CN107330120B - Inquire answer method, inquiry answering device and computer readable storage medium - Google Patents
Inquire answer method, inquiry answering device and computer readable storage medium Download PDFInfo
- Publication number
- CN107330120B CN107330120B CN201710575024.3A CN201710575024A CN107330120B CN 107330120 B CN107330120 B CN 107330120B CN 201710575024 A CN201710575024 A CN 201710575024A CN 107330120 B CN107330120 B CN 107330120B
- Authority
- CN
- China
- Prior art keywords
- inquiry
- correlation
- degree
- inquiry message
- user
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/36—Creation of semantic tools, e.g. ontology or thesauri
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/332—Query formulation
- G06F16/3329—Natural language query formulation or dialogue systems
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Mathematical Physics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Human Computer Interaction (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
A kind of inquiry answer method of present invention offer, inquiry answering device and computer readable storage medium, the inquiry answer method include：Semantic processes step (S101) carries out semantic processes, with the user view of the inquiry purpose of reaction of formation inquiry message and for carrying out retrieving retrieval information used according to inquiry message to inquiry message input by user；Searching step (S102) is based on the retrieval information, and the data retrieval based on participle is carried out from database, obtains the list of candidate solid data；Sequence step (S103) is ranked up processing based on the degree of correlation between candidate solid data and user view to candidate solid data；And first result determine step (S104), will the candidate solid data with the highest degree of correlation in list, be determined as the response result for user's inquiry message.
Description
Technical field
The present invention relates to inquiry answer method, inquiry answering device and computer-readable storage mediums based on semantic understanding
Matter.
Background technology
Currently, fuzzy semantics understanding is a very universal problem in information retrieval and semantic analysis, if cannot be very
Good carries out it semantic identification, and the result of return may not be very much the result that user wants greatly.Voice input is just becoming more
Carry out more universal interactive mode, although having benefited from the promotion of computing capability and the accumulation of mass data, the use of deep learning is big
Width reduces identification error rate, but still has the error rate of 4%-5%, is particularly acute in the field that certain neologisms occur frequently, this is just
So that fuzzy semantics understanding seems critically important.Still further aspect, due to information huge explosion, man memory power is limited, when very much
Waiting possibly can not accurately say whole information, this is but also fuzzy semantics understand a necessary part as system.
In view of the above-mentioned problems, the Chinese patent application that application publication number is CN106294875A proposes a kind of fuzzy inspection of entity
Rope method and system, but this method is relatively simple, does not account for the factor of phonetic error correction etc, it is difficult to solve current Vague language
The problem of reason and good sense solution.
Separately have, it is crucial in a kind of network search procedure of Chinese patent application proposition that application publication number is CN101206673A
The intelligent correction system and method for word.The system is applied on internet platform, establish language model, corresponding dictionary and
Data directory database, calculates sound character error and fuzzy matching calculates morphological pattern error correction, and degree of correlation filtering and sequence are carried out to result,
Obtain immediate several results.This method is to be used for web search, is not applied for taking turns the fuzzy search in dialogue more, cannot
Solve the problem of that the error correction of fuzzy phoneme cannot solve state transition in more wheel dialogues, cannot solving retrieval result, there is no optimal
In the case of error correction, also not to coming to nothing when how to deal with and be defined, also influence of the error correction result to display, such as
Prompt message etc..
Invention content
The present invention is had developed in view of the above problem in the prior art.The present invention is intended to provide one kind can carry out Vague language
The system and method for reason and good sense solution are sent out inaccurate the problems such as user is because of the fuzzy expression of speech intonation, mistake or does not remember clearly
When true instruction, system remains to make correct semantic understanding and smoothly completes information retrieval on this basis.It is suitable for institute
Some need obscures the scene of error correction, includes the fuzzy semantics error correction in the semantic error correction of web search, and more wheel dialogues.
The first aspect of the present invention provides a kind of inquiry answer method based on semantic understanding, the inquiry answer method packet
It includes：Semantic processes step (S101) carries out semantic processes, with the inquiry of reaction of formation inquiry message to inquiry message input by user
Ask the user view of purpose and for carrying out retrieving retrieval information used according to inquiry message；Searching step (S102), is based on
The retrieval information carries out the data retrieval based on participle from database, obtains the list of candidate solid data；Sequence step
(S103), based on the degree of correlation between candidate solid data and user view, processing is ranked up to candidate solid data；And
First result determines step (S104), and the candidate solid data with the highest degree of correlation in list is determined as asking for user
Ask the response result of information.
The second aspect of the present invention provides a kind of inquiry answering device based on semantic understanding, the inquiry answering device packet
It includes：Semantic processing unit (1101) carries out semantic processes, with the inquiry of reaction of formation inquiry message to inquiry message input by user
Ask the user view of purpose and for carrying out retrieving retrieval information used according to inquiry message；Retrieval unit (1102), is based on
The retrieval information carries out the data retrieval based on participle from database, obtains the list of candidate solid data；Sequencing unit
(1103), based on the degree of correlation between candidate solid data and user view, processing is ranked up to candidate solid data；And
Candidate solid data with the highest degree of correlation in list is determined as asking for user by the first result determination unit (1104)
Ask the response result of information.
