JP2021005391A - Text monitoring system, method for monitoring text, and program - Google Patents

Abstract

To provide an analysis support system, a method for supporting analysis, and a program which can increase accuracy of detecting an event in a monitor using a text.SOLUTION: An analysis support system (monitoring system 1) includes: text acquiring means 10 for acquiring first texts; reception means for receiving designation of second texts; monitoring means 70 for monitoring to determine whether or not the number of ones of the acquired first texts that include designated second texts satisfies a predetermined condition; and display control means 80 for displaying a screen which can show that a predetermined condition is satisfied when the condition is satisfied. A relation between the second texts and the first texts including the second texts shows that the content of the second texts is true if the content of the first texts is true.SELECTED DRAWING: Figure 2

Description

The present invention relates to a text monitoring system, a text monitoring method, and a recording medium, and more particularly to a text monitoring system, a text monitoring method, and a recording medium for monitoring using text.

A method of monitoring the number of texts containing a specific keyword is known as a technique for monitoring the occurrence of an event such as a defect report or a complaint for a product from the collected texts. For example, Patent Document 1 discloses a technique for extracting a rumor expression from a collected document in accordance with a rumor rule. In monitoring text containing a specific keyword, it is necessary to define the keyword of the event to be monitored in advance.

As a monitoring technique that does not require the definition of keywords, for example, a method of classifying texts into clusters based on the degree of sharing of keywords between texts and monitoring the number of texts classified in each cluster can be considered.

However, in general, the text to be monitored may have a mixture of viewpoints. For this reason, even if texts are classified based on the degree of sharing of keywords, the viewpoints of each cluster may become unclear due to oversight of viewpoints or classification of texts from different viewpoints into the same cluster. .. Therefore, when text is monitored using a cluster generated based on the degree of keyword sharing, the event detection accuracy may be low.

FIG. 25 is a diagram showing an example of clustering results based on the degree of sharing of keywords. The text in FIG. 25 is a text relating to a malfunction of an automobile. These texts are classified into the same cluster because they share keywords such as "fault", "start", and "periphery". However, even if an alert is notified when the number of texts classified in this cluster increases, the text contained in the cluster is not the text related to the same event, so the user can see what kind of event has increased. I can't figure out exactly.

As a related technique, Non-Patent Document 1 discloses an implication clustering technique for extracting implications between texts and classifying the texts having implications into the same cluster. Further, Patent Document 2 discloses a technique for generating an implication graph showing an implication relationship based on the implication relationship between texts. Patent Document 3 discloses a technique of extracting utterances from a set of dialogue texts and extracting utterances having an implication relationship as utterance clusters.

Japanese Unexamined Patent Publication No. 2003-271609 Japanese Patent No. 5494999 Japanese Unexamined Patent Publication No. 2013-19991

"NEC develops technology to automatically group a large amount of document data with the same meaning", [online], NEC Corporation, [Search on February 9, 2015], Internet <URL: http: // jpn. nec.com/press/201411/20141118_02.html>

As described above, when event monitoring is performed using a cluster generated based on the degree of keyword sharing, there is a technical problem that the event detection accuracy is low.

An object of the present invention is to provide a text monitoring system, a text monitoring method, and a recording medium capable of solving the above-mentioned technical problems and improving the detection accuracy of an event in monitoring using text.

The analysis support system according to one aspect of the present invention has received the designation of the text acquisition means for acquiring the first text, the reception means for accepting the designation of the second text, and the acquired first text. A monitoring means for monitoring whether or not the number of the first texts including the second text satisfies a predetermined condition, and a screen capable of grasping that the predetermined condition is satisfied when the predetermined condition is satisfied. The relationship between the second text and the first text implying the second text is such that if the content of the first text is true, the relationship between the second text and the first text is provided. Indicates that the content of the second text is true.

In the analysis support method according to one aspect of the present invention, the text acquisition means provided in the computer acquires the first text, the reception means provided in the computer receives the designation of the second text, and the computer. The monitoring means provided in the above monitors whether or not the number of the first texts including the second text that has received the designation among the acquired first texts satisfies a predetermined condition. An analysis support method for displaying a screen on which it is possible to grasp that the predetermined condition is satisfied when the display control means provided in the computer satisfies the predetermined condition, the second text and the second text. The relationship with the first text, which implies the second text, indicates that if the content of the first text is true, then the content of the second text is true.

The program according to one aspect of the present invention acquires the first text, accepts the designation of the second text, and includes the second text of the acquired first text that has accepted the designation. A program that monitors whether or not the number of first texts to be used satisfies a predetermined condition, and if the predetermined condition is satisfied, executes a process of displaying a screen that can be grasped that the predetermined condition is satisfied. The relationship between the second text and the first text implying the second text is that if the content of the first text is true, the content of the second text is true. Indicates that.

The technical effect of the present invention is that the accuracy of event detection can be improved in monitoring using text.

