JP6868576B2 - Event presentation system and event presentation device - Google Patents

Description

The present invention relates to an event presentation system and an event presentation device, and is suitable for application to, for example, an event presentation system and an event presentation device that output information related to an abstract event.

A system for automatically generating new knowledge from existing knowledge is disclosed (see Patent Document 1). The system described in Patent Document 1 includes an "input event" input by a user, a "first event" similar to the input event, a "second event" having a causal relationship with the first event, and a second event. The "third event" similar to is searched sequentially. The above system regards a third event whose domain is similar to that of the input event as a "causal event" that has a cause and effect with the input event and associates them. In addition, the above system compares the predicate of an event with the tendency of words and phrases in a dependency relationship to determine the similarity between events, and compares the predicate of an event with the tendency of words and phrases in a co-occurrence relationship between events. Determine domain similarity. In the above system, the known causal relationship between the first event and the second event is acquired from the causal relationship DB.

Japanese Unexamined Patent Publication No. 2009-59323

In the system described in Patent Document 1, a plurality of similar events are grouped into one group in order to detect a new causal event. However, the meaning of the group and the outline thereof are not explained, and it is not easy for the user who analyzes the transition of the event to read the meaning and the outline of the group. Therefore, when the user analyzes the event at the group level, there is a problem that the meaning of the event at the group level cannot be understood and the analysis becomes difficult.

The present invention has been made in consideration of the above points, and an object of the present invention is to propose an event presentation system and an event presentation device that present an event so that an event can be analyzed at a high level.

In order to solve such a problem, in the present invention, the event corresponding to the character string input for the search is acquired as an abstract event from the event information including the event information, and based on the abstract event, before and after the event. The parent-child relationship of the event is associated with the first specific part that identifies the event after the abstract event from the context information with which the relationship is associated, and the parent-child relationship of the event based on the abstract event acquired by the first specific part. The second specific part that identifies the concrete event indicating the child event of the abstract event from the parent-child relationship information and specifies the information of the case related to the concrete event from the case information including the information of the case related to the event, and the first 1 The abstract event acquired in the specific part, the event after the abstract event specified in the first specific part, and the information of the case related to the concrete event of the abstract event specified in the second specific part. , And an output unit to output, are provided.

According to the above configuration, it is possible to output an abstract event, an event after the abstract event, and information on a case related to a concrete event of the abstract event.

Other problems disclosed in the present application and solutions thereof will be clarified by the columns and drawings of embodiments of the invention.

According to the present invention, an event can be presented so that the event can be highly analyzed.

It is a figure which shows an example of the structure of the event presentation system by 1st Embodiment. It is a figure which shows an example of the flowchart which concerns on the event presentation processing by 1st Embodiment. It is a figure which shows an example of the causal model DB by the 1st Embodiment. It is a figure which shows an example of the causal model DB by the 1st Embodiment. It is a figure which shows an example of the causal model DB by the 1st Embodiment. It is a figure which shows an example of the causal model DB by the 1st Embodiment. It is a figure which shows an example of the causal model DB by the 1st Embodiment. It is a figure which shows an example of the article / document DB by 1st Embodiment. It is a figure which shows an example of the article / document DB by 1st Embodiment. It is a figure which shows an example of the text analysis result DB by 1st Embodiment. It is a figure which shows an example of the text analysis result DB by 1st Embodiment. It is a figure which shows an example of the management DB by 1st Embodiment. It is a figure which shows an example of the management DB by 1st Embodiment. It is a figure which shows an example of the management DB by 1st Embodiment. It is a figure which shows an example of the event search screen by 1st Embodiment. It is a figure which shows an example of the search result display screen of the event by 1st Embodiment. It is a figure which shows an example of the flowchart which concerns on the 1st event abstraction processing by 1st Embodiment. It is a figure which shows an example of the flowchart which concerns on the 2nd event abstraction processing by 1st Embodiment. It is a figure which shows an example of the flowchart which concerns on the user abstraction degree setting process by 1st Embodiment.

Hereinafter, one embodiment of the present invention will be described in detail with reference to the drawings.

(1) First embodiment In this embodiment, various information is collected and the collected information is collected when analyzing the current situation regarding corporate trends, social conditions, etc., future prospects, etc. We will explain the configuration (method) that supports the analysis of the current situation recognition, future prospects, etc. by the user by generating information on the combination, current situation recognition as new knowledge, future prospects, etc.

Here, in order to analyze the current situation, future prospects, etc., various events occurring in the world are extracted, the causal relationships between the events are clarified, and the current events are obtained from the causal relationships of the past events. It is necessary to estimate the causal relationship of the above and to estimate the transition of events including unoccurred events.

However, the frequency of events is low, and it is not easy to analyze the causal relationship between events. In this respect, in the present embodiment, a plurality of events are regarded as the same event in consideration of the similarity of the events such as the background and meaning before and after, and it is easy to give a causal relationship between the events. As a result, a model (causal model) in which a causal relationship is given between events can be created, and the transition of events can be predicted based on the causal model. Here, integrating a plurality of similar events is referred to as event abstraction.

Therefore, in the present embodiment, the abstracted event is given meaning, and the user's analysis is supported by using the causal model. Here, the user has various purposes. One user tries to analyze from a global perspective, and another user tries to analyze from a detailed perspective. Therefore, causal models of events should also be constructed according to different levels of abstraction. Therefore, in the present embodiment, a causal model is constructed according to various levels of abstraction, an appropriate causal model of abstraction is presented according to the purpose of the user, and the analysis of the user is supported.

Hereinafter, the details of the present embodiment will be described with reference to FIGS. 1 to 11. However, the present invention is not limited to the present embodiment, and the present invention can be applied to any situation presenting a causal relationship of an event.

In FIG. 1, 100 indicates an event presentation system according to the first embodiment as a whole. The event presentation system 100 is communicably connected to the map information providing system 160, the vocabulary information providing system 170, and the like via the communication line 150. Each of the event presentation system 100, the map information providing system 160, and the vocabulary information providing system 170 can be realized by one or more general computers (computers).

The event presentation system 100 includes a main storage device 110, an auxiliary storage device 130, a central control device 140, a communication device 141, and an input / output device 142. These are connected to each other via a bus.

