information processing apparatus 100 are realized by controlling the hardware shown in FIG. 2 using a computer program.

  Note that the technology according to the second embodiment can be realized by using one information processing apparatus having the hardware illustrated in FIG. 2, but a plurality of information processing apparatuses, storage apparatuses, and the like are provided. It can also be realized by a system connected via a network. Such a modification naturally belongs to the technical scope of the second embodiment.

  As shown in FIG. 2, this hardware mainly includes a CPU 902, a ROM (Read Only Memory) 904, a RAM 906, a host bus 908, and a bridge 910. Further, this hardware includes an external bus 912, an interface 914, an input unit 916, an output unit 918, a storage unit 920, a drive 922, a connection port 924, and a communication unit 926.

  The CPU 902 functions as, for example, an arithmetic processing unit or a control unit, and controls the overall operation of each component or a part thereof based on various programs recorded in the ROM 904, the RAM 906, the storage unit 920, or the removable recording medium 928. . The ROM 904 is an example of a storage device that stores a program read by the CPU 902, data used for calculation, and the like. The RAM 906 temporarily or permanently stores, for example, a program read by the CPU 902 and various parameters that change when the program is executed.

  These elements are connected to each other via, for example, a host bus 908 capable of high-speed data transmission. On the other hand, the host bus 908 is connected to an external bus 912 having a relatively low data transmission speed via a bridge 910, for example. As the input unit 916, for example, a mouse, a keyboard, a touch panel, a touch pad, a button, a switch, a lever, or the like is used. Furthermore, as the input unit 916, a remote controller capable of transmitting a control signal using infrared rays or other radio waves may be used.

  As the output unit 918, for example, a display device such as a CRT, LCD, PDP, or ELD is used. As the output unit 918, an audio output device such as a speaker or headphones, or a printer may be used. In other words, the output unit 918 is a device that can output information visually or audibly.

  The storage unit 920 is a device for storing various data. As the storage unit 920, for example, a magnetic storage device such as an HDD is used. Further, as the storage unit 920, a semiconductor storage device such as an SSD (Solid State Drive) or a RAM disk, an optical storage device, a magneto-optical storage device, or the like may be used.

  The drive 922 is a device that reads information recorded on a removable recording medium 928 that is a removable recording medium or writes information on the removable recording medium 928. As the removable recording medium 928, for example, a magnetic disk, an optical disk, a magneto-optical disk, or a semiconductor memory is used.

  The connection port 924 is a port for connecting an external connection device 930 such as a USB (Universal Serial Bus) port, an IEEE 1394 port, a SCSI (Small Computer System Interface), an RS-232C port, or an optical audio terminal. For example, a printer or the like is used as the external connection device 930.

  The communication unit 926 is a communication device for connecting to the network 932. As the communication unit 926, for example, a communication circuit for wired or wireless LAN (Local Area Network), a communication circuit for WUSB (Wireless USB), a communication circuit or router for optical communication, an ADSL (Asymmetric Digital Subscriber Line) Communication circuits, routers, communication circuits for mobile phone networks, and the like are used. A network 932 connected to the communication unit 926 is a wired or wireless network, and includes, for example, the Internet, a LAN, a broadcast network, a satellite communication line, and the like.

The hardware of the information processing apparatus 100 has been described above.
[2-2. function]
Next, functions of the information processing apparatus 100 will be described with reference to FIG. FIG. 3 is a block diagram illustrating an example of functions of the information processing apparatus according to the second embodiment.

  As illustrated in FIG. 3, the information processing apparatus 100 includes a storage unit 101, a source text input unit 102, a morpheme analysis unit 103, a syntax analysis unit 104, a summary sentence generation unit 105, a sentence output unit 106, a designation reception unit 107, and a range A control unit 108 is included.

  Note that the function of the storage unit 101 can be realized by using the above-described RAM 906, the storage unit 920, or the like. The functions of the original text input unit 102 and the designation receiving unit 107 can be realized by using the functions of the input unit 916 described above. The functions of the morphological analysis unit 103, the syntax analysis unit 104, the summary sentence generation unit 105, and the range control unit 108 can be realized by using the above-described CPU 902 or the like. The function of the sentence output unit 106 can be realized by using the output unit 918 described above.

(2-2-1. Generation of syntax tree, summary sentence, and corresponding data)
The storage unit 101 stores information such as the original text 101a and the analysis result 101b. The original text 101a is text data of the original text that is the source of the summary text.

  For example, the original text input unit 102 stores the original text 101 a input by the user using the input unit 916 in the storage unit 101. When the original text 101 a is in a storage device externally connected to the information processing apparatus 100 or a storage area on the network, the original text input unit 102 acquires the original text and stores it in the storage unit 101.

  The morpheme analysis unit 103 performs morpheme analysis on the original text 101a, and stores in the storage unit 101 information (see FIG. 4) obtained by adding parts of speech to each morpheme extracted from the original text 101a as a part of the analysis result 101b. To do. FIG. 4 is a diagram illustrating an example of a morphological analysis result according to the second embodiment. When the original text 101a is a sentence “a hot day at the end of August last year, a single expert who was heavy on the economic forum passed away in the middle”, the morpheme analysis unit 103 uses the morpheme illustrated in FIG. Output analysis results.

  The syntax analysis unit 104 performs syntax analysis (dependency analysis) of the original text 101a based on the morpheme analysis result output by the morpheme analysis unit 103. Parsing is a method of analyzing the structure of a sentence in units of phrases and clauses according to grammatical rules. In this paper, phrases and clauses that are units of parsing are simply called “phrases”. The syntax analysis unit 104 associates information (see FIG. 5) such as a phrase notation, a dependency destination, and a dependency type for each phrase obtained by the syntax analysis, and stores the information in the storage unit 101 as a part of the analysis result 101b. To do. FIG. 5 is a diagram illustrating an example of a syntax analysis (dependency analysis) result according to the second embodiment.