The third aspect of the present invention provides a kind of inquiry response system (100) based on semantic understanding, the system comprises
User terminal (1001) and the server (1002) being connect with user terminal, the user terminal include：Input receiving unit
(10011), receive inquiry message input by user；Semantic processing unit (10012) carries out semantic processes to inquiry message, with
The user view of the inquiry purpose of reaction of formation inquiry message and for being carried out retrieving retrieval information used according to inquiry message；
Inquiry message, the user view of the inquiry message and retrieval information are sent to by transmission unit (10013) in a manner of associated
Server, and the response result for inquiry message is received from server, the server includes：Receiving unit (10021), from
User terminal receives inquiry message and user view associated with the inquiry message and retrieval information；Retrieval unit (10022),
Based on the retrieval information, the data retrieval based on participle is carried out from database, obtains the list of candidate solid data；Sequence
Unit (10023) is ranked up candidate solid data based on the degree of correlation between candidate solid data and user view；With
And the candidate solid data with the highest degree of correlation in list is determined as inquiring for user by result determination unit (10024)
The response result of information, and response result is sent to user terminal.
The fourth aspect of the present invention provides a kind of computer readable storage medium, stores computer program, the calculating
When being executed by processor, realization includes the steps that machine program according in above-mentioned inquiry answer method.
According to the present invention, even if can be smooth if the inquiry message of input error the problems such as due to user's vagueness in memory
Completion information retrieval so that user can obtain and the closer retrieval result of the intention of user.In addition, even if there is no
In the case of optimal retrieval result, error correction can be also carried out, and provide a user the result after error correction.
Description of the drawings
In order to more clearly explain the technical solutions in the embodiments of the present application, make required in being described below to embodiment
Attached drawing is briefly described, it should be apparent that, the accompanying drawings in the following description is only some implementations described in the application
Example, without creative efforts, can also be according to these attached drawings for this field or those of ordinary skill
Obtain other attached drawings.
Fig. 1 is the figure for the hardware construction for showing the inquiry answering device in the present invention.
Fig. 2 is the flow chart for illustrating inquiry answer method according to a first embodiment of the present invention.
Fig. 3 is the flow chart for the semantic processes step for illustrating inquiry answer method according to the present invention.
Fig. 4 is the flow chart for the sequence step for illustrating inquiry answer method according to the present invention.
Fig. 5 is the block diagram for the software configuration for illustrating inquiry answering device according to first embodiment.
Fig. 6 is the flow chart for illustrating inquiry answer method according to a second embodiment of the present invention.
Fig. 7 is the block diagram for the software configuration for illustrating inquiry answering device according to second embodiment.
Fig. 8 is the flow chart for illustrating inquiry answer method according to the preferred embodiment of the invention.
Fig. 9 is the block diagram for the software configuration for illustrating the inquiry answering device according to preferred embodiment.
Figure 10 is the schematic diagram for illustrating the inquiry response system of the present invention.
Specific implementation mode
Hereinafter describe the embodiment of the present invention in detail with reference to the accompanying drawings.It should be appreciated that following embodiments and unawareness
The figure limitation present invention, also, about the means according to the present invention solved the problems, such as, it is not absolutely required to be retouched according to following embodiments
The whole combinations for the various aspects stated.For simplicity, to identical structure division or step, identical label or mark have been used
Number, and the description thereof will be omitted.
[hardware configuration of inquiry answering device]
Fig. 1 is the figure for the hardware construction for showing the inquiry answering device in the present invention.In the present embodiment, with smart phone
Example as inquiry answering device provides description.Although it is noted that illustrating smart phone in the present embodiment as inquiry
Ask answering device 1000, but it is clear that without being limited thereto, inquiry answering device of the invention can be mobile terminal (smart mobile phone,
Smartwatch, Intelligent bracelet, music player devices), laptop, tablet computer, PDA (personal digital assistant), fax dress
Set, printer or be with inquiry answering internet device (such as digital camera, refrigerator, television set etc.)
Etc. various devices.
First, the hardware configuration of block diagram description inquiry answering device 1000 (2000,3000) referring to Fig.1.In addition, at this
Following construction is described as example in embodiment, but the inquiry answering device of the present invention is not limited to construction shown in FIG. 1.
Inquire that answering device 1000 includes input interface 101, CPU 102, the ROM being connected to each other via system bus
103, RAM 105, storage device 106, output interface 104, communication unit 107 and short-range wireless communication unit 108 and display
Unit 109.Input interface 101 is the interface executed instruction for receiving data and function that user is inputted, and is
For being received from data input by user via the operating unit (not shown) of such as microphone, button, button or touch screen and
The interface of operational order.It note that the display unit 109 being described later on and operating unit can be at least partly integrated, also,
For example, it may be carrying out picture output in same picture and receiving the construction of user's operation.
CPU 102 is system control unit, and generally comprehensively answering device 1000 is inquired in control.In addition, for example,
CPU 102 carries out the display control of the display unit 109 of inquiry answering device 1000.ROM 103 stores all of the execution of CPU 102
Such as fixed data of tables of data and control program and operating system (OS) program.In the present embodiment, it is stored in ROM 103
Each control program carry out at such as scheduling, task switching and interruption for example, under the management of the OS stored in ROM 103
The software of reason etc. executes control.
The constructions such as SRAM (static RAM), DRAM by needing stand-by power supply of RAM 105.This
In the case of, RAM 105 can store the significant data of control variable of program etc. in a non-volatile manner.In addition, RAM 105
Working storage as CPU 102 and main memory.
Storage device 106 stores model trained in advance (for example, word error correction mode, physical model, Rank models, semanteme
Model etc.), the database for being retrieved and for execute it is according to the present invention inquiry answer method application program etc..