The main storage device 110 stores the program. Search word reception unit 111, text analysis unit 112, causal model / event search / acquisition unit 113, article / document search / acquisition unit 114, attribute information search / acquisition unit 115, analysis result output unit 116, management data reception The unit 117, the abstraction degree determination unit 118, the analysis target acquisition unit 120, the abstract event generation unit 121, the causal relationship assigning unit 122, and the causal strength / branch number determination unit 123 are program functions (functions of the event presentation system 100). is there. When the subject is described as "○○ part", the central control device 140 shall read each program from the auxiliary storage device 130, load each program into the main storage device 110, and then execute each program. The functions of each program (details below) are realized.

Although the details will be described later, the causal model / event search / acquisition unit 113 acquires the event corresponding to the character string input for the search as an abstract event from the event information including the event information. This is an example of a first specific unit that identifies an event after the abstract event from the context information associated with the context of the event based on the abstract event. The article / document search / acquisition unit 114 identifies the concrete event of the abstract event from the parent-child relationship information associated with the parent-child relationship of the event based on the abstract event acquired in the first specific unit, and the event This is an example of the second specific part that specifies the information of the case related to the above-mentioned concrete event from the case information including the information of the case related to. The attribute information search / acquisition unit 115 is an example of a third specific unit that specifies information related to the information of the case specified by the user. The analysis result output unit 116 includes an abstract event acquired in the first specific unit, an event after the abstract event specified in the first specific unit, and an abstract event specified in the second specific unit. This is an example of an output unit that outputs information on a case related to a concrete event of.

The auxiliary storage device 130 stores (stores) the causal model DB (database) 131, the article / document DB 132, the text analysis result DB 133, and the management DB 134.

The map information providing system 160 is a system that provides the event presenting system 100 with map information and the like around the place based on the information related to the place provided by the event presenting system 100. The map information providing system 160 communicates with the event presenting system 100 via the communication line 150 to provide map information and the like.

The vocabulary information providing system 170 is a system that provides the event presentation system 100 with a degree of abstraction of words based on the words provided by the event presentation system 100. The vocabulary information providing system 170 communicates with the event presenting system 100 via the communication line 150 to provide the abstraction degree of the word.

As the communication line 150, in addition to LAN (Local Area Network), various networks such as dedicated line, WAN (Wide Area Network), power line network, wireless network, public line network, mobile phone network, satellite communication line, etc. are adopted. can do.

In FIG. 1, the event presentation system 100 is supposed to perform input / output independently, but it may be connected to another terminal by the communication line 150 and input / output / output may be performed by the input / output device of the other terminal. Good. In addition, the function of the event presentation system 100 may be realized by a plurality of computers.

Subsequently, a series of processes for presenting the causal relationship of the events in the present embodiment will be described. The process of presenting the causal relationship of the abstract event is shown in FIG. Details of the database used in the process of FIG. 2 will be described later in FIGS. 3A to 6C.

FIG. 2 is a flowchart relating to a process (event presentation process) of estimating the transition before and after the current event based on the causal relationship of the past event and presenting the transition of the event to the user with respect to the event to be searched by the user. An example is shown. Note that FIG. 7 shows an image of an event search screen for the user to input an event to be searched, and FIG. 8 shows an image of an event search result display screen that presents the transition of the event to the user.

In step S200, the event presentation system 100 starts the event presentation process.

In step S201, the search word receiving unit 111 includes a search word indicating the event to be searched (a character string input for the search) and an attribute of the event to be searched (search attribute: place, person / organization, event). An input with any of the occurrence dates and times, or a combination thereof) is received from the user via the input / output device 142 (see FIG. 7).

In step S202, the causal model / event search / acquisition unit 113 searches for an abstract event including the search word received from the user. More specifically, the causal model / event search / acquisition unit 113 divides the search word into words by the morphological analysis technique, searches each word into words in the word information table 500 of the text analysis result DB 133, and finds the corresponding word. Acquires the word ID 501 of the record to be recorded. The causal model / event search / acquisition unit 113 searches for the word ID: noun 303 and the word ID: sa-variant noun 304 in the event table 300 of the causal model DB 131 with the acquired word ID 501, and if there is a corresponding record, The event ID 301 of the corresponding record is stored in the memory area of the main storage device 110 as an event (abstract event) to be searched. Further, the causal model / event search / acquisition unit 113 stores the attributes to be searched in the memory area of the main storage device 110.

In step S203, the causal model / event search / acquisition unit 113 determines whether or not the event ID 301 could be searched in step S202. If it is determined that the causal model / event search / acquisition unit 113 has been able to search, the process is transferred to step S205, and if it is determined that the search is not possible, the process is transferred to step S204.

In step S204, the text analysis unit 112 extracts the meaning vector of the word using the word embedding vector technique based on the word divided from the search word in step S202, and the word closest to the meaning vector of the word. The meaning vector of is searched from the meaning vector 503 of the word information table 500 of the text analysis result DB 133.

Here, instead of searching for the one with the shortest distance, the one with the distance within the threshold value may be substituted. In that case, for example, the threshold value is described in the setting content 602 of the setting table 600 of the management DB 134. It is assumed that it is set by the set value 603 of "word distance threshold".

Then, the text analysis unit 112 replaces the word divided from the search word with the word having the shortest distance (or the word having the distance within the threshold value). After that, as in step S203, the causal model / event search / acquisition unit 113 searches for the word 502 in the word information table 500 of the text analysis result DB 133, acquires the word ID 501 of the corresponding record, and acquires the word ID 501. The word ID: noun 303 and the word ID: sa-hen noun 304 of the event table 300 of the causal model DB 131 are searched by the generated word ID 501, and the event ID 301 of the corresponding record is stored in the memory area of the main storage device 110 as the event to be searched. To do.

In step S205, the causal model / event search / acquisition unit 113 acquires the abstraction degree 305 of the same record for the event ID 301, specifies the abstraction degree of the event, and sets it as the abstraction degree of the event presented to the user. .. Further, the causal model / event search / acquisition unit 113 searches for the pre-event ID 312 of the causal table 310 based on the event ID 301, and sets the post-event ID 313 of the corresponding record as the event (post-event) following the event ID 301. Get as. Further, the causal model / event search / acquisition unit 113 searches for the post-event ID 313 based on the event ID 301, and acquires the pre-event ID 312 of the corresponding record as an event (pre-event) preceding the event ID 301. The causal model / event search / acquisition unit 113 acquires the causal model by repeating the search for the subsequent event and the previous event. It is assumed that the number of repetitions follows the set threshold value, and the threshold value is set to, for example, the set value 603 of the "causal display threshold value" of the setting content 602 of the setting table 600 of the management DB 134.