It is a block diagram which shows the basic structure of the 1st Embodiment of this invention. It is a block diagram which shows the structure of the monitoring system 1 in the 1st Embodiment of this invention. It is a block diagram which shows the structure of the monitoring system 1 realized by the computer in the 1st Embodiment of this invention. It is a flowchart which shows the operation of the monitoring system 1 in the 1st Embodiment of this invention. It is a figure which shows the example of the text data in the 1st Embodiment of this invention. It is a figure which shows the example of the extraction result of the implication relation in the 1st Embodiment of this invention. It is a figure which shows the example of the representative text information in the 1st Embodiment of this invention. It is an example of the number of texts implying each representative text in the first embodiment of the present invention. It is a figure which shows the example of the notification screen 90 in the 1st Embodiment of this invention. It is another example of the number of texts implying each representative text for each monitoring period in the first embodiment of the present invention. It is a block diagram which shows the structure of the monitoring system 1 in the 2nd Embodiment of this invention. It is a figure which shows the example of the text data in the 2nd Embodiment of this invention. It is a figure which shows the example of the extraction result of the implication relation in the 2nd Embodiment of this invention. It is a figure which shows the example of the representative text information in the 2nd Embodiment of this invention. It is a figure which shows the example of the object type information in the 2nd Embodiment of this invention. It is an example of the total number of texts implying different representative texts in the second embodiment of the present invention. It is a figure which shows the example of the notification screen 100 in the 2nd Embodiment of this invention. It is a flowchart which shows the operation of the monitoring system 1 in the 3rd Embodiment of this invention. It is a figure which shows the example of the text data in the 3rd Embodiment of this invention. It is a figure which shows the example of the extraction result of the implication relation in the 3rd Embodiment of this invention. It is a figure which shows the example of the representative text information in the 3rd Embodiment of this invention. It is an example of the number of texts implying each representative text in the third embodiment of the present invention. It is a figure which shows the example of the comparison screen 110 in the 3rd Embodiment of this invention. It is a figure which shows the example of the relationship between the representative text and the element text in embodiment of this invention. It is a figure which shows the example of the clustering result based on the degree of sharing of a keyword.

First, implication clustering, which is a text clustering method used in the embodiment of the present invention, will be described. In implication clustering, as described in Non-Patent Document 1, clustering is performed based on the implication relationship, which is the relationship of meaning between texts. In the embodiment of the present invention, the implication relationship is defined as follows, as in Patent Document 2. That is, if the content of the first text is true, then if the content of the second text is true, then the first text is defined as entailing the second text. Further, when the content of the second text can be read from the content of the first text, it may be defined that the first text implies the second text. By using implication clustering, it is possible to extract all the viewpoints included in the text to be analyzed, the texts in the cluster have common implications, and the representative text representing the outline of the cluster.

In order to facilitate understanding of implications, explanations will be given using specific examples.
<Specific example 1>
First text: President Obama lives in the White House.
Second text: President Obama lives in the United States.

In this case, if the content of the first text is true, then the content of the second text is true, so it can be said that the first text implies the second text.

<Specific example 2>
First text: Prime Minister Tsuyoshi Inukai was assassinated by naval officers.
Second text: Prime Minister Tsuyoshi Inukai died In this case, if the content of the first text is true, then the content of the second text is true, so if the first text implies the second text I can say.

Here, "representative text" and "element text" are defined. When the implication clustering process is executed on a set of texts, the representative text and the element texts are determined. The relationship between the representative text and the element text is that if the content of the element text is true, then the content of the representative text is true. That is, the relationship between the representative text and the element text is that the element text implies the representative text.

FIG. 24 is a diagram showing an example of the relationship between the representative text and the element text in the embodiment of the present invention. In order to facilitate the understanding of the representative text and the element text, the description will be made with reference to FIG. FIG. 24 shows how the implication clustering process was executed for 11 texts from T1 to T11. The circular symbol in FIG. 24 indicates a piece of text. The arrow in FIG. 24 indicates that the original text of the arrow implies the text at the tip of the arrow. In FIG. 24, the texts T6, T7, T11 imply text T1. Similarly, texts T2, T3, T7, and T10 imply text T5, and texts T2, T4, T7, and T8 imply text T9. At this time, the texts T6, T7, and T11 are element texts of the representative text T1. Similarly, the texts T2, T3, T7, and T10 are element texts of the representative text T5. Similarly, the texts T2, T4, T7, and T8 are element texts of the representative text T9.

Here, the representative text itself may be treated as an element text. For example, the texts T1, T6, T7, and T11 may be element texts of the representative text T1.

(First Embodiment)
The first embodiment of the present invention will be described.

First, the configuration of the first embodiment of the present invention will be described.

FIG. 2 is a block diagram showing the configuration of the monitoring system 1 according to the first embodiment of the present invention.

Referring to FIG. 2, the monitoring system 1 according to the first embodiment of the present invention includes a text acquisition unit 10, a text storage unit 20, an implication relationship extraction unit 30, a representative text extraction unit 40, a representative text storage unit 50, and a determination. A unit 60, a monitoring unit 70, and a display control unit 80 are included. The monitoring system 1 is an embodiment of the text monitoring system of the present invention.