It note that database here can also be stored in the external device (ED) of such as server.In addition, storage device 106 store it is all
It is such as used to carry out the information for sending/receiving transmission/receiving control program via communication unit 107 and communication device (not shown)
Various programs and the various information that use of these programs.In addition, storage device 106 also stores inquiry answering device 1000
Setting information, inquire answering device 1000 management data etc..
Output interface 104 is for being controlled display unit 109 to show the display picture of information and application program
The interface in face.Display unit 109 is for example constructed by LCD (liquid crystal display).Have such as by being arranged on display unit 109
The soft keyboard of the key of numerical value enter key, mode setting button, decision key, cancel key and power key etc. can receive single via display
The input from the user of member 109.
Inquire answering device 100 via communication unit 107 for example, by wireless communications such as Wi-Fi (Wireless Fidelity) or bluetooth
Method holds row data communication with external device (ED) (not shown).
In addition, inquiry answering device 1000 can also via short-range wireless communication unit 108, in short-range with
External device (ED) etc. is wirelessly connected and holds row data communication.And short-range wireless communication unit 108 by with communication unit
107 different communication means are communicated.It is, for example, possible to use its communication range is shorter than the communication means of communication unit 107
Communication means of the Bluetooth Low Energy (BLE) as short-range wireless communication unit 108.In addition, as short-distance wireless communication list
The communication means of member 108, for example, it is also possible to perceive (Wi-Fi Aware) using NFC (near-field communication) or Wi-Fi.
[first embodiment]
[inquiry answer method according to first embodiment]
Inquiry answer method according to the present invention can be by inquiring that the readings of CPU 102 of answering device 1000 are stored in
ROM 103 or control program on storage device 106 or via communication unit 107 from passing through network and inquiry answering device
The network server (not shown) of 1000 connections and the control program downloaded are realized.
Before carrying out inquiry answer method according to the present invention, first preparation model and database are needed.Detailed process is such as
Under：
(1) crawl of related data：The crawl of the crawl and associated data such as label etc. of solid data,
In, solid data refers to the entity in certain field (such as video field), such as film " private savings of husbands ", " the Mi months pass ", and
Label is exactly the word for describing the entity：Such as " social forest ", " love ".
(2) training of model：Word error correcting model：The mapping for establishing the pinyin table and fuzzy phoneme of word, passes through the instruction to language material
Practice, calculates the transition probability model between the probabilistic model of word and word；Physical model：Using language model to including solid data
Language material be trained, it is established that identify the model of entity；Rank models：Pass through ready data and feature extraction, training
At the decision-tree model of GBDT；Semantic model：By language model and training corpus, semantic model can be extracted by being trained to.
Sample needed for model above training process can be crawled from public network.
(3) the index storage of data：To the field modeling, be based on existing data and model, be processed into for retrieval and
The structural data and storage of semantic understanding.
Next, being illustrated to inquiry answer method according to a first embodiment of the present invention in conjunction with Fig. 2 to Fig. 4.Wherein,
Fig. 2 is the flow chart for illustrating inquiry answer method according to a first embodiment of the present invention；Fig. 3 is to illustrate inquiry according to the present invention
The flow chart of the semantic processes step of answer method；Fig. 4 is the sequence step for illustrating inquiry answer method according to the present invention
Flow chart.
As shown in Fig. 2, first, in semantic processes step S101, language is carried out to inquiry message input by user (query)
Justice processing, it is used with the user view of the inquiry purpose of reaction of formation inquiry message and for according to inquiry message retrieve
Retrieve information.Preferably, as shown in figure 3, semantic processes step S101 further comprises：User view identification step S1011 is right
Inquiry message carries out user view identification, obtains the user view corresponding to inquiry message；Entity recognition step S1012, passes through
Trained physical model in advance, identifies solid data from inquiry message；And semantic understanding step S1013, by advance
Trained semantic model carries out semantic understanding, to obtain retrieval information to inquiry message.Here, inquiry message be user for example
Pass through the text message of keyboard input, the text envelope for example generated by the voice messaging of microphone input by converting user
One in text message made of breath and text message input by user and the text combination for being converted into user speech information
Kind.For example, user can input inquiry message " I will see The Shawshank Redemption ", at this point, by Entity recognition step, Ke Yicong
Entity " The Shawshank Redemption " is identified in the inquiry message, and by semantic understanding step, semantic reason is carried out to the inquiry message
Solution, can obtain retrieval information, retrieval information here using the intelligible slot value pair of computer form, such as " title=
The Shawshank Redemption ".
Next, in searching step S102, it is based on the retrieval information, the data based on participle are carried out from database
Retrieval obtains the list of candidate solid data.Here, first by the slot value obtained in semantic understanding step S1013 to conversion
At the sentence that can be retrieved (for example, " title=The Shawshank Redemption " is converted into " film title:The Shawshank Redemption "),
Then retrieval request is sent to return the result list to database.It, can be according to pre-prepd participle mould in retrieving
Type segments the value (such as " The Shawshank Redemption ") in retrieval information, and in the preparation of database, it also can be in library
Each solid data carries out participle and is indexed with falling to sort, and this makes it possible to matched result is found out by the result of participle
Come.The advantages of being retrieved based on participle be, though user due to vagueness in memory and the inquiry message of input error (such as
" cucurbit baby brother "), it is " cucurbit baby " and " brother " to be segmented inquiry message by participle model, can also be examined from database
Rope goes out desired result (such as " Calabash Brothers ")；Alternatively, user may input incomplete inquiry message (such as " Xiao Shenke
Redeem "), it is " Xiao Shenke " and " redeeming " to be segmented inquiry message by participle model, and the phase can be also retrieved from database
The result (such as " The Shawshank Redemption ") of prestige.