Step S205 is not limited to the above-mentioned processing contents. For example, the causal model / event search / acquisition unit 113 may acquire a causal model of only subsequent events, or may acquire a causal model of only preceding events. In addition, when the causal model of both the preceding event and the succeeding event is acquired, the analysis of the event becomes easier than the case of acquiring the causal model of either one event.

Further, for example, when the causal model / event search / acquisition unit 113 determines that the abstraction level of the specified event and the abstraction level of the user (see FIG. 11) are different, the causal model matches the abstraction level of the user. You may get a model. When the causal model / event search / acquisition unit 113 determines that the user's abstraction level is higher, the causal model / event search / acquisition unit 113 searches the concrete event ID 322 of the abstract event relationship table 320 of the causal model DB 131, and searches for the abstract event ID 323 of the corresponding record. Acquire a causal model as a new abstract event (abstract event suitable for the user). Further, when the causal model / event search / acquisition unit 113 determines that the degree of abstraction of the user is lower, it searches the abstract event ID 323 of the abstract event relationship table 320 of the causal model DB 131 and searches for one of the corresponding records or All concrete event IDs 322 are set as new abstract events (abstract events suitable for the user), and a causal model is acquired. According to such a process, a causal model can be constructed according to various levels of abstraction, an appropriate causal model of abstraction can be presented according to the purpose of the user, and the analysis of the user can be assisted.

In step S206, the causal model / event search / acquisition unit 113 relates to the event ID (event ID 301, pre-event ID 312, post-event ID 313) of the causal model acquired in step S205, and is an abstract event relationship table 320 of the causal model DB 131. The abstract event ID 323 of the above is searched, and the concrete event ID 322 of the corresponding record is acquired. More specifically, the causal model / event search / acquisition unit 113 traces back the parent-child relationship of the abstract event relationship table 320 until it reaches the lowest (lowest abstraction) concrete event, and the lowest concrete event. Acquires all the concrete event IDs 322 of. A data structure in which an article / document is associated with an event of each abstraction degree may be adopted, and in that case, the concrete event ID 322 corresponding to the concrete event of the abstract event may be acquired.

In step S207, the article / document search / acquisition unit 114 searches the event ID 331 of the event attribute table 330 of the causal model DB 131 based on the concrete event ID 322 searched in step S206. When the event attribute is input to the memory of the main storage device 110 in step S202, the article / document search / acquisition unit 114 narrows down the corresponding records of location 332, person / organization 333, and date / time 334 by the attribute. .. That is, the event ID 331 of the record as a result of the search and narrowing down is acquired. The article / document search / acquisition unit 114 searches the event ID 412 of the article / document-event relationship table 410 of the article / document DB 132 based on the event ID 331, and acquires the article / document ID 411 of the corresponding record. The article / document search / acquisition unit 114 searches the article / document ID 401 of the article / document table 400 based on the acquired article / document ID 411, and the article / document 402, publication date 403, and publication source of the corresponding record. Get 404.

In step S208, the analysis result output unit 116 prepares for screen output. More specifically, the analysis result output unit 116 searches for the abstract event, the pre-event, the post-event, the context (causal model) of the event acquired in step S205, and the article / document acquired in step S207. As a result, it is stored in the main storage device 110 as data to be displayed on the search result display screen of the event.

In step S209, the attribute information search / acquisition unit 115 determines the location 332 of the concrete event acquired in step S206 (for example, the search attribute of the search attribute) with respect to each of the abstract event, the pre-event, and the post-event displayed in step S208. Location), the surrounding map is acquired from the external map information providing system 160.

In step S210, the attribute information search / acquisition unit 115 relates to the person / organization 333 who executed the concrete event acquired in step S206 for each of the abstract event, the pre-event, and the post-event displayed in step S208. The name 342 of the person / organization management table 340 is searched, and the person / organization ID 341 of the corresponding record is acquired. Further, the attribute information search / acquisition unit 115 searches for the parent ID (person / organization) 343 based on the person / organization ID 341 and holds it as child person / organization information. Further, the attribute information search / acquisition unit 115 searches for the person / organization ID 341 based on the parent ID (person / organization) 343 of the corresponding record, and holds the information as the parent person / organization information. The attribute information search / acquisition unit 115 repeats the same process to acquire the person / organization configuration.

In step S211 the analysis result output unit 116 includes the abstract event, the pre-event, the post-event, the context (causal model) of the event, the article / document, and the periphery acquired in step S209, which are stored in the main storage device 110 in step S208. The map of the above and the composition of the person / organization acquired in step S210 are displayed as search results on the search result display screen of the event (see FIG. 8).

In step S212, the analysis result output unit 116 accepts an article / document selection from the user. When the analysis result output unit 116 determines that there is a selection from the user, the process is transferred to step S213, and when it is determined that there is no selection, the analysis result output unit 116 shifts the process to step S214.

In step S213, the analysis result output unit 116 identifies the event ID 331, the place 332, the person / organization 333, and the date and time 334 associated with the article / document selected by the user from the result of step S207, and in step S209. Move the process.

In step S214, the event presentation system 100 ends the event presentation process.

According to the event presentation process, an abstract event whose meaning can be read by grouping a plurality of similar events and a causal model thereof are presented to the user, so that the user can support the analysis of the event.

Next, the details of each database will be described.

3A to 3E show an example of the causal model DB131. The causal model DB 131 stores data related to the causal model of the event. The causal model DB 131 includes an event table 300 (FIG. 3A), a causal table 310 (FIG. 3B), an abstract event relationship table 320 (FIG. 3C), an event attribute table 330 (FIG. 3D), and a person / organization management table 340. (Fig. 3E) is included.

Data related to the event is stored in the event table 300. The event table 300 is an example of event information, and is configured to include information of event ID 301, event 302, word ID: noun 303, word ID: sa-variant noun 304, and abstraction degree 305. The key for identifying each record is event ID 301.

The event ID 301 is an identifier for uniquely identifying the event. Event 302 is a simple expression of an event, and is composed of a noun indicating the object of the event and a sa-variant noun indicating the action of the event. Word ID: The noun 303 is an identifier for uniquely identifying a noun word indicating the object of an event. Word ID: The sa-hen noun 304 is an identifier for uniquely identifying a word of the sa-hen noun indicating the action of an event. The abstraction degree 305 indicates the abstraction degree of the event, and is calculated from the abstraction degree of each word of the word ID: noun 303 and the word ID: sa-variant noun 304.