The text acquisition unit 10 acquires the text to be monitored.

The text storage unit 20 stores text data indicating the text acquired by the text acquisition unit 10.

FIG. 5 is a diagram showing an example of text data in the first embodiment of the present invention. The example of FIG. 5 is an example in which the text to be monitored is a natural language text related to the "phenomenon" in the defect report of the automobile. The text data includes the acquisition date and time of the text and the text. The code in parentheses before the text indicates the text identifier.

The text to be monitored is extracted from, for example, a document (defect report, etc.). In this case, the text is extracted, for example, by acquiring the description for the category (phenomenon) to be monitored in the document described for each of a plurality of categories (phenomenon, cause, countermeasure, etc. of the defect) according to a predetermined format. Will be done. In addition, the text may be extracted from the document described in free form by specifying the description part related to the category to be monitored. Further, the text may be extracted from a call log generated by voice recognition of a conversation in a call center or the like, for example. In addition, the text may be extracted from word-of-mouth sites, blogs, and SNS (social networking services).

The implication relationship extraction unit 30 extracts the implication relationship by using the text of the implication relationship extraction target (hereinafter, also referred to as the extraction target text) among the texts to be monitored. Here, as the text to be extracted, for example, all texts of a specific monitoring period such as the first monitoring period in a plurality of monitoring periods of a predetermined length (regular monitoring period), or one of the specific monitoring periods. The text of the part (predetermined number or predetermined subperiod) is used.

The representative text extraction unit 40 extracts the representative text from the extracted implication relation, and generates a cluster in which the representative text and the element text including the representative text are set.

The representative text storage unit 50 stores representative text information indicating information related to the cluster generated by the representative text extraction unit 40.

The determination unit 60 determines, for each monitoring period, whether each newly acquired text (added to the text data) implies each representative text (belongs to each cluster).

The monitoring unit 70 monitors the number of texts including each representative text for each monitoring period, and outputs the monitoring result via the display control unit 80. In the first embodiment of the present invention, as the output of the monitoring result, a notification is given when the number of texts including the representative text satisfies a predetermined notification condition. Here, as the notification condition, for example, a lower threshold value related to the number of texts (notification if the number of texts is equal to or greater than the threshold value) is used.

The display control unit 80 generates a notification screen 90 for displaying the monitoring result (notification content) and displays it to the user or the like.

The monitoring system 1 may be a computer that includes a CPU (Central Processing Unit) and a storage medium that stores a program, and operates under control based on the program.

FIG. 3 is a block diagram showing a configuration of a monitoring system 1 realized by a computer according to the first embodiment of the present invention.

The monitoring system 1 includes a CPU 2, a storage device 3 (storage medium) such as a hard disk or memory, a communication device 4 that communicates with other devices, an input device 5 such as a mouse or keyboard, and an output device 6 such as a display. Including.

The CPU 2 executes a computer program for realizing the functions of the text acquisition unit 10, the implication relationship extraction unit 30, the representative text extraction unit 40, the determination unit 60, the monitoring unit 70, and the display control unit 80. The storage device 3 stores the data of the text storage unit 20 and the representative text storage unit 50. The output device 6 outputs a notification screen 90 to the user or the like. The input device 5 receives, for example, designation of notification conditions and the like from the user and the like. Further, the communication device 4 may output the notification screen 90 to another device or the like.

Further, each component of the monitoring system 1 shown in FIG. 2 may be an independent logic circuit.

Further, each component of the monitoring system 1 shown in FIG. 2 may be distributedly arranged in a plurality of physical devices connected by wire or wirelessly.

Next, the operation of the first embodiment of the present invention will be described.

Here, the text acquisition unit 10 extracts the text to be monitored from the document related to the defect report by periodically accessing, for example, a storage unit (not shown), and stores the text data as shown in FIG. 5 as text. It is assumed that it is stored in part 20.

It is also assumed that text monitoring is performed every monitoring period (1 month). Further, it is assumed that the text to be extracted with implications (text to be extracted) is the first predetermined number of texts in the text of the first monitoring period.

FIG. 4 is a flowchart showing the operation of the monitoring system 1 according to the first embodiment of the present invention.

First, the implication relationship extraction unit 30 extracts the implication relationship between the texts to be extracted stored in the text storage unit 20 during the first monitoring period (step S101).

Here, the implication relationship extraction unit 30 extracts the implication relationship between texts by, for example, performing the same determination processing as in Patent Document 2. In this case, the implication relationship extraction unit 30 determines the presence or absence of the implication relationship by comparing the content words contained in the text and calculating the coverage ratio. If the implication relationship extraction unit 30 can extract the implication relationship between texts, the implication relationship between texts may be determined by a determination process different from that of Patent Document 2.

FIG. 6 is a diagram showing an example of the extraction result of the implication relationship in the first embodiment of the present invention. In FIG. 6, the original text of the arrow indicates that it implies the previous text. In the example of FIG. 6, the texts T6, T7, T11, ... Imply the text T1 "warning light turned on". Similarly, texts T2, T3, T7, T10, ... Imply text T5 "abnormal noise", and texts T2, T4, T7, T8, ... Imply text T9 "stall". ing.