Next, in sequence step S103, based on the degree of correlation between candidate solid data and user view, to candidate
Solid data is ranked up processing.Preferably, as shown in figure 4, sequence step S103 further comprises：Relatedness computation step
S1031 calculates the degree of correlation between candidate solid data and user view according to GBDT models；And relevancy ranking step
S1032 is based on the calculated degree of correlation, is ranked up to candidate solid data using Rank models.Here, candidate is being calculated in fact
When the degree of correlation between volume data and user view, first by context state, entity static information (such as label, name,
Classification etc.), the multidate information (such as distance of temperature, marking and current time) of entity calculate characteristic value, then will own
Characteristic value calculate the last degree of correlation by pre-prepd GBDT models.Here, the characteristic value of static information is logical
Information is crossed with the matching degree of inquiry message input by user come what is calculated, this matching degree can pass through phonetic (including fuzzy phoneme)
Editing distance, the editing distance of word, semantic editing distance etc. determine that and the characteristic value of multidate information can be by certain
Formula calculate.
Finally, in the first result determines step S104, by the candidate solid data with the highest degree of correlation in list, really
It is set to the response result for user's inquiry message.Here it is possible to by display unit 109, by the time with the highest degree of correlation
Solid data is selected to return to user as optimal result.
Inquiry answer method according to a first embodiment of the present invention carries out base by being based on retrieval information from database
In the data retrieval of participle, the list of candidate solid data is obtained, and based on the phase between candidate solid data and user view
Guan Du is ranked up processing to candidate solid data, can obtain following technique effect：Even if a. due to user's vagueness in memory or
Input error and input incomplete inquiry message, can also retrieve ideal result；B. it allows users to obtain and use
The closer retrieval result of intention at family.
[software configuration of inquiry answering device according to first embodiment]
Fig. 5 is the block diagram for the software configuration for illustrating inquiry answering device according to first embodiment.As shown in figure 5, inquiry
Answering device 1000 includes that semantic processing unit 1101, retrieval unit 1102, sequencing unit 1103 and the first result determine list
Member 1104.
Specifically, semantic processing unit 1101 includes：User view recognition unit 11011, uses inquiry message
Family intention assessment obtains the user view corresponding to inquiry message；Entity recognition unit 11012 passes through entity trained in advance
Model identifies solid data from inquiry message；And semantic understanding unit 11013, by semantic model trained in advance,
Semantic understanding is carried out to inquiry message, to obtain retrieval information.Retrieval unit 1102 is based on the retrieval information, from database
The data retrieval based on participle is carried out, the list of candidate solid data is obtained.Sequencing unit 1103 includes：Correlation calculating unit
11031, the degree of correlation between candidate solid data and user view is calculated according to GBDT models；And relevancy ranking unit
11032, it is based on the calculated degree of correlation, candidate solid data is ranked up.First result determination unit 1104, will be in list
Candidate solid data with the highest degree of correlation is determined as the response result for user's inquiry message.
[second embodiment]
[inquiry answer method according to second embodiment]
Inquiry answer method according to a second embodiment of the present invention is illustrated with reference to Fig. 6.Wherein, Fig. 6 is example
Show the flow chart of inquiry answer method according to a second embodiment of the present invention.
As shown in fig. 6, inquiry answer method according to second embodiment and the inquiry answer method according to first embodiment
Difference lies in increase the first judgment step S204, the second judgment step S205 and the second result and determine step S206.
Specifically, in the first judgment step S204, according to similarity distance, the list obtained in step s 103 is calculated
In first degree of correlation between candidate solid data and inquiry message with the highest degree of correlation, and whether judge first degree of correlation
Less than first threshold.Here, the different attribute of solid data is equivalent to the different slots of semantic understanding, and the inquiry of attribute and user
Ask that the degree of correlation of information is determined by similarity distance, similarity distance here include phonetic (including fuzzy phoneme) editor away from
The editing distance etc. of editing distance and semanteme from, word, wherein the editing distance of word is for example since font is close, unisonance is different
Word lacks situations such as word multiword and generates.If first degree of correlation is less than first threshold (being "Yes" in step S204), then it represents that from
Differing greatly between the desired result of optimal result and user retrieved in database, at this moment, processing proceed to the second knot
Fruit determines step S206, and the solid data identified in Entity recognition step S1012, which is determined as response result, to be returned to
User so that even if in the database it is not anticipated that ideal result can be obtained in the case of result if user.For example, with
Family inputs inquiry message " I will see dear Interpreter Officer " in step S101, and in the database without the film, at this moment
The entity " dear Interpreter Officer " identified in Entity recognition step S1012 can be returned to user.
On the other hand, if first degree of correlation is greater than or equal to first threshold (in step S204 be "No"), handle into
Row is to the second judgment step S205, to judge whether first degree of correlation is more than second threshold.If first degree of correlation is more than second
Threshold value (being "Yes" in step S205), then it represents that the optimal result retrieved from database is consistent with the desired result of user,
And it handles and proceeds to step S104, by the optimal result, be determined as the response result for user's inquiry message.So that
User can obtain satisfied response result.