The causal table 310 stores data on causal relationships between events. The causal table 310 is an example of contextual information, and includes information of link ID 311, pre-event ID 312, post-event ID 313, and intensity 314. The key for identifying each record is the link ID 311.

The link ID 311 is an identifier for uniquely identifying a link indicating a causal relationship between events. The pre-event ID 312 is an identifier for uniquely identifying the event (event indicating the cause) that is the cause of the cause and effect among the causal relationships between the events. The post-event ID 313 is an identifier for uniquely identifying the event (event indicating the result) that precedes the causality among the causal relationships between the events. Intensity 314 indicates the intensity of the causal relationship between events. In the event abstraction, a causal relationship is given in advance by the user (intensity 314 is also given) for the original event, and for example, the strength 314 of the causal relationship between the abstract events of the original event is given. It is calculated by counting the number of concrete events of the next lower abstraction level belonging to the abstract event.

The abstract event relationship table 320 stores data on the relationship between the event before abstraction (child event) and the event after abstraction (parent event). The abstract event relationship table 320 is an example of parent-child relationship information, and includes information on the abstract event ID 321, the concrete event ID 322, and the abstract event ID 323. The key for identifying each record is the abstraction ID 321.

The abstraction ID 321 is an identifier for uniquely identifying the relationship between events when an event is abstracted into another event. The concrete event ID 322 is an identifier indicating an event to be abstracted. The abstract event ID 323 is an identifier indicating an abstracted event.

The event attribute table 330 stores data related to the attributes of the place, person / organization, and date and time related to the event. The event attribute table 330 is an example of the narrowing down information to which the event and the search attribute are associated with each other, and includes information of the event ID 331, the place 332, the person / organization 333, and the date and time 334. The key for identifying each record is event ID 331.

The event ID 331 is an identifier for uniquely identifying the event. Location 332 indicates the location of the event. Person / organization 333 indicates the subject who performed the event. The date and time 334 indicates the date and time when the event occurred.

The person / organization management table 340 stores data related to the structure of the person / organization. The person / organization management table 340 is configured to include information of a person / organization ID 341, a name 342, and a parent ID (person / organization) 343. The key for identifying each record is the person / organization ID 341.

The person / organization ID 341 is an identifier for uniquely identifying the person / organization. Name 342 indicates the name of the person / organization. The parent ID (person / organization) 343 is an identifier for uniquely identifying a person / organization (parent person / organization) that is one level higher in composition when viewed from the person / organization of the person / organization ID 341.

4A and 4B show an example of the article / document DB 132. In the article / document DB 132, data related to articles and documents for extracting events are stored. The article / document DB 132 includes an article / document table 400 (FIG. 4A) and an article / document-event relationship table 410 (FIG. 4B).

In the article / document table 400, data or the like related to an article or a document for extracting an event is stored. The article / document table 400 is an example of case information, and includes information of an article / document ID 401, an article / document 402, a publication date 403, and a publication source 404. The key for identifying each record is the article / document ID 401.

The article / document ID 401 is an identifier for uniquely identifying the article or document. Article / document 402 indicates the specific content of an article or document for extracting an event. The publication date 403 indicates the date and time when the article or document was published in newspapers, documents, or the like. The publication source 404 indicates a newspaper, a document name, or the like in which the article or document is published.

The article / document-event relationship table 410 stores an article or a document and data indicating the relationship between the events extracted from the article or document. The article / document-event relationship table 410 includes information on the article / document ID 411 and the event ID 412. The keys for identifying each record are the article / document ID 411 and the event ID 412.

The article / document ID 411 is an identifier for uniquely identifying the article or document. The event ID 412 is an identifier for uniquely identifying one of the events extracted from the article or the document.

5A and 5B show an example of the text analysis result DB133. The text analysis result DB 133 stores data related to the analysis result obtained by analyzing the meaning, part of speech, etc. of the word described in the article or document. The text analysis result DB 133 includes a word information table 500 (FIG. 5A) and a part of speech table 510 (FIG. 5B).

The word information table 500 stores data related to the meanings of words described in articles or documents. The word information table 500 includes information of a word ID 501, a word 502, a meaning vector 503, an appearance frequency 504, and an abstraction degree 505. The key for identifying each record is the word ID 501.

The word ID 501 is an identifier for uniquely identifying a word. The word 502 indicates the word itself. The meaning vector 503 expresses the meaning of a word from the text of an article or document with a low-dimensional vector such as 200 dimensions by using the "word embedding vector technology" established as a general technique. The frequency of appearance 504 indicates the frequency with which a word appears in an article, newspaper, or the like. The abstraction degree 505 indicates the abstraction degree of the word calculated from the appearance frequency or the abstraction degree of the word obtained from the vocabulary information providing service.

The part-speech table 510 stores data on the part-speech of a word extracted from an article or document. The part-speech table 510 is configured to include information about the word ID 511 and the part-speech 512. The key for identifying each record is the word ID 511.

The word ID 511 is an identifier for uniquely identifying a word. Part of speech 512 indicates the part of speech of a word such as a noun, verb, or s-irregular noun, and the part of speech of the word was extracted from the text of an article or document using the "morphological analysis technique" established as a general technique. It is a thing.

6A to 6C show an example of the management DB 134. The management DB 134 stores data related to threshold values, set values, user information, search history, and the like. The management DB 134 includes a setting table 600 (FIG. 6A), a user information table 610 (FIG. 6B), and a search history table 620 (FIG. 6C).

Data regarding a set value (for example, a threshold value) is stored in the setting table 600. The setting table 600 includes information on the setting ID 601, the setting content 602, and the setting value 603. The key for identifying each record is the setting ID 601.

The setting ID 601 is an identifier for uniquely identifying the setting value related to the system. Setting content 602 indicates the setting content. The setting value 603 indicates a setting value related to the setting content.

Data related to user information is stored in the user information table 610. The user information table 610 is configured to include information about the user ID 611, the matter of interest 612, and the degree of abstraction 613. The key for identifying each record is the user ID 611.

The user ID 611 is an identifier for uniquely identifying the user. Interest 612 is a set of keywords entered by the user and indicates what the user is interested in. The abstraction degree 613 indicates the abstraction degree of the user's interest, and is calculated from the abstraction degree of each word of the interest matter 612.