  As shown in FIG. 5, each word / phrase is assigned a node number, and each word / phrase can be specified based on the node number. The dependency relationship between words is expressed by a node number described in the dependency destination column. For example, the phrase “last year” of the node number 1 has a dependency relationship with the phrase “of the end of August” of the node number 2 as a destination. An expression of the tree structure representing the dependency relationship for each word is called a syntax tree, and the syntax tree corresponding to FIG. 5 is as shown in FIG.

  FIG. 6 is a diagram showing an example of a syntax tree and summary text according to the second embodiment. Each block of the syntax tree illustrated in FIG. 6 represents a node. A line connecting the blocks is a branch, and each branch represents a connection relationship between nodes (that is, a dependency relationship). In the example of FIG. 6, the root node of the syntax tree is a node corresponding to the phrase “Leave”. In addition, the terminal nodes are nodes corresponding to the phrases “last year”, “hot”, “in the economic forum”, “heavy”, “one person”, “the world”, and “zhi”, respectively.

  Further, the syntax analysis unit 104 generates correspondence data (see FIG. 7) that associates each character included in the original text 101a with a node corresponding to a word including the character. FIG. 7 is a diagram showing an example of correspondence data between the original text and the syntax analysis result according to the second embodiment. For example, as shown in FIG. 7, the syntax analysis unit 104 includes a number assigned to each character included in the original text 101a (hereinafter referred to as an original character number), a character notation, and a node corresponding to the character. Corresponding data in which numbers are associated is generated as a part of the analysis result 101 b and the corresponding data is stored in the storage unit 101.

  The summary sentence generation unit 105 generates summary sentence text based on a series of nodes connecting from the root node to the set node. In the example of FIG. 6, a summary sentence “one of the authors is half-hearted” generated based on a series of nodes that connect from the root node to the nodes corresponding to the phrases “one person”, “the world”, and “will”. I passed away. " As described above, the summary sentence generation unit 105 identifies the phrases corresponding to the nodes on the path to the set node, and generates the summary sentence text by arranging the identified phrases in the order of the phrases on the original text 101a. In the process of automatically generating summary text from the syntax analysis result, various methods using word importance, word N-gram, dependency type, etc. are set as methods for setting nodes to be included in the summary or nodes to be deleted in the summary. Although the method is known, the present invention does not specify any method. In addition, all nodes may be included in the summary sentence, and the same summary sentence text as the original text may be generated, or all nodes may be deleted and a character string of 0 characters may be used as the summary sentence text.

  The summary sentence generation unit 105 turns on the deletion flag of the phrases that are not included in the summary sentence text among the phrases included in the syntax analysis result (see FIG. 5). The deletion flag is managed in association with each word included in the syntax analysis result, and is used for the summary sentence editing process. The summary sentence generation unit 105 inputs the summary sentence text to the sentence output unit 106.

  Further, the summary sentence generation unit 105 generates correspondence data (see FIG. 8) for associating each character included in the summary sentence text with a node corresponding to the phrase including the character. FIG. 8 is a diagram illustrating an example of correspondence data between a summary sentence and a syntax analysis result according to the second embodiment. For example, as illustrated in FIG. 8, the summary sentence generation unit 105 includes a number assigned to each character included in the summary sentence text (hereinafter, a summary sentence character number), a character notation, and a node corresponding to the character. Is generated as part of the analysis result 101b, and the corresponding data is stored in the storage unit 101.

  The sentence output unit 106 acquires the original text 101 a from the storage unit 101 and displays the original text 101 a together with the summary text input from the summary generation unit 105. At this time, the sentence output unit 106 displays the original text 101a and the summary text in a format that allows the user to specify the characters on the original text 101a and the characters on the summary text.

(2-2-2. Editing summary sentences)
The designation receiving unit 107 receives a designation operation for the original text 101a or the summary text displayed by the sentence output unit 106. The designation operation is an operation of designating characters or phrases included in the original text 101a or the summary text (see FIGS. 9A and 10A). FIG. 9 is a diagram illustrating an example of the designation operation and processing for adding words according to the second embodiment. FIG. 10 is a diagram illustrating an example of a designation operation and processing for deleting words according to the second embodiment. Here, a case where a designation operation for designating characters is accepted will be described.

  When the designation receiving unit 107 receives a designation operation on the characters of the original text 101a, the designation receiving unit 107 refers to correspondence data (see FIG. 7) of the original text and the syntax analysis result, and identifies a node corresponding to the designated character. On the other hand, when receiving a designation operation for the characters of the summary sentence text, the designation receiving unit 107 refers to the correspondence data (see FIG. 8) of the summary sentence and the syntax analysis result, and identifies a node corresponding to the designated character. . The designation receiving unit 107 inputs the specified node information to the range control unit 108. At this time, the designation receiving unit 107 notifies the range control unit 108 whether the target of the designation operation is the original text 101a or the summary text.

  When the target of the designating operation is the original text 101a, the range control unit 108, as shown in FIG. 9B, the range of words to be added to the summary text based on the parsing result (hereinafter, additional range). To decide. At this time, the range control unit 108 extracts nodes on the path from the node corresponding to the designated character to the root node, and adds a node other than the node corresponding to the phrase already included in the summary text. Include in

  In the example of FIG. 9B, among the nodes on the path from the node corresponding to the phrase “heavy” to the root node corresponding to the phrase “leaved”, the phrase “discussion is” is included in the summary text. Therefore, the added range is a node corresponding to the words “heavy” and “was present”. In this case, as shown in FIG. 9C, the phrase “it was heavy” (underlined) is added to the summary text.

  On the other hand, when the target of the designation operation is a summary sentence text, the range control unit 108, as shown in FIG. Range). At this time, the range control unit 108 extracts a node on the path from the node corresponding to the designated character to the terminal node, and selects the node corresponding to the phrase already included in the summary text from the extracted nodes. Include in deletion range.

  In the example of FIG. 10B, of the nodes on the path from the node corresponding to the phrase “mid” to the terminal node corresponding to the phrase “zhi”, the phrases “middle” and “zhi” are summary sentences. Since it is included in the text, the deletion range is a node corresponding to the words “middle” and “will”. In this case, as shown in FIG. 10 (C), the phrase “mid-shi” is deleted from the summary text.