On the other hand, if first degree of correlation is no more than second threshold (being "No" in step S205), then it represents that from data
There are still differences between the desired result of optimal result and user retrieved in library, and at this moment, processing proceeds to step S206, with
The solid data identified in Entity recognition step S1012 is determined as response result and returns to user.
It note that the above first threshold and second threshold are advance in the performance of training, verification and test according to model
Determining, to ensure in the performance recalled with had in accuracy rate.
In addition, in above-mentioned second judgment step S205, if the candidate solid data with the highest degree of correlation in list
First degree of correlation between inquiry message is not more than second threshold, can also further judge there is the second high correlation in list
Whether the degree of correlation between the candidate solid data and inquiry message of degree is more than second threshold, and the case where being judged as "Yes"
Under, proceed to step S104.It can avoid leading to miss optimal response knot due to sequencing errors in step s 103 in this way
Fruit.In the case where not appreciably affecting processing speed, can in step S205 successively the preceding N in calculations list (for example, N=
3) degree of correlation between the candidate solid data and inquiry message of position.
Inquiry answer method according to a second embodiment of the present invention, by calculating the phase between optimal result and inquiry message
Guan Du carrys out the response result that certainly directional user returns, can obtain following technique effect：Even if making in the database without pre-
In the case of phase result, user can also obtain ideal result.
[software configuration of inquiry answering device according to second embodiment]
Fig. 7 is the block diagram for the software configuration for illustrating inquiry answering device according to second embodiment.As shown in fig. 7, according to
Difference lies in increase the inquiry answering device 2000 of second embodiment with inquiry answering device 1000 according to first embodiment
First judging unit 1204, the second result determination unit 1206 and second judgment unit 1205.
Specifically, the first judging unit is according to the candidate entity number with the highest degree of correlation in similarity distance calculations list
According to first degree of correlation between inquiry message, and judge whether first degree of correlation is less than first threshold.Second result determines single
Member identifies the Entity recognition unit in the case where first judging unit judges that first degree of correlation is less than first threshold
The solid data gone out, is determined as response result.Second judgment unit, judges whether first degree of correlation is more than second threshold, wherein
In the case where the second judgment unit judges that first degree of correlation is more than second threshold, the first result determination unit will have
The candidate solid data for having the highest degree of correlation, is determined as response result, and wherein, and the similarity distance includes the editor of phonetic
At least one of the editing distance of distance, the editing distance of word and semanteme.
[preferred embodiment]
[according to the inquiry answer method of preferred embodiment]
Inquiry answer method according to the preferred embodiment of the invention is illustrated with reference to Fig. 8.Fig. 8 is to illustrate basis
The flow chart of the inquiry answer method of the preferred embodiment of the present invention.
As shown in figure 8, according to the inquiry answer method of preferred embodiment and inquiry answer method according to first embodiment
Difference lies in increase pretreatment and error correction step S301.
Specifically, in pretreatment and error correction step S301, inquiry message is pre-processed, and by instructing in advance
Experienced word error correcting model, to carrying out correction process by pretreated inquiry message.Here, the pretreatment includes believing inquiry
The deletion of the stop words and spoken word that include in breath and the capital and small letter conversion of the letter and number to including in inquiry message
Deng.For example, when in inquiry message input by user include some colloquial words when, carry out semantic processes step S101 it
Before, it needs to remove these colloquial words.For example, in the feelings that inquiry message input by user is " I will see dear diplomat "
Under condition, colloquial word " I will see " can be deleted by pretreatment first.It then, will be pre- by word error correcting model trained in advance
Treated, and inquiry message " dear diplomat " is corrected as " dear Interpreter Officer ".Next, to by pretreatment and error correction
Treated, and inquiry message carries out subsequent processing.In addition, user is also possible to due to pronunciation mistake and the inquiry message of input error,
Such as it in the case where inquiry message input by user is " Xiao Shengke's redeems ", is entangled by the fuzzy phoneme in word correction process
It is wrong, additionally it is possible to be corrected as " The Shawshank Redemption ".
According to the inquiry answer method of preferred embodiment by being carried out before carrying out semantic processes at pretreatment and error correction
Reason, can be corrected inquiry message input by user, to improve the accuracy of later retrieval.
[according to the software configuration of the inquiry answering device of preferred embodiment]
Fig. 9 is the block diagram for the software configuration for illustrating the inquiry answering device according to preferred embodiment.As shown in figure 9, according to
Difference lies in increase the inquiry answering device 3000 of preferred embodiment with inquiry answering device 1000 according to first embodiment
Pretreatment and error correction unit 1301.
Specifically, pretreatment and error correction unit 1301 pre-process inquiry message, and pass through training in advance
Word error correcting model, to carrying out correction process by pretreated inquiry message.
In addition, the present invention also provides a kind of inquiry response systems based on semantic understanding.Figure 10 is to illustrate the present invention
The schematic diagram of inquiry response system.As shown in Figure 10, inquiry response system 100 includes user terminal 1001 and server 1002,
User terminal 1001 is connect with server 1002 via network 1003, and network 1003 can be cable network or wireless network.
User terminal 1001 includes input receiving unit 10011, semantic processing unit 10012 and transmission unit 10013.Clothes
Business device 1002 includes receiving unit 10021, retrieval unit 10022, sequencing unit 10023 and result determination unit 10024.