For example, the implication relationship extraction unit 30 extracts the implication relationship as shown in FIG. 6 from the extraction target text whose acquisition date and time is within the monitoring period “2015/1” in FIG.

The representative text extraction unit 40 extracts the representative text from the extracted implications and generates a cluster (step S102). The representative text extraction unit 40 generates clusters based on the implication relationships extracted by the implication relationship extraction unit 30, for example, as in the technique of Non-Patent Document 1. Here, among the texts to be extracted, the text implied by other texts is set as the representative text of the cluster, and the text implying the representative text is set as the element text of the cluster. If the text implies more than one representative text, the text is set to the element text of multiple clusters. In the embodiment of the present invention, the text itself set as the representative text of a certain cluster is also set as the element text implying the representative text of the cluster. The representative text extraction unit 40 stores the representative text information in which the identifier of the representative text of each cluster is associated with the identifier of the element text of the cluster in the representative text storage unit 50.

FIG. 7 is a diagram showing an example of representative text information in the first embodiment of the present invention. In the example of FIG. 7, the texts T1 "warning light turned on", T5 "abnormal noise", and T9 "stall" are set as the representative texts of clusters C1, C2, and C3, respectively. There is. Further, the texts T6, T7, T11 ... Implying the text T1 and the text T1 are set as the element texts of the cluster C1. Similarly, the text including the text T5 and the text T5 is set as the element text of the cluster C2, and the text including the text T9 and the text T9 is set as the element text of the cluster C3.

For example, the representative text extraction unit 40 generates the representative text information as shown in FIG. 7 based on the implication relationship of FIG.

The representative text extraction unit 40 may further integrate the different clusters into one cluster based on the degree of duplication of element texts between different clusters.

Next, in each monitoring period, the determination unit 60 means that each text newly acquired (added to the text data) stored in the text storage unit 20 during the monitoring period implies each representative text. Whether or not it is determined (step S103).

The monitoring unit 70 totals the number of texts including the representative text for each representative text (step S104). Here, when the text of the monitoring period includes the text to be extracted, the number of the texts to be extracted is also used as the text to be aggregated. In addition, the monitoring unit 70 totals the number of texts that are not implied in any representative text as the number of "other" texts.

The monitoring unit 70 determines whether or not the number of texts including each representative text in the monitoring period satisfies a predetermined notification condition (step S105).

In step S105, when there is a representative text satisfying a predetermined notification condition (step S105 / Y), the monitoring unit 70 notifies via the display control unit 80 (step S106). Here, the display control unit 80 generates a notification screen 90 and displays it on the user or the like.

Further, while the number of "other" texts is less than the predetermined extraction threshold, the process from step S103 is repeated for each monitoring period (step S107 / N).

FIG. 8 is an example of the number of texts implying each representative text in the first embodiment of the present invention.

For example, the monitoring unit 70 totals the number of texts implying the representative texts T1, T5, and T9 as shown in FIG. 8 in each of the monitoring periods “2015/1”, “2015/2”, and so on.

Here, for example, when the notification condition is the lower limit threshold value “100 or more” related to the number of texts, the number of texts including the representative text T5 “abnormal noise” in the monitoring period “2015/5” is notified. Satisfy the conditions. Therefore, the monitoring unit 70 notifies the event "abnormal noise" in the monitoring period "2015/5".

FIG. 9 is a diagram showing an example of the notification screen 90 in the first embodiment of the present invention.

In the example of FIG. 9, the notification screen 90 includes a notification area 91, a representative text display area 92, a time series display area 93, and a text display area 94.

In the notification area 91, the monitoring period of the notification target (monitoring period in which the notification threshold is detected to be exceeded) and the representative text of the notification target (representative text exceeding the notification threshold) are displayed.

In the "cluster" column of the representative text display area 92, for example, the representative text of each cluster is displayed. Further, in the "number of cases" column, for example, the number of texts including each representative text in the monitoring period to be notified is displayed.

In the time series display area 93, for example, a graph showing the number of texts (time series) including each representative text for each monitoring period is displayed.

In the "detailed text" field of the text display area 94, for example, texts including the representative text of the notification target in the monitoring period of the notification target are displayed in the order of acquisition date and time.

The display control unit 80 displays the notification screen 90 as shown in FIG. 9 on the user or the like.

The user or the like can refer to the notification area 91 of the notification screen 90 of FIG. 9 and grasp the events with a large number (or a small number of occurrences) at the summary level. In addition, the user or the like can refer to the text display area 94 and grasp the details of the event to be notified.

Further, when the number of "other" texts is equal to or greater than the predetermined extraction threshold value (step S107 / Y), the processing from step S101 is performed.

FIG. 10 is another example of the number of texts implying each representative text for each monitoring period in the first embodiment of the present invention.