  The summary sentence generation unit 105 executes the above-described addition range addition processing and deletion range deletion processing. The information on the addition range or the deletion range determined by the range control unit 108 is input to the summary sentence generation unit 105, and the summary sentence generation unit 105 edits the summary sentence text based on this information, and the edited summary sentence text is displayed. Displayed by the sentence output unit 106. In this way, the edited summary text as shown in FIG. 9C or FIG. 10C is displayed.

  As described above, according to the information processing apparatus 100, the summary text can be easily edited by performing the designation operation on the original text 101a or the summary text. At the time of addition, words corresponding to a series of nodes connected in the direction of the root node of the syntax tree are set as an addition range, and at the time of deletion, a word corresponding to a series of nodes connected in the direction of the terminal node is set as a deletion range. Therefore, a set of words and phrases that are likely to be added or deleted at a time are processed together by a single designation operation, so that the summary text can be edited more easily, contributing to a reduction in the burden of editing work.

The function of the information processing apparatus 100 has been described above.
[2-3. Processing flow]
Next, the flow of processing executed by the information processing apparatus 100 will be described with reference to FIGS. 11 and 12. FIG. 11 is a first flowchart illustrating a process flow regarding the operation of the information processing apparatus according to the second embodiment. FIG. 12 is a second flowchart illustrating a process flow regarding the operation of the information processing apparatus according to the second embodiment.

  (S101) The original text input unit 102 acquires the original text 101a and stores it in the storage unit 101. For example, the original text input unit 102 stores the original text 101 a input by the user using the input unit 916 in the storage unit 101. When the original text 101 a is in a storage device externally connected to the information processing apparatus 100 or a storage area on the network, the original text input unit 102 acquires the original text from the original text and stores it in the storage unit 101.

  (S102) The morpheme analysis unit 103 performs morpheme analysis on the original text 101a, and stores information (see FIG. 4) in which each morpheme extracted from the original text 101a has a part of speech as a part of the analysis result 101b. 101. For example, if the original text 101a is a sentence "A hot day at the end of August last year, a single expert who was heavy on the economic platform passed away in the middle", the morphological analysis results shown in Fig. 4 were obtained. It is done.

  (S103) The syntax analysis unit 104 performs syntax analysis (dependency analysis) of the original text 101a based on the morpheme analysis result output by the morpheme analysis unit 103. Parsing is a method of analyzing the structure of a sentence in units of phrases and clauses according to grammatical rules. The syntax analysis unit 104 associates information (see FIG. 5) such as a phrase notation, a dependency destination, and a dependency type for each phrase obtained by the syntax analysis, and stores the information in the storage unit 101 as a part of the analysis result 101b. To do.

  (S104) The syntax analysis unit 104 generates correspondence data (see FIG. 7) for associating each character included in the original text 101a with a node corresponding to the phrase including the character. For example, as shown in FIG. 7, the syntax analysis unit 104 associates the original character number of each character included in the original text 101a, the character notation, and the node number of the node corresponding to the character. Is generated as a part of the analysis result 101b.

  (S105) The syntax analysis unit 104 turns off all the deletion flags associated with each word / phrase included in the syntax analysis result (see FIG. 5) (initialization). The deletion flag is set to OFF for words included in the summary sentence text and ON for words not included in the summary sentence text.

  (S106) The summary sentence generation unit 105 sets the deletion flag corresponding to the node to be deleted when generating the summary sentence to ON. For example, the summary sentence generation unit 105 identifies a phrase corresponding to a node on a path from the root node to a node set for summary sentence generation, and turns on a deletion flag corresponding to a node other than the identified node. Set to.

(S107) The summary sentence generation unit 105 generates a summary sentence text by concatenating words and phrases corresponding to the node whose deletion flag is OFF in the order of appearance on the original text 101a.
(S108) The summary sentence generation unit 105 generates correspondence data (see FIG. 8) for associating each character included in the summary sentence text with a node corresponding to the phrase including the character. For example, as illustrated in FIG. 8, the summary sentence generation unit 105 associates the summary sentence character number of each character included in the summary sentence text, the character notation, and the node number of the node corresponding to the character. Corresponding data is generated as a part of the analysis result 101b.

  (S109) The sentence output unit 106 acquires the original text 101a from the storage unit 101, and outputs the original text 101a together with the summary text generated by the summary sentence generation unit 105. At this time, the sentence output unit 106 displays the original text 101a and the summary text in a format that allows the user to specify the characters on the original text 101a and the characters on the summary text.

  (S110) The designation receiving unit 107 determines whether the summary sentence text output by the sentence output unit 106 has been confirmed. For example, the designation receiving unit 107 determines whether or not a summary sentence text editing end operation has been performed. When the summary sentence text is confirmed, the series of processes shown in FIGS. 11 and 12 ends. On the other hand, if the summary text has not been confirmed, the process proceeds to S111.

  (S111) The designation receiving unit 107 determines whether or not a character of the original text 101a has been designated. When the character of the original text 101a is designated, the process proceeds to S112. On the other hand, when the character of the original text 101a is not designated, the process proceeds to S114.

  (S112) When the designation receiving unit 107 receives a designation operation on the characters of the original text 101a, the designation receiving unit 107 refers to the correspondence data (see FIG. 7) between the original text and the syntax analysis result, and corresponds to the designated character (designated character). Identify the node.

  (S113) The range control unit 108 turns OFF the deletion flag corresponding to each node from the node identified by the designation receiving unit 107 to the root node based on the syntax analysis result. That is, as shown in FIG. 9B, the range control unit 108 extracts a node on the path from the node corresponding to the designated character to the root node based on the syntax analysis result, and has already added it to the summary text. Include nodes other than those corresponding to the included words in the additional range. When the process of S113 is completed, the process proceeds to S107.

  (S114) The designation receiving unit 107 determines whether characters of the summary text are designated. If characters of the summary text are designated, the process proceeds to S115. On the other hand, when the characters of the summary text are not designated, the process proceeds to S110.

  (S115) When the designation receiving unit 107 receives a designation operation for the characters of the summary sentence text, the designation receiving unit 107 refers to the correspondence data (see FIG. 8) of the summary sentence and the syntax analysis result, and identifies the node corresponding to the designated character.