Specifically, in user terminal 1001, input receiving unit 10011 receives inquiry message input by user；Language
Adopted processing unit 10012 carries out semantic processes to inquiry message, with the user view of the inquiry purpose of reaction of formation inquiry message
With for being carried out retrieving retrieval information used according to inquiry message；Transmission unit 10013 is by inquiry message, the inquiry message
User view and retrieval information are sent to server in a manner of associated, and receive the response for inquiry message from server
As a result.
On the other hand, in server 1002, receiving unit 10021 from user terminal receive inquiry message and with the inquiry
The associated user view of information and retrieval information；Retrieval unit 10022 is based on the retrieval information, and base is carried out from database
In the data retrieval of participle, the list of candidate solid data is obtained；Sequencing unit 10023 is based on candidate solid data and anticipates with user
The degree of correlation between figure is ranked up candidate solid data；As a result determination unit 10024 will have the highest degree of correlation in list
Candidate solid data, be determined as the response result for user's inquiry message, and response result is sent to user terminal.
Although exemplary embodiments describe the present invention for reference above, above-described embodiment is only to illustrate this hair
Bright technical concepts and features, it is not intended to limit the scope of the present invention.It is all to be done according to spirit of the invention
Any equivalent variations or modification, should be covered by the protection scope of the present invention.
Claims (16)
1. a kind of inquiry answer method based on semantic understanding, the inquiry answer method include：
Semantic processes step (S101) carries out semantic processes, with reaction of formation inquiry message to inquiry message input by user
It inquires the user view of purpose and for being carried out retrieving retrieval information used according to inquiry message, and is identified from inquiry message
Go out solid data；
Searching step (S102) is based on the retrieval information, and the data retrieval based on participle is carried out from database, obtains candidate
The list of solid data；
Sequence step (S103) carries out candidate solid data based on the degree of correlation between candidate solid data and user view
Sequence is handled；
First judgment step (S204), according to the candidate solid data in similarity distance calculations list with the highest degree of correlation and inquiry
It asks first degree of correlation between information, and judges whether first degree of correlation is less than first threshold；
Second result determines step (S206), judges that first degree of correlation is less than the feelings of first threshold in first judgment step
Under condition, the solid data that will be identified in the semantic processes step is determined as response result；
Second judgment step (S205) judges the case where first degree of correlation is not less than first threshold in first judgment step
Under, judge whether first degree of correlation is more than second threshold；And
First result determines step (S104), judges that first degree of correlation is more than the feelings of second threshold in second judgment step
Under condition, by the candidate solid data with the highest degree of correlation in list, it is determined as the response result for user's inquiry message, and
It, will be in the semantic processes step in the case of judging that first degree of correlation is not more than second threshold in second judgment step
In the solid data that identifies, be determined as response result.
2. inquiry answer method according to claim 1, wherein the semantic processes step (S101) includes：
User view identification step (S1011) carries out user view identification to inquiry message, obtains the use corresponding to inquiry message
Family is intended to；
Entity recognition step (S1012) identifies solid data by physical model trained in advance from inquiry message；With
And
Semantic understanding step (S1013) carries out semantic understanding by semantic model trained in advance to inquiry message, to obtain
Retrieve information.
3. inquiry answer method according to claim 1,
Wherein, the similarity distance include the editing distance of phonetic, word editing distance and semanteme editing distance at least
One.
4. inquiry answer method according to any one of claim 1 to 3, wherein the sequence step (S103) includes：
Relatedness computation step (S1031) calculates the degree of correlation between candidate solid data and user view according to GBDT models；
And
Relevancy ranking step (S1032) is based on the calculated degree of correlation, is ranked up to candidate solid data.
5. inquiry answer method according to any one of claim 1 to 3, the inquiry answer method is at the semanteme
Further include before reason step (S101)：
Pretreatment and error correction step (S301), pre-process inquiry message, and the word error correcting model by training in advance,
To carrying out correction process by pretreated inquiry message.
6. inquiry answer method according to claim 5, the pretreatment includes the stop words to including in inquiry message
The capital and small letter of deletion with spoken word and the letter and number to including in inquiry message is converted.
7. inquiry answer method according to any one of claim 1 to 3, the inquiry message is text input by user
Information, the text message generated by converting voice messaging input by user and text message input by user with will use
One kind in text message made of the text combination that family voice messaging is converted into.
8. a kind of inquiry answering device based on semantic understanding, the inquiry answering device include：
Semantic processing unit (1101) carries out semantic processes, with reaction of formation inquiry message to inquiry message input by user
It inquires the user view of purpose and for being carried out retrieving retrieval information used according to inquiry message, and is identified from inquiry message
Go out solid data；
Retrieval unit (1102) is based on the retrieval information, and the data retrieval based on participle is carried out from database, obtains candidate
The list of solid data；
Sequencing unit (1103) carries out candidate solid data based on the degree of correlation between candidate solid data and user view
Sequence is handled；
First judging unit (1204), according to the candidate solid data in similarity distance calculations list with the highest degree of correlation and inquiry
It asks first degree of correlation between information, and judges whether first degree of correlation is less than first threshold；
Second result determination unit (1206) judges the case where first degree of correlation is less than first threshold in first judging unit
Under, the solid data that the semantic processing unit is identified is determined as response result；
Second judgment unit (1205) judges the case where first degree of correlation is not less than first threshold in first judging unit
Under, judge whether first degree of correlation is more than second threshold；And
First result determination unit (1104) judges the case where first degree of correlation is more than second threshold in the second judgment unit
Under, by the candidate solid data with the highest degree of correlation in list, it is determined as the response result for user's inquiry message；And
The second judgment unit judges that no more than in the case of second threshold, the semantic processing unit is identified for first degree of correlation
Solid data, be determined as response result.