The search history table 620 stores data related to the history searched by the user. The search history table 620 is configured to include information of the user ID 621, the date and time 622, the search word 623, and the search attribute 624. The keys for identifying each record are the user ID 621 and the date and time 622.

The user ID 621 is an identifier for uniquely identifying the user. The date and time 622 indicates the date and time when the user searched. The search word 623 indicates the content searched by the user. The search attribute indicates the attribute (location, person / organization, date and time when the event occurred) of the search target specified when the user searches.

FIG. 7 shows an example of the event search screen (input screen) displayed in step S201 of FIG. The event search screen is a screen for the user to input information about the event to be searched for in order to analyze the transition of the future event, and is displayed on the input / output device 142 of the event presentation system 100. The event search screen includes a search word 701, a search attribute 702, a search button 703, and a past search 704.

The search word 701 is an input field for inputting a keyword related to the event to be searched. The search attribute 702 is an input field for narrowing down the events to be searched by place, person / organization, date and time. The search button 703 is a button for executing a search under the search conditions input by the search word 701 and the search attribute 702. By such a search screen, the condition of the search target of the event from the user is accepted. The past search 704 displays a list of search words 701 and search attributes 702 searched in the past.

In addition, not only one word but a plurality of words may be input to the search word 701, and in that case, AND condition, OR condition and the like may be set.

Further, not only one attribute but also a plurality of attributes may be input to the search attribute 702. In that case, the AND condition, the OR condition, and the like may be set.

Further, although the person / organization 333 acquired in step S210 has been described as the person / organization that executes the event, it is assumed that the person / organization of the object of the event is stored in the person / organization of the object of the event. The configuration may be acquired.

FIG. 8 shows an example of the search result display screen of the event displayed in step S211 of FIG. When displaying the event searched by the user, the event search result display screen also displays the events before and after due to the causal relationship, and is displayed on the input / output device 142 of the event presentation system 100.

The event search result display screen includes the corresponding event 801 and the search content 802, the article / document 803 related to the corresponding event, the article / document 804 of the previous event, the post-event (prediction result) 805, and the bird's-eye view map 806. It includes a person / organization relationship 807, a causal model 808, and a back button 809.

Corresponding event 801 indicates an event searched as an event to be searched. The search content 802 indicates a search word and a search attribute when the search is performed. Article / document 803 of the relevant event shows a list of articles / documents associated with the specific event of the relevant event 801. The article / document 804 of the previous event indicates an event determined to be a previous event with respect to the corresponding event based on a causal relationship, and indicates an article / document related thereto. The post-event (prediction result) 805 indicates an event that is determined to be a post-event with respect to the relevant event based on a causal relationship, and indicates an article / document related thereto. In FIG. 8, articles or documents for a predetermined period (for example, the latest one year) are targeted and displayed.

The bird's-eye view map 806 is a map of the area around the place where the event / document event that the user focused on in the article / document 803 of the relevant event 801 occurs (an example of information related to the information of the case specified by the user). Is shown. For example, when a country is entered in the search attribute, the map around the country is initially displayed. The person / organization relationship 807 indicates the structure of the person / organization to which the person / organization that executed the event of the article / document focused by the user belongs (an example of information related to the information of the case specified by the user). For example, when an organization is entered in the search attribute, the configuration of the organization (which may be an administrative division or the like) is initially displayed. The causal model 808 shows a causal model showing a causal relationship between the corresponding event 801 and the pre-event shown in the article / document 804 of the pre-event, the post-event shown in the post-event (prediction result) 805, and the post-event. The back button 809 is a button for the user to return to the event search screen (FIG. 7). Information on the event is presented to the user on such a screen.

Here, as a preparatory step for the event presentation process shown in FIG. 2, an example of a data generation process related to an abstract event stored in the database will be described with reference to FIGS. 9 and 10.

FIG. 9 shows an example of a flowchart relating to the first event abstraction process in which a plurality of similar events are integrated to abstract the events.

In step S900, the event presentation system 100 starts the first event abstraction process.

In step S901, the management data reception unit 117 receives the setting value of the level division of the abstraction degree of the event from the management user via the input / output device 142, and “sets the abstraction degree” of the setting content 602 of the setting table 600 of the management DB 134. It is stored in the set value 603 of "number".

In step S902, the abstraction degree determination unit 118 acquires the level division number of the word abstraction degree of the vocabulary information provision system 170 from the vocabulary information provision system 170, and the acquired level division number and the set value 603 stored in step S901. Compare with. If the abstraction degree determination unit 118 determines that the number of levels divided by the vocabulary information providing system 170 is larger, the process is transferred to step S903, and if it is determined that the number is not large, the process is transferred to step S904.

In step S903, the abstraction degree determination unit 118 integrates the number of level divisions according to the set value 603. For example, when the set value is "5" and the number of level divisions is "8", the abstraction degree determination unit 118 integrates the second and third level divisions and integrates the fourth and fifth levels. Then integrate the 6th and 7th. As a result, it is possible to match the degree of abstraction of the set value as the first, second, third, fourth, fifth, sixth, -7th, and eighth levels of the level division. There may be a plurality of patterns for the method of integration, but any pattern may be used. In this way, the abstraction degree determination unit 118 resets the abstraction degree of the word based on the level division number according to the abstraction degree of the set value, and sets the set abstraction degree as the word information of the text analysis result DB133. It is stored in the word 502 and the abstraction degree 505 of the table 500.

In step S904, the abstraction degree determination unit 118 allocates the number of level divisions to the plurality of abstraction degrees in duplicate with respect to the abstraction degree according to the set value 603. For example, when the set value is "10" and the level division number is "8", the abstraction degree determination unit 118 sets the fourth level division number to the fourth and fifth levels of the abstraction degree of the set value. Duplicate allocation, the 5th level division number is duplicated and assigned to the 6th and 7th levels of abstraction of the set value. As a result, it is set as 10 levels of the 1st, 2nd, 3rd, 4th, 4th (overlapping allocation), 5th, 5th (overlapping allocation), 6th, 7th, and 8th level division numbers. It can be adjusted to the level of abstraction of the value. There may be a plurality of patterns for overlapping, but any pattern may be used. In this way, the abstraction degree determination unit 118 resets the abstraction degree of the word based on the level division number according to the abstraction degree of the set value (duplicates are allowed as described above), and the set abstraction degree. Is stored in the word 502 and the abstraction degree 505 of the word information table 500 of the text analysis result DB 133.