  (S116) The range control unit 108 turns on the deletion flag corresponding to each node from the node specified by the designation receiving unit 107 to the terminal node based on the syntax analysis result. That is, as shown in FIG. 10B, the range control unit 108 extracts a node on the path from the node corresponding to the designated character to the terminal node based on the syntax analysis result, and has already added it to the summary text. Include nodes corresponding to the included words in the deletion range. When the process of S116 is completed, the process proceeds to S107.

The flow of processing executed by the information processing apparatus 100 has been described above.
According to the above processing method, the summary sentence text can be easily edited by performing the designation operation on the original sentence text 101a or the summary sentence text. At the time of addition, words corresponding to a series of nodes connected in the direction of the root node of the syntax tree are set as an addition range, and at the time of deletion, a word corresponding to a series of nodes connected in the direction of the terminal node is set as a deletion range. Therefore, a set of words and phrases that are likely to be added or deleted at a time are processed together by a single designation operation, so that the summary text can be edited more easily, contributing to a reduction in the burden of editing work.

The second embodiment has been described above.
<3. Third Embodiment>
Next, a third embodiment will be described. However, detailed description of the same parts as those described in the second embodiment will be omitted.

[3-1. function]
The information processing apparatus 100 according to the third embodiment determines the addition range and the deletion range in consideration of a co-occurrence probability indicating a high possibility that a plurality of phrases co-occur in the same text. In the information processing apparatus 100, the second embodiment described above is that the co-occurrence probability table indicating information on the co-occurrence probability of the node is stored in the storage unit 101, and that the range control unit 108 uses the co-occurrence probability. Different from form. Hereinafter, these differences will be mainly described.

(3-1-1. Co-occurrence probability)
The co-occurrence probability table will be described with reference to FIG. FIG. 13 is a diagram illustrating an example of a co-occurrence probability table according to the third embodiment. As illustrated in FIG. 13, the co-occurrence probability table is a table that associates the co-occurrence probability for each combination of the content word included in the dependency source node and the content word included in the dependency destination node. The co-occurrence probability table is stored in the storage unit 101 in advance. A content word is a word that has almost no grammatical functions, such as nouns, verbs, and adjectives, and mainly represents lexical meaning.

The co-occurrence probability is calculated using a text corpus based on the following equation (1). In the following formula (1), A → B means that node A relates to node B. In other words, node A is the source node and node B is the destination node. P (A → B) represents the co-occurrence probability of nodes A and B having a dependency relationship of A → B. M _A and M _B represent content words included in the nodes A and B, respectively. * Represents an arbitrary content word. N (M _A → M _B ), N (M _A → *), and N (* → M _B ) respectively represent the number of occurrences of M _A → M _B , M _A → *, * → M _{B in} the text corpus. .

(3-1-2. Editing summary text)
Next, the summary sentence editing process using the co-occurrence probability table will be described with reference to FIGS. FIG. 14 is a diagram illustrating an example of a specifying operation and processing for adding words according to the third embodiment. FIG. 15 is a diagram illustrating an example of a designation operation and processing for deleting a phrase according to the third embodiment.

  When the target of the designation operation is the original text 101a (see FIG. 14A), the range control unit 108 determines an additional range based on the syntax analysis result as shown in FIG. 14B. At this time, the range control unit 108 starts from a node corresponding to the designated character, extracts a node on the path from the starting point to the root node, and nodes other than the node corresponding to the phrase already included in the summary text Is included in the additional range.

  Further, the range control unit 108 refers to the co-occurrence probability table, and extracts nodes having a co-occurrence probability equal to or higher than a preset threshold among nodes connected in the direction from the starting point to the terminal node. Further, the range control unit 108 uses the extracted node as a new starting point, and extracts a node having a co-occurrence probability equal to or higher than a preset threshold among nodes connected in the direction from the starting point toward the terminal node. Then, the range control unit 108 repeats this extraction process and adds the extracted node to the additional range. The threshold value may be a statistically valid value based on the co-occurrence probability distribution obtained from the text corpus, or may be arbitrarily set by the user.

  In the example of FIG. 14B, among the nodes on the path from the node corresponding to the phrase “existing” corresponding to the designated character to the root node corresponding to the phrase “leaving”, the phrase “discussion is” Is included in the summary text, the node corresponding to the phrase “was present” is included in the additional range.

  In addition, as nodes from the node corresponding to the phrase “existing” to the terminal node, there are two nodes corresponding to the phrases “in the economic platform” and “heavy”. In this example, since the co-occurrence probability between nodes corresponding to the phrases “in the economic forum” and “was present” is less than the threshold, this node is not included in the additional range. On the other hand, in this example, since the co-occurrence probability between nodes corresponding to the phrases “heavy” and “was present” is equal to or greater than the threshold, this node is included in the additional range. As a result, as shown in FIG. 14C, the phrase “it was heavy” (underlined portion) is added to the summary text.

On the other hand, when the target of the designation operation is a summary text, the range control unit 108 determines the deletion range based on the syntax analysis result as shown in FIG. At this time, the range control unit 108 starts from the node corresponding to the designated character, extracts nodes on the path from the start point to the end node, and deletes nodes corresponding to the words and phrases already included in the summary text Include in
In addition, the range control unit 108 extracts a node on the path from the node corresponding to the designated character to the root node, and sets the node corresponding to the phrase already included in the summary text among the extracted nodes as candidates. . Then, when the co-occurrence probability between the node corresponding to the designated character and the candidate is equal to or higher than the threshold, the range control unit 108 includes the candidate in the deletion range.

  In the example of FIG. 15B, since the node corresponding to the phrase “zhi” corresponding to the designated character string is itself a terminal node, the deletion range is not further extracted in the terminal direction. On the other hand, of the nodes on the path from the node corresponding to the phrase “zhi” to the root node corresponding to the phrase “departed”, the phrases “middle” and “departed” are included in the summary text. . In this example, since the co-occurrence probability of the nodes corresponding to the phrases “zhi” and “middle” is greater than or equal to the threshold value, the deletion range is a node corresponding to the words “zhi” and “middle”. As a result, as shown in FIG. 15 (C), the phrase “mid-disease” is deleted from the summary text. Note that the threshold value used at the time of addition and the threshold value used at the time of deletion are set to the same value, for example.