9. inquiry answering device according to claim 8, wherein the semantic processing unit includes：
User view recognition unit (11011) carries out user view identification to inquiry message, obtains the use corresponding to inquiry message
Family is intended to；
Entity recognition unit (11012) identifies solid data by physical model trained in advance from inquiry message；With
And
Semantic understanding unit (11013) carries out semantic understanding by semantic model trained in advance to inquiry message, to obtain
Retrieve information.
10. inquiry answering device according to claim 8, wherein the similarity distance includes the editing distance of phonetic, word
Editing distance and semantic at least one of editing distance.
11. the inquiry answering device according to any one of claim 8 to 10, wherein the sequencing unit includes：
Correlation calculating unit (11031) calculates the degree of correlation between candidate solid data and user view according to GBDT models；
And
Relevancy ranking unit (11032) is based on the calculated degree of correlation, is ranked up to candidate solid data.
12. the inquiry answering device according to any one of claim 8 to 10, the inquiry answering device further include：
Pretreatment and error correction unit (1301), pre-process inquiry message, and the word error correcting model by training in advance,
To carrying out correction process by pretreated inquiry message.
13. inquiry answering device according to claim 12, the pretreatment includes being deactivated to include in inquiry message
The capital and small letter of the deletion of word and spoken word and the letter and number to including in inquiry message is converted.
14. the inquiry answering device according to any one of claim 8 to 10, the inquiry message is text input by user
This information, the text message generated by converting voice messaging input by user and text message input by user and general
One kind in text message made of the text combination that user speech information is converted into.
15. a kind of inquiry response system (100) based on semantic understanding, the system comprises user terminal (1001) and and users
The server (1002) of terminal connection,
The user terminal includes：
Receiving unit (10011) is inputted, inquiry message input by user is received；
Semantic processing unit (10012) carries out semantic processes, with the inquiry purpose of reaction of formation inquiry message to inquiry message
User view and for being carried out retrieving retrieval information used according to inquiry message, and entity number is identified from inquiry message
According to；
Transmission unit (10013) sends out inquiry message, the user view of the inquiry message and retrieval information in a manner of associated
Server is given, and the response result for inquiry message is received from server,
The server includes：
Receiving unit (10021) receives inquiry message and user view associated with the inquiry message and inspection from user terminal
Rope information；
Retrieval unit (10022) is based on the retrieval information, the data retrieval based on participle is carried out from database, is waited
Select the list of solid data；
Sequencing unit (10023) carries out candidate solid data based on the degree of correlation between candidate solid data and user view
Sequence；
First judging unit (1204), according to the candidate solid data in similarity distance calculations list with the highest degree of correlation and inquiry
It asks first degree of correlation between information, and judges whether first degree of correlation is less than first threshold；
Second result determination unit (1206) judges the case where first degree of correlation is less than first threshold in first judging unit
Under, the solid data that the semantic processing unit is identified is determined as response result, and response result is sent to user's end
End；
Second judgment unit judges in the case where first judging unit judges that first degree of correlation is not less than first threshold
Whether first degree of correlation is more than second threshold；And
As a result determination unit (10024), in the case where the second judgment unit judges that first degree of correlation is more than second threshold,
By the candidate solid data with the highest degree of correlation in list, it is determined as the response result for user's inquiry message, and will answer
It answers result and is sent to user terminal；And judge the case where first degree of correlation is not more than second threshold in the second judgment unit
Under, the solid data that the semantic processing unit is identified is determined as response result, and response result is sent to user's end
End.
16. a kind of computer readable storage medium, stores computer program, the computer program is being executed by processor
When, it realizes in inquiry answer method according to any one of claim 1 to 7 and includes the steps that.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710575024.3A CN107330120B (en) | 2017-07-14 | 2017-07-14 | Inquire answer method, inquiry answering device and computer readable storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710575024.3A CN107330120B (en) | 2017-07-14 | 2017-07-14 | Inquire answer method, inquiry answering device and computer readable storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107330120A CN107330120A (en) | 2017-11-07 |
| CN107330120B true CN107330120B (en) | 2018-09-18 |
Family
ID=60226783
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710575024.3A Active CN107330120B (en) | 2017-07-14 | 2017-07-14 | Inquire answer method, inquiry answering device and computer readable storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107330120B (en) |
Families Citing this family (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108021554A (en) * | 2017-11-14 | 2018-05-11 | 无锡小天鹅股份有限公司 | Audio recognition method, device and washing machine |