Here, for example, when the extraction threshold value is “10”, the number of “other” texts in the monitoring period “2016/4” is equal to or greater than the extraction threshold value. Therefore, the implication relationship extraction unit 30 extracts the implication relationship again from, for example, the extraction target text of the next monitoring period “2016/5”. Then, the representative text extraction unit 40 extracts the representative text again from the extracted implications and generates a cluster. Here, it is assumed that the representative text extraction unit 40 extracts a new representative text T201 "oil is leaking" in addition to the representative texts T1, T5, and T9.

In the monitoring period "2016/8", the number of texts implying the representative text T201 "oil is leaking" satisfies the notification condition. Therefore, the monitoring unit 70 notifies the event "oil is leaking" in the monitoring period "2016/8".

As described above, the operation of the first embodiment of the present invention is completed.

In the first embodiment of the present invention, a case where the text to be clustered is a text related to a defect report of an automobile has been described as an example. However, the text to be clustered is not limited to this, and the text related to any content such as various phenomena, causes, countermeasures, opinions, evaluations, complaints, requests, etc. may be used.

Further, in the first embodiment of the present invention, the text (all texts or a part of the texts) of a specific monitoring period was used as the text to be extracted (extracted text). Then, a representative text was extracted based on the implication relationship extracted from the text of the monitoring period, and the representative text was used for monitoring the monitoring period after the monitoring period. Further, when the number of other representative texts was equal to or more than a predetermined extraction threshold value, the implication relation was re-extracted from the text of the new monitoring period, and the representative text was re-extracted based on the implication relation.

However, the present invention is not limited to this, and the text of each monitoring period may be used as the text to be extracted. In this case, for each monitoring period, the implication relationship is re-extracted from the text of the monitoring period and the representative text is re-extracted, and the representative text is used for monitoring the monitoring period.

Further, the present invention is not limited to this, and the text acquired by each monitoring period (text before the monitoring period) may be used as the text to be extracted. In this case, for each monitoring period, the text of the monitoring period is added to the extraction target text, and the implication relationship is re-extracted and the representative text is re-extracted.

Further, when a new representative text is extracted by re-extracting the representative text, the representative text extraction unit 40 may notify the new representative text via the display control unit 80.

Further, the representative text extraction unit 40 receives the designation of the representative text to be monitored from the user or the like among the extracted representative texts, and the monitoring unit 70 monitors the number of implied texts for the designated representative text. You may.

Further, in the first embodiment of the present invention, the lower limit threshold value related to the number of texts is used as the notification condition. However, not limited to this, as a notification condition, an upper threshold (notification if the number of texts is less than or equal to the threshold) or a lower threshold of the increase or decrease (the increase or decrease in the number of texts is the threshold). If the above is the case, notification) may be used. Further, as the notification condition, a predetermined statistic or distribution condition related to the number of texts or the amount of increase / decrease may be set in addition to the threshold value of the number of texts or the amount of increase / decrease.

Next, the basic configuration of the first embodiment of the present invention will be described.

FIG. 1 is a block diagram showing a basic configuration of the first embodiment of the present invention. Referring to FIG. 1, the monitoring system 1 (text monitoring system) of the present invention includes a text acquisition unit 10, a determination unit 60, and a monitoring unit 70. The monitoring system 1 is accessiblely connected to a storage unit that stores representative text, which is the text implied by other texts, extracted based on the implications between the texts. The text acquisition unit 10 acquires the text to be monitored. The determination unit 60 determines whether the acquired text implies a representative text. The monitoring unit 70 monitors the number of texts including the representative text and outputs the monitoring result.

Next, the effect of the first embodiment of the present invention will be described.

According to the first embodiment of the present invention, the accuracy of detecting an event can be improved in monitoring using text. The reason is that the determination unit 60 determines whether the acquired text implies the representative text, and the monitoring unit 70 monitors the number of texts including the representative text and outputs the monitoring result. This allows the user to grasp the events with a high (or low) number of occurrences for each clear summary level viewpoint.

Further, according to the first embodiment of the present invention, in monitoring using a text, monitoring can be performed without defining a monitoring target in advance. The reason is that the representative text extraction unit 40 extracts the representative text from the implications between the texts to be monitored.

(Second Embodiment)
Next, a second embodiment of the present invention will be described.

A second embodiment of the present invention differs from the first embodiment of the present invention in that it monitors the total number of texts implying a particular type of representative text.

First, the configuration of the second embodiment of the present invention will be described.

FIG. 11 is a block diagram showing the configuration of the monitoring system 1 according to the second embodiment of the present invention.

Referring to FIG. 11, the monitoring system 1 of the second embodiment of the present invention includes the target type storage unit 65 in addition to the configuration of the monitoring system 1 of the first embodiment of the present invention.

The target type storage unit 65 stores target type information indicating the type of the representative text to be monitored.

FIG. 15 is a diagram showing an example of target type information in the second embodiment of the present invention. The target type information includes the type of the representative text (“type” column), the notification condition (“notification condition” column), and the representative text classified into the type (“cluster column”). In the example of FIG. 15, "bad evaluation" and "good evaluation" are set as the types of representative texts. Further, as a notification condition, a lower limit threshold value relating to the total number of texts implied in each type of representative text is set.