  As described above, in the third embodiment, as in the case of the second embodiment, the summary text can be easily edited by performing the designation operation on the original text 101a or the summary text. .

  At the time of addition, words corresponding to a series of nodes connected in the direction of the root node of the syntax tree are added as an additional range, and further words corresponding to a series of nodes having a high co-occurrence probability are set as an additional range in the direction of the end node. . At the time of deletion, words corresponding to a series of nodes connected in the terminal node direction are set as the deletion range, and words corresponding to a series of nodes having a high co-occurrence probability in the root node direction are set as the deletion range. Therefore, a set of words and phrases that are likely to be added or deleted at a time are processed together by a single designation operation, so that the summary text can be edited more easily, contributing to a reduction in the burden of editing work. Furthermore, in order to consider the co-occurrence probability, a set of words and phrases that are particularly likely to be added or deleted in one operation is processed in one designation operation, which contributes to further reducing the burden of editing work.

[3-2. Processing flow]
Next, the flow of processing executed by the information processing apparatus 100 will be described with reference to FIGS. FIG. 16 is a first flowchart illustrating a process flow regarding the operation of the information processing apparatus according to the third embodiment. FIG. 17 is a second flowchart illustrating a processing flow regarding the operation of the information processing apparatus according to the third embodiment. FIG. 18 is a third flowchart illustrating a process flow regarding the operation of the information processing apparatus according to the third embodiment.

  (S201) The original text input unit 102 acquires the original text 101a and stores it in the storage unit 101. For example, the original text input unit 102 stores the original text 101 a input by the user using the input unit 916 in the storage unit 101. When the original text 101 a is in a storage device externally connected to the information processing apparatus 100 or a storage area on the network, the original text input unit 102 acquires the original text from the original text and stores it in the storage unit 101.

  (S202) The morpheme analysis unit 103 performs morpheme analysis on the original text 101a, and stores information (see FIG. 4) in which each morpheme extracted from the original text 101a has a part of speech as a part of the analysis result 101b. 101. For example, if the original text 101a is a sentence "A hot day at the end of August last year, a single expert who was heavy on the economic platform passed away in the middle", the morphological analysis results shown in Fig. 4 were obtained. It is done.

  (S203) The syntax analysis unit 104 performs syntax analysis (dependency analysis) of the source text 101a based on the morpheme analysis result output by the morpheme analysis unit 103. Parsing is a method of analyzing the structure of a sentence in units of phrases and clauses according to grammatical rules. The syntax analysis unit 104 associates information (see FIG. 5) such as a phrase notation, a dependency destination, and a dependency type for each phrase obtained by the syntax analysis, and stores the information in the storage unit 101 as a part of the analysis result 101b. To do.

  (S204) The syntax analysis unit 104 generates correspondence data (see FIG. 7) that associates each character included in the original text 101a with a node corresponding to a word including the character. For example, as shown in FIG. 7, the syntax analysis unit 104 associates the original character number of each character included in the original text 101a, the character notation, and the node number of the node corresponding to the character. Is generated as a part of the analysis result 101b.

  (S205) The syntax analysis unit 104 turns OFF all the deletion flags associated with each word / phrase included in the syntax analysis result (see FIG. 5) (initialization). The deletion flag is set to OFF for words included in the summary sentence text and ON for words not included in the summary sentence text.

  (S206) The summary sentence generation unit 105 sets the deletion flag corresponding to the node to be deleted when generating the summary sentence to ON. For example, the summary sentence generation unit 105 identifies a phrase corresponding to a node on a path from the root node to a node set for summary sentence generation, and turns on a deletion flag corresponding to a node other than the identified node. Set to.

(S207) The summary sentence generation unit 105 generates a summary sentence text by concatenating words and phrases corresponding to the node with the deletion flag OFF in the order of appearance on the original text 101a.
(S208) The summary sentence generation unit 105 generates correspondence data (see FIG. 8) for associating each character included in the summary sentence text with a node corresponding to the phrase including the character. For example, as illustrated in FIG. 8, the summary sentence generation unit 105 associates the summary sentence character number of each character included in the summary sentence text, the character notation, and the node number of the node corresponding to the character. Corresponding data is generated as a part of the analysis result 101b.

  (S209) The sentence output unit 106 acquires the original text 101a from the storage unit 101, and outputs the original text 101a together with the summary text generated by the summary sentence generation unit 105. At this time, the sentence output unit 106 displays the original text 101a and the summary text in a format that allows the user to specify the characters on the original text 101a and the characters on the summary text.

  (S210) The designation receiving unit 107 determines whether the summary sentence text output from the sentence output unit 106 has been confirmed. For example, the designation receiving unit 107 determines whether or not a summary sentence text editing end operation has been performed. When the summary sentence text is confirmed, the series of processing shown in FIGS. 16 and 17 ends. On the other hand, if the summary text has not been confirmed, the process proceeds to S211.

  (S211) The designation receiving unit 107 determines whether a character of the original text 101a is designated. If the character of the original text 101a is designated, the process proceeds to S212. On the other hand, when the character of the original text 101a is not designated, the process proceeds to S217.

  (S212) Upon receiving a designation operation for the characters of the original text 101a, the designation receiving unit 107 refers to correspondence data (see FIG. 7) between the original text and the syntax analysis result, and identifies a node corresponding to the designated character. Then, the designation receiving unit 107 sets the identified node as the examination target node.

  (S213) The range control unit 108 executes the process of subroutine S (see FIG. 19). Subroutine S [...] is a processing unit with "..." as an argument. In the process of S213, the range control unit 108 executes the process of the subroutine S by specifying the examination target node set in S212 as an argument. The processing of the subroutine S will be described later. An additional starting node is set in the subroutine S, which will also be described later.

  (S214, S215, S216) The process of S215 is executed for each additional starting node for all the additional starting nodes set in the process of S213. In the processing of S215, the range control unit 108 sets the deletion flag corresponding to each node from the additional origin node to the root node to OFF based on the syntax analysis result. When the process of S215 is completed for all the additional starting nodes, the process proceeds to S207.