In step S905, the text analysis unit 112 extracts words from the learning data (articles, documents, etc.) designated by the management user by the morphological analysis technique, extracts the part of speech of the words, and counts the frequency of appearance of the words. , Generate word meaning vectors with word embedding vector technology (learning). The text analysis unit 112 stores the extracted words, the meaning vector, and the appearance frequency in the word 502, the meaning vector 503, and the appearance frequency 504 of the word information table 500 of the text analysis result DB 133, assigns the word ID 501, and assigns the word ID 501 to the part of speech table. The same word ID is stored in the word ID 511 in 510, and the extracted part of speech is stored in the part of speech 512.

In step S906, the abstraction degree determination unit 118 sets the words having the highest appearance frequency 504 in the word information table 500 based on the setting value 603 of the "abstraction degree setting number" of the setting content 602 of the setting table 600 of the management DB 134. Group by the number of values 603 in order, and assign the degree of abstraction so that the degree of abstraction of the group with the highest frequency of occurrence increases. The abstraction degree assigned here is overwritten with the abstraction degree 505. In the process of step S906, the abstraction degree of the vocabulary information providing system 170 set in step S903 may be used as it is.

In step S907, the text analysis unit 112 extracts words from the event in the event list to be analyzed designated by the management user by the morphological analysis technique. In step S907, words may be extracted from all the events by the morphological analysis technique.

In step S908, the abstract event generation unit 121 classifies (clusters) words having similar meanings into the same category with respect to the sa-variant noun and the noun. More specifically, in the abstract event generation unit 121, the distance of the meaning vector 503 is relative in the record in which the word ID 501 of the word information table 500 matches with respect to the word ID 511 in which the part of speech 512 is the sa-variant noun in the part of speech table 510. Word ID 501 close to is classified into the same category as words having similar meanings. Similarly, in the record in which the word ID 501 of the word information table 500 matches with respect to the word ID 511 in which the part of speech 512 is a noun in the part of speech table 510, the word ID 501 having a relatively short distance of the meaning vector 503 is a word having a similar meaning. Classify into the same category as.

In step S909, the abstract event generation unit 121 searches for the word ID 501 within the threshold distance from the center of gravity of the category in the categories classified in step S908 based on the meaning vector 503, and among the searched words, the word ID 501 is searched. A word having an abstraction level 505 of the information table 500 one level higher (a word having an abstraction level one level higher than the word having the lowest abstraction level) is selected as a representative word.

In step S910, the abstract event generation unit 121 sets the representative word of the sa-variant noun as a word representing the action of the event, combines the representative word of the noun with the word representing the object of the event, and raises the combined event by one level. Generated as an abstract event of the degree of abstraction. The abstract event generation unit 121 assigns the event ID 301 to the generated abstract event information (event 302, word ID: noun 303, word ID: sa-variant noun 304, abstraction degree 305) and registers it in the event table 300. Further, the abstract event generation unit 121 has one abstraction degree 505 of the word information table 500 in the category to which the representative word of the sa-variant noun belongs, which is the original word of the sa-variant noun that is the source of the abstract event of the next higher abstraction degree. The other word below) and the original word of the noun that is the source of the abstract event of the next higher abstraction level (the other word below the abstraction level 505 of the word information table 500 in the category to which the representative word of the noun belongs) The event that is combined with the word) is generated as a concrete event. The abstract event generation unit 121 assigns the event ID 301 to the generated concrete event information (event 302, word ID: noun 303, word ID: sa-variant noun 304, abstraction degree 305) and registers it in the event table 300. Then, the abstract event generation unit 121 associates the generated abstract event of the higher degree of abstraction with the concrete event and stores it in the concrete event ID 322 and the abstract event ID 323 of the abstract event relation table 320 of the causal model DB 131. ..

In step S911, the causal relationship assigning unit 122 assigns a causal relationship between the abstract events of the next higher degree of abstraction based on the causal relationship between the concrete events. In the assignment of the causal relationship, a new link ID 311 is generated in the causal table 310 of the causal model DB 131, and the pre-event ID 312 and the post-event ID 313 are registered as the events for which the causal relationship is assigned. It is assumed that the causal relationship between the lowest abstraction events (concrete events) is preset by the user.

In step S912, the abstract event generation unit 121 classifies words of abstract events having similar meanings into the same category. More specifically, the abstract event generation unit 121 has a meaning in the record in which the word ID 501 of the word information table 500 matches with respect to the word ID 511 of the sa-variant noun indicating the operation of the abstract event generated in step S910 or step S914. Word ID 501, which is relatively close to each other in the vector 503, is classified into the same category as words having similar meanings. Similarly, with respect to the noun word ID 511 that indicates the object of the abstract event generated in step S910 or step S914, the words whose meaning vectors 503 are relatively close to each other in the records in which the word ID 501 of the word information table 500 matches. ID 501 is classified into the same category as words having similar meanings.

In step S913, the abstract event generation unit 121 searches for the word ID 501 within the threshold distance from the center of gravity of the category in the categories classified in step S912 based on the meaning vector 503, and among the searched words, the word ID 501 is searched. A word whose abstraction level 505 of the information table 500 is one higher is selected as a representative word.

In step S914, the abstract event generation unit 121 sets the representative word of the sa-variant noun as a word representing the action of the event, combines the representative word of the noun with the word representing the object of the event, and raises the combined event by one level. Generated as an abstract event of the degree of abstraction. The abstract event generation unit 121 assigns the event ID 301 to the generated abstract event information (event 302, word ID: noun 303, word ID: sa-variant noun 304, abstraction degree 305) and registers it in the event table 300. Further, the abstract event generation unit 121 has one abstraction degree 505 of the word information table 500 in the category to which the representative word of the sa-variant noun belongs, which is the original word of the sa-variant noun that is the source of the abstract event of the next higher abstraction degree. The other word below) and the original word of the noun that is the source of the abstract event of the next higher abstraction level (the other word below the abstraction level 505 of the word information table 500 in the category to which the representative word of the noun belongs) The original event combined with the word) is generated as a concrete event. The abstract event generation unit 121 assigns the event ID 301 to the generated concrete event information (event 302, word ID: noun 303, word ID: sa-variant noun 304, abstraction degree 305) and registers it in the event table 300. Then, the abstract event generation unit 121 associates the generated abstract event of the higher degree of abstraction with the concrete event, and stores it in the concrete event ID 322 and the abstract event ID 323 of the abstract event relation table 320 of the causal model DB 131.