The representative text extraction unit 40 classifies the extracted representative text into the types indicated by the target type information. Here, the representative text extraction unit 40 classifies the extracted representative texts into types according to a predetermined classification rule. In this case, in the classification rule, for example, a keyword or expression is defined for each type, and the representative text including the keyword or expression is classified into the corresponding type. Further, the representative text extraction unit 40 may output the extracted representative text to the user or the like and accept the designation of the type of the representative text from the user or the like.

The monitoring unit 70 notifies each type indicated by the target type information when the total number of texts including the representative text of the type satisfies the notification condition set for the type.

The display control unit 80 generates a notification screen 100 for displaying the monitoring result (notification content) and displays it to the user or the like.

Next, the operation of the second embodiment of the present invention will be described.

In step S102 described above, the representative text extraction unit 40 classifies the extracted representative text into the types indicated by the target type information.

In step S104, the monitoring unit 70 totals the total number of texts including the representative text of the type for each type indicated by the target type information.

In step S105, the monitoring unit 70 determines whether or not the total number of texts for each type in the monitoring period satisfies the notification condition set for the type. If the notification condition is satisfied, the monitoring unit 70 notifies in step S106.

Next, a specific example of the operation of the second embodiment of the present invention will be described.

FIG. 12 is a diagram showing an example of text data in the second embodiment of the present invention. The example of FIG. 12 is an example in which the text to be monitored is the text related to the "evaluation" of the television product. Here, it is assumed that the text acquisition unit 10 stores the text data as shown in FIG. 12 in the text storage unit 20. Further, it is assumed that the target type information has the type as shown in FIG. 15 and the notification condition.

FIG. 13 is a diagram showing an example of the extraction result of the implication relationship in the second embodiment of the present invention.

The implication relationship extraction unit 30 extracts the implication relationship from the extraction target text whose acquisition date and time is within the monitoring period “2015/1” in FIG. 12, for example, as shown in FIG.

FIG. 14 is a diagram showing an example of representative text information in the second embodiment of the present invention.

The representative text extraction unit 40 generates representative text information as shown in FIG. 14 based on the implication relationship of FIG. Then, the representative text extraction unit 40 classifies the representative text T304 "high price" into the type "bad evaluation", the representative text T305 "high functionality", and T306 "good design" into the type "good evaluation". , As shown in FIG. 15, the target type information is set.

FIG. 16 is an example of the total number of texts implying different representative texts in the second embodiment of the present invention.

For example, the monitoring unit 70 includes representative texts of the types "bad evaluation" and "good evaluation" in each of the monitoring periods "2015/1", "2015/2", ..., As shown in FIG. Aggregate the total number of texts.

Here, the total number of texts including the representative text of the type "good evaluation" in the monitoring period "2015/5" satisfies the notification condition. Therefore, the monitoring unit 70 notifies the event of the type "good evaluation" of the monitoring period "2015/5".

FIG. 17 is a diagram showing an example of the notification screen 100 in the second embodiment of the present invention.

In the example of FIG. 17, the notification screen 100 includes a notification area 101, a type display area 102, a representative text display area 103, a time series display area 104, and a text display area 105.

In the notification area 101, the monitoring period of the notification target and the type of the notification target are displayed.

In the "type" column of the type display area 102, for example, each type is displayed. Further, in the "number of cases" column, for example, the total number of texts including the representative texts of each type in the monitoring period to be notified is displayed.

In the "cluster" column of the representative text display area 103, for example, the representative text of the type to be notified is displayed. Further, in the "number of cases" column, for example, the number of texts including each representative text in the monitoring period to be notified is displayed.

In the time-series display area 104, for example, a graph showing the total number of texts (time-series) including representative texts for each type is displayed for each monitoring period.

In the "detailed text" field of the text display area 105, texts including the representative text of the type of the notification target in the monitoring period of the notification target are displayed in the order of acquisition date and time, for example.

The display control unit 80 displays the notification screen 100 as shown in FIG. 17 on the user or the like.

The user or the like can refer to the notification area 101 of the notification screen 100 of FIG. 17 and grasp the types of events with a large number (or a small number of occurrences). If the specific type is "good evaluation" or "bad evaluation" related to the product, the user or the like can easily grasp what kind of evaluation is given to the product. In addition, the user or the like can refer to the text display area 105 and confirm the specific contents of the "good evaluation" and the "bad evaluation".

As described above, the operation of the second embodiment of the present invention is completed.

Next, the effect of the second embodiment of the present invention will be described.

According to the second embodiment of the present invention, the types of events with a large number of occurrences (or a small number of occurrences) can be easily grasped. The reason is that the monitoring unit 70 monitors the total number of texts including the representative texts belonging to a predetermined type.

(Third Embodiment)
Next, a third embodiment of the present invention will be described.