  (S217) The designation receiving unit 107 determines whether characters of the summary text are designated. If characters of the summary text are designated, the process proceeds to S218. On the other hand, when the characters of the summary text are not designated, the process proceeds to S210.

  (S218) Upon receiving the designation operation for the characters of the summary text, the designation receiving unit 107 refers to the correspondence data (see FIG. 8) of the summary text and the syntax analysis result, and identifies the node corresponding to the designated character. Then, the designation receiving unit 107 sets the identified node as the deletion start node.

  (S219) The range control unit 108 determines whether or not the deletion start node is a root node. If the deletion start node is the root node, the process proceeds to S224. On the other hand, if the deletion origin node is not the root node, the process proceeds to S220.

(S220) The range control unit 108 acquires, from the co-occurrence probability table, a co-occurrence probability of a node (connected node) that is connected in the direction from the deletion start node to the root node.
(S221) The range control unit 108 determines whether or not the co-occurrence probability acquired in S220 is equal to or greater than a preset threshold value. If the co-occurrence probability is greater than or equal to the threshold, the process proceeds to S222. On the other hand, if the co-occurrence probability is less than the threshold, the process proceeds to S224.

  (S222) The range control unit 108 determines whether or not the connected node has another node in which the deletion flag connected in the direction toward the end node is OFF. If the connected node has the other node, the process proceeds to S224. On the other hand, if the connected node does not have the other node, the process proceeds to S223.

(S223) The range control unit 108 sets a connection node as a deletion start node. When the process of S223 is completed, the process proceeds to S219.
(S224) The range control unit 108 sets the deletion flag corresponding to each node from the deletion start node to the end node to ON. When the process of S224 is completed, the process proceeds to S207.

(Subroutine S)
Here, the processing of the subroutine S will be described with reference to FIG. FIG. 19 is a fourth flowchart illustrating a process flow regarding the operation of the information processing apparatus according to the third embodiment. Note that a node designated as an argument of the subroutine S is an “input node”.

  (S251) The range control unit 108 determines whether or not there is a node (connected node) connected in the direction from the input node toward the terminal node. If there is a connected node, the process proceeds to S252. On the other hand, if there is no connected node, the series of processes shown in FIG. 19 ends.

  (S252, S256) The range control unit 108 executes the processing from S252 to S256 for each connected node for all connected nodes connected in the direction from the input node to the end node. That is, the range control unit 108 repeatedly executes the processing from S253 to S255 while changing the connection node to be processed. When the process is completed for all connected nodes, the process proceeds to S257.

  (S253, S254) The range control unit 108 acquires the co-occurrence probability between the input node and the connected node from the co-occurrence probability table, and determines whether the co-occurrence probability is equal to or higher than a preset threshold value. If the co-occurrence probability is greater than or equal to the threshold, the process proceeds to S255. On the other hand, if the co-occurrence probability is less than the threshold, the process proceeds to S256.

  (S255) The range control unit 108 executes the process of the subroutine S with the connected node as an argument. That is, the range control unit 108 executes a series of processes shown in FIG. 19 with the currently processed connection node as an input node.

  (S257) The range control unit 108 determines whether or not the co-occurrence probability is lower than a threshold value for all connected nodes. That is, the range control unit 108 determines whether or not the process of S255 has been executed for all the connected nodes. If the co-occurrence probability is lower than the threshold value for all connected nodes, the process proceeds to S258. On the other hand, when there is a connected node whose co-occurrence probability is higher than the threshold value, the series of processes illustrated in FIG. 19 ends.

(S258) The range control unit 108 sets the input node as an additional starting node. When the process of S258 is completed, the series of processes shown in FIG. 19 ends.
The flow of processing executed by the information processing apparatus 100 has been described above.

According to the above processing method, the summary sentence text can be easily edited by performing the designation operation on the original sentence text 101a or the summary sentence text.
At the time of addition, words corresponding to a series of nodes connected in the direction of the root node of the syntax tree are set as an additional range, and further, words corresponding to a series of nodes having a high co-occurrence probability in the direction of the terminal node are set as an additional range. . At the time of deletion, words corresponding to a series of nodes connected in the terminal node direction are set as the deletion range, and words corresponding to a series of nodes having a high co-occurrence probability in the root node direction are set as the deletion range. Therefore, a set of words and phrases that are likely to be added or deleted at a time are processed together by a single designation operation, so that the summary text can be edited more easily, contributing to a reduction in the burden of editing work. Furthermore, in order to consider the co-occurrence probability, a set of words and phrases that are particularly likely to be added or deleted in one operation is processed in one designation operation, which contributes to further reducing the burden of editing work.

The third embodiment has been described above.
<4. Fourth Embodiment>
Next, a fourth embodiment will be described. However, detailed description of the same parts as those described in the second or third embodiment will be omitted. The technique of the fourth embodiment can be used in combination with the second or third embodiment.

[4-1. function]
When the information processing apparatus 100 according to the fourth embodiment adds a phrase to the summary text, the information processing apparatus 100 selects a node satisfying a specific condition from nodes on the path from the node corresponding to the designated character to the root node. to add. That is, the information processing apparatus 100 performs a shortcut to the connection relationship between nodes located at distant positions on the syntax tree.

  Hereinafter, the shortcut will be described with reference to FIGS. 20 and 21. FIG. FIG. 20 is a diagram illustrating an example of a syntax analysis (dependency analysis) result according to the fourth embodiment. FIG. 21 is a diagram illustrating an example of a specifying operation and processing for adding a phrase according to the fourth embodiment.

  As shown in FIG. 20, the shortcut information added to the syntax analysis result is used as the shortcut. In the example of FIG. 20, the node with the node number “1” is shortcut to the node with the node number “13”. For example, when the phrase “last year” corresponding to the node number “1” is specified (see FIG. 21A), the node with the node number “13” (the phrase “departure” among the nodes on the path leading to the root node). Nodes up to “)” are skipped and excluded from addition targets. Therefore, only “Last Year” is added to the summary text as shown in FIG.