In step S915, the causal relationship assigning unit 122 assigns a causal relationship between the abstract events of the next higher level of abstraction, as in step S911. The causal relationship between the concrete events of the abstract event of the next higher level of abstraction is set in step S911 or step S915.

In step S916, the abstract event generation unit 121 compares the number of repetitions of the processes from step S912 to step S915 with the set value 603 of the "abstraction degree setting number" of the setting content 602 of the setting table 600 of the management DB 134. When the abstract event generation unit 121 determines that the number of repetitions is small, the process is transferred to step S912, and when it is determined that the number of repetitions is the set value 603 or more, the process is transferred to step S917.

In step S917, the event presentation system 100 ends the first event abstraction process.

In step S905, an example of learning articles, documents, etc. specified by the management user has been described. However, the event presentation system 100 automatically searches the Internet, the in-house system, etc., and searches for new articles, documents, etc. It may be acquired automatically and learned automatically.

Further, in step S907, the word is extracted from the event list of the analysis target designated by the management user, but the event is a combination of the sa-variant noun that can be the action of the event and the noun that can be the object of the event. Is generated, and the word used for generating the event may be extracted as a word.

FIG. 10 shows an example of a flowchart relating to a second event abstraction process that abstracts an event based on the strength of the causal relationship and the small number of branches. The event abstraction may be performed by the first event abstraction process, the second event abstraction process, or both the first and second event abstraction processes. You may be abstracted.

In step S1000, the event presentation system 100 starts the second event abstraction process.

In step S1001, the abstract event generation unit 121 acquires the causal table 310 of the causal model DB 131.

In step S1002, the causal strength / branch number determination unit 123 acquires the strength 314 of each link ID 311 in the causal table 310 acquired in step S1001.

In step S1003, the causal intensity / branch number determination unit 123 counts the number of records of the post-event ID 313 associated with the pre-event ID 312 in the causal table 310 acquired in step S1001, and changes from the pre-event to the post-event. Obtained as the number of branches of.

In step S1004, the causal intensity / number of branches determination unit 123 determines the strength of the causal relationship 314 and the number of branches as threshold values. Therefore, in the causal strength / branch number determination unit 123, the setting content 602 of the setting table 600 of the management DB 134 sets the set value 603 of the “high threshold value of the link strength” and the “low threshold value of the link strength” to the strength 314 of the causal relationship. The setting content 602 acquires the set values 603 of the "high threshold of the number of branches" and the "low threshold of the number of branches" as the threshold of the number of branches. The causal intensity / branch number determination unit 123 determines based on these acquired threshold values, the causal relationship intensity 314 acquired in step S1002, and the number of branches acquired in step S1003.

When the causal strength / number of branches determination unit 123 determines that the strength 314 is larger than the low threshold value of the link strength and the number of branches is smaller than the low threshold value of the number of branches, the process proceeds to step S1006. When the causal strength / number of branches determination unit 123 determines that the strength 314 is larger than the high threshold value of the link strength and the number of branches is smaller than the high threshold value of the number of branches, the process proceeds to step S1005. When the causal strength / branch number determination unit 123 determines that the above is other than the above, the process shifts to step S1008.

In step S1005, the causal intensity / branch number determination unit 123 duplicates an event dummy so that one event can be integrated into a plurality of abstract events. Dummy events are represented by the branch number of the event ID. According to the determination in step S1004, even if the number of branches is larger than the low threshold value and abstracted, the event will span a plurality of abstract events. However, since the intensity 314 is larger than the high threshold value, it is different from other events. It is judged that integration is also possible.

In step S1006, the abstract event generation unit 121 has a strong causal relationship between the events and the number of branches to other events is small. Therefore, it is assumed that these events form one mass between the events, and these events are regarded as abstract events (1). Integrate into one mass).

In step S1007, the abstract event generation unit 121 selects the event at the center between the events as the name of the abstract event in the abstract event integrated in step S1006. More specifically, in the causal model DB 131, the abstract event generation unit 121 stores the non-centered event in the abstract event in the concrete event ID 322 of the abstract event relation table 320, and stores the central abstract event in the abstract event ID 323. Store. Further, the abstract event generation unit 121 generates a record in which the abstraction degree 305 of the event ID 301 of the event table 300 that matches the abstract event ID 323 is increased by one. Further, the abstract event generation unit 121 registers the event ID (branch number of the original event) of the dummy event generated in step S1005 in the event ID 301 of the event table 300. For the other items of the event table 300 (event 302, word ID: noun 303, word ID: sa-variant noun 304, abstraction degree 305), the data of the original event is copied.

In step S1008, the abstract event generation unit 121 determines whether or not the processing has been completed for all the events. When the abstract event generation unit 121 determines that there is an unprocessed event, the process is transferred to step S1002, and when it is determined that all the events are processed, the process is transferred to step S1009.

In step S1009, the event presentation system 100 ends the second event abstraction process.

Here, in the process of FIG. 2, the degree of abstraction of the event presented to the user is determined based on the search word input by the user. In addition to or in place, the level of abstraction suitable for the user may be predetermined from the user's interests. In addition, the event of the degree of abstraction determined from the matter of interest and the causal relationship may also be presented to the user. An example of the processing for that purpose is shown in FIG.

FIG. 11 shows an example of a flowchart relating to a user abstraction degree setting process in which the user's abstraction degree is determined and set based on the user's interests.

In step S1100, the event presentation system 100 starts the user abstraction degree setting process.

In step S1101, the abstraction degree determination unit 118 receives the input of the user's interests from the user at the input / output device 142. The input of the user's interests may be a search word in the event presentation system 100, a search word when a target web page, information, or the like is searched on the Internet, or the like. It may be an input of.

In step S1102, the abstraction degree determination unit 118 extracts words of interest, lists them for each user, and accumulates them. More specifically, the abstraction degree determination unit 118 divides the interests input by the user in step S1101 into words by the morphological analysis technique, and stores each word in the interests 612 of the user information table 610 of the management DB 134. .. At that time, the abstraction degree determination unit 118 stores the user ID 611 in association with the user ID 611. The abstraction degree determination unit 118 may register the input of the user's interests as the search history in the search history table 620.