| CN109800407B (en) * | 2017-11-15 | 2021-11-16 | 腾讯科技（深圳）有限公司 | Intention recognition method and device, computer equipment and storage medium |
| KR102517219B1 (en) * | 2017-11-23 | 2023-04-03 | 삼성전자주식회사 | Electronic apparatus and the control method thereof |
| KR102561712B1 (en) * | 2017-12-07 | 2023-08-02 | 삼성전자주식회사 | Apparatus for Voice Recognition and operation method thereof |
| CN110472058B (en) * | 2018-05-09 | 2023-03-03 | 华为技术有限公司 | Entity searching method, related equipment and computer storage medium |
| CN110737756B (en) * | 2018-07-03 | 2023-06-23 | 百度在线网络技术（北京）有限公司 | Method, apparatus, device and medium for determining answer to user input data |
| JP7068962B2 (en) * | 2018-08-13 | 2022-05-17 | 株式会社日立製作所 | Dialogue methods, dialogue systems and programs |
| CN109597993B (en) * | 2018-11-30 | 2021-11-05 | 深圳前海微众银行股份有限公司 | Statement analysis processing method, device, equipment and computer readable storage medium |
| CN110457423A (en) * | 2019-06-24 | 2019-11-15 | 平安科技（深圳）有限公司 | A kind of knowledge mapping entity link method, apparatus, computer equipment and storage medium |
| CN110334347B (en) * | 2019-06-27 | 2024-06-28 | 腾讯科技（深圳）有限公司 | Information processing method based on natural language recognition, related equipment and storage medium |
| CN110456339B (en) * | 2019-08-12 | 2021-09-14 | 四川九洲电器集团有限责任公司 | Inquiring and responding method and device, computer storage medium and electronic equipment |
| CN112396481A (en) * | 2019-08-13 | 2021-02-23 | 北京京东尚科信息技术有限公司 | Offline product information transmission method, system, electronic device, and storage medium |
| CN110647987B (en) * | 2019-08-22 | 2024-09-13 | 腾讯科技（深圳）有限公司 | Method and device for processing data in application program, electronic equipment and storage medium |
| CN110765342A (en) * | 2019-09-12 | 2020-02-07 | 竹间智能科技（上海）有限公司 | Information query method and device, storage medium and intelligent terminal |
| CN111597808B (en) * | 2020-04-24 | 2023-07-25 | 北京百度网讯科技有限公司 | Instrument panel drawing processing method and device, electronic equipment and storage medium |
| CN111538894B (en) * | 2020-06-19 | 2020-10-23 | 腾讯科技（深圳）有限公司 | Query feedback method and device, computer equipment and storage medium |
| CN112527819B (en) * | 2020-12-08 | 2024-06-04 | 北京百度网讯科技有限公司 | Address book information retrieval method and device, electronic equipment and storage medium |
| CN114328655B (en) * | 2021-12-14 | 2024-11-01 | 上海金仕达软件科技股份有限公司 | Intelligent business processing method and system based on deep learning |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8301633B2 (en) * | 2007-10-01 | 2012-10-30 | Palo Alto Research Center Incorporated | System and method for semantic search |
| CN104598445B (en) * | 2013-11-01 | 2019-05-10 | 腾讯科技（深圳）有限公司 | Automatically request-answering system and method |
| CN104050256B (en) * | 2014-06-13 | 2017-05-24 | 西安蒜泥电子科技有限责任公司 | Initiative study-based questioning and answering method and questioning and answering system adopting initiative study-based questioning and answering method |
-
2017
- 2017-07-14 CN CN201710575024.3A patent/CN107330120B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN107330120A (en) | 2017-11-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107330120B (en) | Inquire answer method, inquiry answering device and computer readable storage medium | |
| US10909969B2 (en) | Generation of language understanding systems and methods | |
| CN110765244B (en) | Method, device, computer equipment and storage medium for obtaining answering operation | |
| US20210311975A1 (en) | Language agnostic command-understanding digital assistant | |
| Schalkwyk et al. | “Your word is my command”: Google search by voice: A case study | |
| CN110111780B (en) | Data processing method and server | |
| JP5926378B2 (en) | Text input method, apparatus, and system | |
| US9582757B1 (en) | Scalable curation system | |
| CN107430616B (en) | Interactive reformulation of voice queries | |
| JP2022547631A (en) | Stopword data augmentation for natural language processing | |
| US9336298B2 (en) | Dialog-enhanced contextual search query analysis | |
| CN110415679B (en) | Voice error correction method, device, equipment and storage medium | |
| KR101751113B1 (en) | Method for dialog management based on multi-user using memory capacity and apparatus for performing the method | |
| US20230186161A1 (en) | Data manufacturing frameworks for synthesizing synthetic training data to facilitate training a natural language to logical form model | |
| CN112948534A (en) | Interaction method and system for intelligent man-machine conversation and electronic equipment | |
| CN112154465A (en) | Method, device and equipment for learning intention recognition model | |
| KR101677859B1 (en) | Method for generating system response using knowledgy base and apparatus for performing the method | |
| CN110147494B (en) | Information searching method and device, storage medium and electronic equipment | |
| CN116501960A (en) | Content retrieval method, device, equipment and medium | |
| CN107424612A (en) | Processing method, device and machine readable media | |
| KR20220109238A (en) | Device and method for providing recommended sentence related to utterance input of user | |
| CN110809796A (en) | Speech recognition system and method with decoupled wake phrases | |
| US20230376700A1 (en) | Training data generation to facilitate fine-tuning embedding models | |
| US12099908B2 (en) | Updating training examples for artificial intelligence | |
| RU2759090C1 (en) | Method for controlling a dialogue and natural language recognition system in a platform of virtual assistants |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20200727 Address after: 518000 Nanshan District science and technology zone, Guangdong, Zhejiang Province, science and technology in the Tencent Building on the 1st floor of the 35 layer Patentee after: TENCENT TECHNOLOGY (SHENZHEN) Co.,Ltd. Address before: 100029, Beijing, Chaoyang District new East Street, building No. 2, -3 to 25, 101, 8, 804 rooms Patentee before: Tricorn (Beijing) Technology Co.,Ltd. |
|
| TR01 | Transfer of patent right |