The shortcut destination is determined based on the following equation (2). D (i, j) represents the distance between the nodes i and j. P (i, j) represents the co-occurrence probability of the nodes i and j. The score Scj of the node j with respect to the node i is defined as D (i, j) × P (i, j). The distance between nodes can be evaluated by the number of branches between the nodes or the number of other nodes between the nodes plus one. The co-occurrence probability is given by the co-occurrence probability table described above. Node _s (i) represents the node number of the node having the maximum score among the nodes on the path from the node i to the root node. L is the number of nodes on the path from the node i to the root node.

  In the example of FIG. 21, the phrase “last year” is specified, and the scores Sc1, Sc2, and Sc3 are referenced for each node on the path leading to the root node. In this example, when Sc1> Sc2, Sc3, the node of the phrase “Leave” corresponding to Sc1 is the shortcut destination. Therefore, the range control unit 108 excludes the words “at the end of August” and “day” from the addition target, and includes the word “last year” not included in the summary text in the additional range. As a result, the summary text shown in FIG. 21C is obtained.

  Thus, according to the function of the information processing apparatus 100 according to the fourth embodiment, the addition target is narrowed down based on the distance between nodes and the co-occurrence probability. If the node corresponding to the specified character is far from the root node, more words will be added to the summary text and the risk of adding unnecessary words will increase, but this risk can be increased by applying the above function. It is reduced. In addition, since the co-occurrence probability is taken into consideration, it is possible to increase the possibility that necessary words remain and unnecessary words / phrases are excluded, and that a set of appropriate words / phrases is added. As a result, the summary text can be edited more easily, which contributes to further reducing the burden of editing work.

[4-2. Processing flow]
Next, the flow of processing executed by the information processing apparatus 100 will be described with reference to FIGS. FIG. 22 is a first flowchart illustrating a process flow regarding the operation of the information processing apparatus according to the fourth embodiment. FIG. 23 is a second flowchart illustrating a process flow regarding the operation of the information processing apparatus according to the fourth embodiment. FIG. 24 is a third flowchart illustrating a process flow of the operation of the information processing apparatus according to the fourth embodiment.

  (S301) The original text input unit 102 acquires the original text 101a and stores it in the storage unit 101. For example, the original text input unit 102 stores the original text 101 a input by the user using the input unit 916 in the storage unit 101. When the original text 101 a is in a storage device externally connected to the information processing apparatus 100 or a storage area on the network, the original text input unit 102 acquires the original text from the original text and stores it in the storage unit 101.

  (S302) The morpheme analysis unit 103 performs morpheme analysis on the original text 101a, and stores information (see FIG. 4) in which the part of speech is added to each morpheme extracted from the original text 101a as a part of the analysis result 101b. 101. For example, if the original text 101a is a sentence "A hot day at the end of August last year, a single expert who was heavy on the economic platform passed away in the middle", the morphological analysis results shown in Fig. 4 were obtained. It is done.

  (S303) The syntax analysis unit 104 performs syntax analysis (dependency analysis) of the original text 101a based on the morpheme analysis result output by the morpheme analysis unit 103. Parsing is a method of analyzing the structure of a sentence in units of phrases and clauses according to grammatical rules. The syntax analysis unit 104 associates information such as phrase notation, dependency destination, dependency type, and the like (see FIG. 20) with each phrase obtained by the syntax analysis, and stores it in the storage unit 101 as a part of the analysis result 101b. To do.

  (S304) The syntax analysis unit 104 generates correspondence data (see FIG. 7) that associates each character included in the original text 101a with a node corresponding to the phrase including the character. For example, as shown in FIG. 7, the syntax analysis unit 104 associates the original character number of each character included in the original text 101a, the character notation, and the node number of the node corresponding to the character. Is generated as a part of the analysis result 101b.

  (S305) The syntax analysis unit 104 turns off all the deletion flags associated with the respective words included in the syntax analysis result (see FIG. 20) (initialization). The deletion flag is set to OFF for words included in the summary sentence text and ON for words not included in the summary sentence text.

  (S306) The summary sentence generation unit 105 sets the deletion flag corresponding to the node to be deleted when generating the summary sentence to ON. For example, the summary sentence generation unit 105 identifies a phrase corresponding to a node on a path from the root node to a node set for summary sentence generation, and turns on a deletion flag corresponding to a node other than the identified node. Set to.

(S307) The summary sentence generation unit 105 generates a summary sentence text by concatenating words and phrases corresponding to the node whose deletion flag is OFF in the order of appearance on the original text 101a.
(S308) The summary sentence generation unit 105 generates correspondence data (see FIG. 8) for associating each character included in the summary sentence text with a node corresponding to the phrase including the character. For example, as illustrated in FIG. 8, the summary sentence generation unit 105 associates the summary sentence character number of each character included in the summary sentence text, the character notation, and the node number of the node corresponding to the character. Corresponding data is generated as a part of the analysis result 101b.

  (S309) The sentence output unit 106 acquires the original text 101a from the storage unit 101, and outputs the original text 101a together with the summary text generated by the summary sentence generation unit 105. At this time, the sentence output unit 106 displays the original text 101a and the summary text in a format that allows the user to specify the characters on the original text 101a and the characters on the summary text.

  (S310) The designation receiving unit 107 determines whether the summary sentence text output by the sentence output unit 106 has been confirmed. For example, the designation receiving unit 107 determines whether or not a summary sentence text editing end operation has been performed. When the summary text is confirmed, the series of processes shown in FIGS. 22 to 24 ends. On the other hand, if the summary text has not been confirmed, the process proceeds to S311.

  (S311) The designation receiving unit 107 determines whether or not a character of the original text 101a is designated. When the character of the original text 101a is designated, the process proceeds to S312. On the other hand, when the character of the original text 101a is not designated, the process proceeds to S320.

  (S312) When the designation receiving unit 107 receives a designation operation on the characters of the original text 101a, the designation receiving unit 107 refers to the correspondence data (see FIG. 7) between the original text and the syntax analysis result, and identifies the node corresponding to the designated character. Then, the designation receiving unit 107 sets the identified node as an additional starting node.