In step S1103, the abstraction degree determination unit 118 determines whether the abstraction degree can be obtained for the word of interest 612 (word in the list) in each user ID 611, and the word 502 of the word information table 500 of the text analysis result DB 133. And the abstraction degree 505 are searched for. If the abstraction degree determination unit 118 determines that the acquisition is possible, the process is transferred to step S1104, and if it is determined that the acquisition is not possible, the process is transferred to step S1105.

In step S1104, the abstraction degree determination unit 118 acquires the abstraction degree 505 searched in step S1103.

In step S1105, the abstraction degree determination unit 118 determines whether or not there are other words of interest 612 in the user ID 611. If the abstraction degree determination unit 118 determines that there is another, the process is transferred to step S1103, and if it is determined that there is no other, the process is transferred to step S1106.

In step S1106, the abstraction degree determination unit 118 calculates the average value of the abstraction degree 505 acquired in step S1104, and sets the integer obtained by rounding off the decimal point of the average value as the abstraction level of the user ID 611 in the abstraction degree 613 of the user information table 610. Store. In the calculation of the user's abstraction level, the average value has been described as an example, but representative values such as the median value and the mode value may be used.

In step S1107, the abstraction degree determination unit 118 determines whether or not there is another user who has not set the abstraction degree 613 with the user ID 611. When the abstraction degree determination unit 118 determines that there is another user, the process is transferred to step S1003, and when it is determined that there is no other user, the process is transferred to step S1108.

In step S1108, the event presentation system 100 ends the user abstraction degree setting process.

As described above, the event presentation system 100 abstracts an event by grouping a plurality of similar events into one group and assigning a word sentence expressing a meaning to the group. It also gives a causal relationship to the abstracted event. According to such a configuration, the abstracted event and its causal relationship can be presented to the user, and the user can easily analyze the transition of the event at the abstracted level.

Further, the event presentation system 100 abstracts an event with various degrees of abstraction and imparts a causal relationship thereof. Therefore, when presenting to the user, it is possible to present the event and the causal relationship of the degree of abstraction suitable for the purpose of the user.

(2) Other Embodiments In the above-described embodiment, the case where the present invention is applied to the event presentation system 100 has been described, but the present invention is not limited to this, and various other systems. It can be widely applied to event presentation devices and event presentation methods.

Further, in the above-described embodiment, as an example of outputting the search result, a case of displaying as a search result display screen has been described, but the present invention is not limited to this, and the search result may be printed on paper or the like. Alternatively, the search result may be output as voice, the search result may be sent by e-mail as a file, or other output may be performed.

Further, in the above-described embodiment, various data have been described using the XX table for convenience of explanation, but the data structure is not limited and may be expressed as XX information or the like.

In addition, the information such as programs, tables, and files that realize each function in the above description is a memory, a hard disk, a storage device such as an SSD (Solid State Drive), or a recording medium such as an IC card, an SD card, or a DVD. Can be placed in. In addition, the control lines and information lines indicate those that are considered necessary for explanation, and do not necessarily indicate all the control lines and information lines in the product.

Further, the above-mentioned configuration may be appropriately changed, rearranged, combined, or omitted as long as it does not exceed the gist of the present invention.

100 ... event presentation system, 110 ... main memory, 111 ... search word reception unit, 112 ... text analysis unit, 113 ... causal model / event search / acquisition unit, 114 ... article / document search -Acquisition unit, 115 ... Attribute information search / acquisition unit, 116 ... Analysis result output unit, 117 ... Management data reception unit, 118 ... Abstract degree determination unit, 120 ... Analysis target acquisition unit, 121 ... Abstract event generation unit, 122 ... causal relationship assignment unit, 123 ... causal strength / branch number determination unit, 130 ... auxiliary storage device, 140 ... central control device, 141 ... communication device, 142 ... input / output device , 150 ... communication line, 160 ... map information providing system, 170 ... vocabulary information providing system.

Claims (6)

From the event information including the event information, the event corresponding to the character string input for the search is acquired as an abstract event, and based on the abstract event, the context information associated with the context of the event is used as described above. The first specific part that identifies the event after the abstract event,
Based on the abstract event acquired in the first specific part, the concrete event indicating the child event of the abstract event is specified from the parent-child relationship information associated with the parent-child relationship of the event, and the information of the case related to the event is obtained. From the included case information, the second specific part that identifies the information of the case related to the concrete event, and
Cases relating to the abstract event acquired in the first specific part, the event after the abstract event specified in the first specific part, and the concrete event of the abstract event specified in the second specific part. Information, an output section that outputs information,
An event presentation system characterized by being equipped with. Based on the acquired abstract event, the first specific unit identifies the event before the abstract event from the context information.
The second specific unit identifies the concrete event of the previous event from the parent-child relationship information based on the event before the abstract event specified in the first specific unit, and from the case information, the previous event. Identify the information of the case related to the concrete event of
The output unit outputs the event before the abstract event specified by the first specific unit and the information of the case related to the concrete event of the previous event specified by the second specific unit.
The event presentation system according to claim 1. When the abstraction degree of the user and the abstraction degree of the acquired abstract event are different from each other, the first specific unit identifies the parent event or the child event of the abstract event from the parent-child relationship information, and sets the specified event. Based on this, the event after the event is identified from the context information, and the event is identified.
The output unit outputs the event specified by the first specific unit and the event after the event.
The event presentation system according to claim 1. The second specific unit narrows down the information of the case related to the specified concrete event based on the character string indicating the attribute of the search and the narrowing down information to which the event and the attribute are associated.
The event presentation system according to claim 1. It has a third specific part that identifies information related to the case information specified by the user.
The output unit outputs the related information specified by the third specific unit.
The event presentation system according to claim 1. From the event information including the event information, the event corresponding to the character string input for the search is acquired as an abstract event, and based on the abstract event, the context information associated with the context of the event is used as described above. The first specific part that identifies the event after the abstract event,
Based on the abstract event acquired in the first specific part, the concrete event indicating the child event of the abstract event is specified from the parent-child relationship information associated with the parent-child relationship of the event, and the information of the case related to the event is obtained. From the included case information, the second specific part that identifies the information of the case related to the concrete event, and
Cases relating to the abstract event acquired in the first specific part, the event after the abstract event specified in the first specific part, and the concrete event of the abstract event specified in the second specific part. Information, an output section that outputs information,
An event presentation device comprising.

Images