In the third embodiment of the present invention, as a monitoring result, an increasing / decreasing tendency of the number of texts including each representative text is output between different monitoring periods. Different from.

First, the configuration of the third embodiment of the present invention will be described.

The block diagram showing the configuration of the third embodiment of the present invention is the same as that of the first embodiment of the present invention (FIG. 2).

As the monitoring result, the monitoring unit 70 outputs an increase / decrease tendency (comparison result) of the number of texts including each representative text between different monitoring periods.

The display control unit 80 generates a comparison screen 110 for displaying the monitoring result (increase / decrease tendency) and displays (outputs) it to the user or the like.

Next, the operation of the third embodiment of the present invention will be described.

FIG. 18 is a flowchart showing the operation of the monitoring system 1 according to the third embodiment of the present invention.

First, the implication relationship extraction unit 30 extracts the implication relationship between the texts to be extracted in the same manner as in step S101 described above (step S201). Here, the implication relationship extraction unit 30 extracts the implication relationship by using all the texts of the plurality of monitoring periods as the extraction target texts.

The representative text extraction unit 40 extracts the representative text from the extracted implications and generates a cluster in the same manner as in step S102 described above (step S202).

Similar to step S103 described above, the determination unit 60 determines whether or not each text acquired during each monitoring period implies each representative text (step S203).

Similar to step S104 described above, the monitoring unit 70 totals the number of texts including the representative text for each representative text (step S204).

The determination unit 60 and the monitoring unit 70 repeat the processes of steps S203 and S204 for all monitoring periods (step S205).

The monitoring unit 70 compares the number of texts implying each representative text between different monitoring periods (step S206).

The monitoring unit 70 outputs an increasing / decreasing tendency (comparison result) of the number of texts including each representative text via the display control unit 80 (step S207). Here, the display control unit 80 generates the comparison screen 110 and displays it to the user or the like.

Next, a specific example of the operation of the third embodiment of the present invention will be described.

FIG. 19 is a diagram showing an example of text data in the third embodiment of the present invention. Here, it is assumed that the text to be monitored is the text related to the "evaluation" of the television product, as in the second embodiment of the present invention. In addition, the monitoring unit 70 has a text that implies a representative text between the one-month monitoring period (before the campaign) before the campaign related to the TV product and the one-month monitoring period (after the campaign) after the campaign. Suppose you want to compare numbers. Then, it is assumed that the text data as shown in FIG. 19 is stored in the text storage unit 20. In addition, it is assumed that the text to be extracted (extracted text) is the entire text of the monitoring period both before and after the campaign.

FIG. 20 is a diagram showing an example of the extraction result of the implication relationship in the third embodiment of the present invention.

The implication relationship extraction unit 30 extracts implication relationships from, for example, all the texts in FIG. 19, as shown in FIG.

FIG. 21 is a diagram showing an example of representative text information in the third embodiment of the present invention. In the example of FIG. 21, the texts T501 "high price", T601 "good design", and T604 "high functionality" are set as the representative texts of the clusters C1, C2, and C3, respectively. ..

The representative text extraction unit 40 generates representative text information as shown in FIG. 21 based on the implication relationship of FIG. 20.

FIG. 22 is an example of the number of texts implying each representative text in the third embodiment of the present invention.

For example, the monitoring unit 70 includes texts including the representative texts T501, T601, and T604 in each of the pre-campaign monitoring period “2015/1” and the post-campaign monitoring period “2015/2”, as shown in FIG. Aggregate the number of.

In addition, the monitoring unit 70 compares the number of texts implying the representative texts T501, T601, and T604 between the monitoring periods “2015/1” and “2015/2”. Here, the number of texts including the representative text T501 was 100 in the monitoring period "2015/1" before the campaign, but 0 in the monitoring period "2015/2" after the campaign. Therefore, the monitoring unit 70 detects the "deletion" of the cluster of the representative text T501. The number of texts including the representative text T601 was 0 before the campaign, but 100 after the campaign. Therefore, the monitoring unit 70 detects the “addition” of the cluster of the representative text T501. Furthermore, the number of texts implying the representative text T604 was 50 before the campaign, but 200 after the campaign. Therefore, an "increase" in the number of texts implying the representative text T604 is detected. Similarly, the monitoring unit 70 may detect a “decrease” in the number of texts including the representative text.

FIG. 23 is a diagram showing an example of the comparison screen 110 in the third embodiment of the present invention.

In the example of FIG. 23, the comparison screen 110 includes a representative text display area 111 and a time series display area 112.

For example, each representative text is displayed in the "cluster" column of the representative text display area 111. Further, in the "number of cases" column, for example, the number of texts including each representative text is displayed for each monitoring period. In the "Comparison result" column, the comparison result of the number of texts including each representative text ("Delete" or "Add" of the cluster, "Increase" or "Decrease" of the number of texts) is displayed.

For example, the display control unit 80 displays the comparison screen 110 as shown in FIG. 23 on the user or the like.