(S313) The range control unit 108 sets a node adjacent to the root node side of the additional origin node as a shortcut determination node. When the process of S313 is completed, the process proceeds to S314.
(S314) The range control unit 108 acquires the co-occurrence probability between the additional origin node and the shortcut determination node from the co-occurrence probability table.

  (S315) The range control unit 108 calculates a score based on the distance between the additional origin node and the shortcut determination node and the co-occurrence probability acquired in S314. For example, if the additional origin node is node i, the shortcut determination node is node j, the distance between nodes i and j is D (i, j), and the co-occurrence probability is P (i, j), the score Scj is D (i , J) × P (i, j). The above distance is a distance on the syntax tree.

  (S316) The range control unit 108 determines whether the deletion flag of the shortcut determination node is ON. If the shortcut flag of the shortcut determination node is ON, the process proceeds to S317. On the other hand, if the deletion flag of the shortcut determination node is OFF, the process proceeds to S318.

  (S317) The range control unit 108 sets a node adjacent to the root node side of the current shortcut determination node as a new shortcut determination node. When the process of S317 is completed, the process proceeds to S314.

(S318) The range control unit 108 sets a shortcut from the additional origin node to the node with the highest score.
(S319) The range control unit 108 sets the deletion flag corresponding to each node from the additional origin node to the root node to OFF while maintaining the deletion flag of the node included in the section to be shortcutted. When the process of S319 is completed, the process proceeds to S307.

  (S320) The designation receiving unit 107 determines whether characters of the summary text are designated. When characters of the summary text are designated, the same processing as that in the second or third embodiment is executed. On the other hand, when the characters of the summary text are not designated, the process proceeds to S310.

  When the same processing as that of the second embodiment is executed, the processing after S115 of FIG. 12 is executed, and when the processing of S116 is completed, the processing proceeds to S307. On the other hand, when the same processing as that of the third embodiment is executed, the processing after S218 in FIG. 18 is executed, and when the processing of S224 is completed, the processing proceeds to S307. Thus, the technique of the fourth embodiment can be combined with the technique of the second or third embodiment.

The flow of processing executed by the information processing apparatus 100 has been described above.
According to the above processing method, the addition target is narrowed down based on the distance between nodes and the co-occurrence probability. If the node corresponding to the specified character is far from the root node, more words will be added to the summary text and the risk of adding unnecessary words will increase, but this risk can be increased by applying the above processing method. Is reduced. In addition, since the co-occurrence probability is taken into consideration, it is possible to increase the possibility that necessary words remain and unnecessary words / phrases are excluded, and that a set of appropriate words / phrases is added. As a result, the summary text can be edited more easily, which contributes to further reducing the burden of editing work.

  The fourth embodiment has been described above.

DESCRIPTION OF SYMBOLS 10 Information processing apparatus 11 Memory | storage part 12 Operation part 13 Display part 20 Syntax tree 21 Additional range 22 Deletion range 31 Original sentence 32 Summary sentence

Claims (6)

A storage unit that stores an original sentence and a syntax tree in which a dependency structure of a phrase based on a syntactic analysis of the original sentence is expressed by a connection relation of nodes corresponding to the phrase;
A display unit that displays the original sentence and a summary sentence that summarizes the original sentence by omitting the word;
When the designation operation for the original sentence is received, the second node connected to the first node corresponding to the word or phrase at the designated location and in the direction toward the root of the syntax tree is specified, and the first and second Add words to the summary sentence that correspond to
When the designation operation for the summary sentence is received, the fourth node connected to the third node corresponding to the word at the designated location and in the direction toward the end of the syntax tree is specified, and the third And an arithmetic unit that deletes the word corresponding to the fourth node from the summary sentence;
An information processing apparatus comprising: The storage unit further stores a co-occurrence probability between nodes connected on the syntax tree,
When the calculation unit receives a designating operation on the original text, the computing unit is connected to the first node and is connected to the first node among nodes in a direction toward the end of the syntax tree. The information processing apparatus according to claim 1, wherein a node having an occurrence probability greater than a set threshold is specified as the second node. When the calculation unit receives a designating operation on the summary sentence, the co-occurrence with the third node among nodes connected to the third node and in a direction toward the root of the syntax tree. The information processing apparatus according to claim 2, wherein a node having a probability greater than the threshold is specified as the fourth node. The storage unit further stores a co-occurrence probability between nodes connected on the syntax tree,
When the calculation unit accepts a designation operation on the original text, the calculation unit is based on an evaluation value that takes a higher value as the distance from the first node on the syntax tree is longer and the co-occurrence probability increases. The information processing apparatus according to claim 1, wherein a node having the highest evaluation value is identified as the second node among nodes connected to the first node and in a direction toward a root of the syntax tree. Computer
From the storage unit, obtain the original sentence and a syntax tree expressing the dependency structure of the phrase based on the syntactic analysis of the original sentence by the connection relation of the nodes corresponding to the phrase;
The display unit displays the original sentence and a summary sentence summarizing the original sentence,
When the designation operation for the original sentence is received, the second node connected to the first node corresponding to the word or phrase at the designated location and in the direction toward the root of the syntax tree is specified, and the first and second Add words to the summary sentence that correspond to
When the designation operation for the summary sentence is received, the fourth node connected to the third node corresponding to the word at the designated location and in the direction toward the end of the syntax tree is specified, and the third And a summary sentence editing method of deleting words corresponding to the fourth node from the summary sentence. On the computer,
From the storage unit, obtain the original sentence and a syntax tree expressing the dependency structure of the phrase based on the syntactic analysis of the original sentence by the connection relation of the nodes corresponding to the phrase;
The display unit displays the original sentence and a summary sentence summarizing the original sentence,
When the designation operation for the original sentence is received, the second node connected to the first node corresponding to the word or phrase at the designated location and in the direction toward the root of the syntax tree is specified, and the first and second Add words to the summary sentence that correspond to
When the designation operation for the summary sentence is received, the fourth node connected to the third node corresponding to the word at the designated location and in the direction toward the end of the syntax tree is specified, and the third And the program which performs the process which deletes the phrase corresponding to a 4th node from the said summary sentence.

Images