The user or the like can refer to the comparison screen 110 of FIG. 23 and grasp the clusters added or deleted during different monitoring periods, or the events in which the number of occurrences increased or decreased. Here, for example, when a cluster of events related to good evaluation is added or the number of events related to good evaluation is increasing, such as the representative texts T601 "good design" and T604 "high functionality", the campaign may be used. , It can be seen that the evaluation has improved. On the other hand, if the cluster of events related to good evaluation is deleted or the number of events related to good evaluation is decreasing, it can be seen that the evaluation has been deteriorated by the campaign. Similarly, when the cluster of events related to bad evaluation was deleted or the number of events related to bad evaluation decreased, for example, the representative text T501 "High price", the evaluation was improved by the campaign. I understand. In addition, when the number of events related to bad evaluation is increasing or the number of events related to bad evaluation is increasing, it can be seen that the evaluation has been deteriorated by the campaign.

As described above, the operation of the third embodiment of the present invention is completed.

Next, the effect of the third embodiment of the present invention will be described.

According to the third embodiment of the present invention, it is possible to easily grasp the clusters added or deleted, or the events in which the number of occurrences increased or decreased during different monitoring periods. The reason is that the monitoring unit 70 outputs the increasing / decreasing tendency of the number of texts including each representative text between different monitoring periods.

Although the present invention has been described above with reference to the embodiments, the present invention is not limited to the above embodiments. Various changes that can be understood by those skilled in the art can be made within the scope of the present invention in terms of the structure and details of the present invention.

Hereinafter, an example of the reference form will be added.

(Appendix 1)
Sources clustered by classifying texts with implications into the same group based on the implications between texts, and specific means of identifying clusters that belong to the same group as the text added to the source. A text monitoring system comprising a presentation means for calculating the number of texts belonging to the identified cluster and presenting them at a predetermined timing.

The present invention can be applied to a system for monitoring an event from a large amount of document data. For example, the present invention can be applied to a system for monitoring call logs, customer opinions, etc. in order to improve products and services, improve marketing, and improve the efficiency of sales activities. The present invention can also be applied to a system for monitoring product defects, evaluations and requests for products, and a system for monitoring the contents of academic literature and the like.

1 Monitoring system 2 CPU
3 Storage device 4 Communication device 5 Input device 6 Output device 10 Text acquisition unit 20 Text storage unit 30 Implication relationship extraction unit 40 Representative text extraction unit 50 Representative text storage unit 60 Judgment unit 70 Monitoring unit 80 Display control unit 90 Notification screen 91 Notification Area 92 Representative text display area 93 Time series display area 94 Text display area 100 Notification screen 101 Notification area 102 Type display area 103 Representative text display area 104 Time series display area 105 Text display area 110 Comparison screen 111 Representative text display area 112 Time series Indicated Area

Claims (8)

Text acquisition means to acquire the first text,
A reception means that accepts the designation of the second text,
A monitoring means for monitoring whether or not the number of the first texts including the second text that has received the designation among the acquired first texts satisfies a predetermined condition, and
A display control means for displaying a screen that can be grasped to satisfy the predetermined condition when the predetermined condition is satisfied
With
The relationship between the second text and the first text implying the second text is that if the content of the first text is true, then the content of the second text is true. An analysis support system that shows that. The analysis support system according to claim 1, wherein the monitoring means monitors whether or not the number of first texts including the second text that has received the designation exceeds a threshold value. The analysis according to claim 1, wherein the monitoring means monitors whether or not the number of first texts including the second text that has received the designation satisfies a predetermined condition in a plurality of predetermined periods. Support system. The monitoring means monitors whether or not the total number of the first texts including the second text belonging to a predetermined type satisfies the predetermined condition in the plurality of predetermined periods. The analysis support system according to claim 3. The predetermined condition is that the number of the first texts is equal to or more than or equal to a predetermined threshold value, or an increase or decrease in the number of the first texts during different periods among the plurality of periods. The analysis support system according to claim 3 or 4, wherein the amount is equal to or greater than a predetermined threshold value. The analysis support system according to claim 1, wherein the monitoring means monitors an increasing / decreasing tendency of the number of the first texts including the second texts that have received the designation. The text acquisition means provided in the computer acquires the first text and
The reception means provided in the computer accepts the designation of the second text,
The monitoring means provided in the computer monitors whether or not the number of the first texts including the second text that has received the designation among the acquired first texts satisfies a predetermined condition. ,
An analysis support method for displaying a screen on which it is possible to grasp that the predetermined condition is satisfied when the display control means provided in the computer satisfies the predetermined condition.
The relationship between the second text and the first text implying the second text is that if the content of the first text is true, then the content of the second text is true. An analysis support method that shows that. On the computer
Get the first text,
Accepts the specification of the second text,
Among the acquired first texts, it is monitored whether or not the number of the first texts including the second texts that have accepted the designation satisfies a predetermined condition.
A program that executes a process of displaying a screen that can be grasped that the predetermined conditions are satisfied when the predetermined conditions are satisfied.
The relationship between the second text and the first text implying the second text is that if the content of the first text is true, then the content of the second text is true. A program that shows that.

Images