JP7265110B2 - Conversation support system and conversation support program - Google Patents

Description

The present invention relates to a conversation support system and a conversation support program that support user conversation.

As a conventional technique, when an individual inputs a sentence containing words that lead to satisfaction such as interest, preference, hope, goal, purpose, expectation such as preference, want, and want, morphological analysis is performed on this sentence. It is known to acquire value indicators and value judgment conditions that serve as criteria for judging personal value by performing natural language processing based on dependency analysis and part-of-speech decomposition (see, for example, Patent Document 1). In the technology described in Patent Document 1, based on the coincidence or similarity of value indicators and value judgment conditions between individuals belonging to an organization, the similarity of value between these individuals is determined, and the value of this organization is determined. I can provide you with the information you have.

Japanese Patent No. 6191892

Normally, in a conversation, even if the speaker does not use words that lead to satisfaction, the listener may predict the speaker's interest from the speaker's utterances and expand the conversation. For example, when the speaker says "I'm from XX", the listener may expand the conversation on the topic of the place "XX" included in the utterance.

However, in the technique described in Patent Literature 1, only words that are related to words indicating the speaker's interest such as liking, wanting, and wanting can be acquired as value indicators. There is a problem that a word that is not related to the indicated word but is likely to be of interest to the speaker cannot be acquired as a value index, and appropriate information cannot be provided.

SUMMARY OF THE INVENTION Accordingly, it is an object of the present invention to provide a conversation support system and a conversation support program capable of appropriately providing information that is highly likely to be of interest to both the user and the user's partner. .

The conversation support system of the present invention includes an interesting word extracting unit that extracts interesting words, which are words that are highly likely to be of interest to the speaker, from the speech record, and a common word extracting unit that is words that are highly likely to be of interest to a plurality of people. a common word-of-interest specifying unit that specifies words of interest; and an information providing unit that provides a user with information related to the common words of interest specified by the common words-of-interest specifying unit. The interesting word extracted from the speech record is registered in the storage unit for each speaker, and the common interesting word specifying unit determines whether the interesting word registered for each of the plurality of persons in the storage unit is If it matches or is similar, the interesting word is specified as the common interesting word, and the interesting word extracting unit extracts an extraction word list, which is a list of words for extracting the interesting word from the speech record. When the word list contains the morpheme of the utterance content contained in the utterance record, or the root form of this morpheme, the interesting word is specified from the word that is the source of the morpheme.

With this configuration, the conversation support system of the present invention can extract morphemes of the utterance content contained in the utterance record, or, if the original form of this morpheme is contained in the word list for extraction, the morpheme is a word that indicates the speaker's interest. Even if there is no such morpheme, the interesting word is identified from the word that is the source of the dependency for this morpheme, so it is possible to appropriately provide information that is highly likely to be of interest to both the user and the user's partner.

In the conversation support system of the present invention, the common word-of-interest identification unit weights each of the common words of interest, and the information provision unit distributes the information to be provided to the user by the common-interest word identification unit. The selection may be based on the weighting value performed for each common-interest word.

With this configuration, the conversation support system of the present invention selects information to be provided to the user based on the weighting value executed for each common interest word related to this information. It is possible to appropriately provide information that is highly likely to be of interest to both the other party.

In the conversation support system of the present invention, the information provision unit acquires a summary sentence of the information related to the common interest word identified by the common interest word identification unit from an external information providing site, and prepares the summary sentence. A positive/negative analysis may be performed on the information, and the information to be provided to the user may be selected based on the degree of positive/negative that is the result of the positive/negative analysis.

With this configuration, the conversation support system of the present invention selects information to be provided to the user based on the positive/negative degree, which is the result of positive/negative analysis of the summary text of the information obtained from the information providing site. , it is possible to appropriately provide information with positive content as information that is highly likely to be of interest to both the user and the user's partner.

The conversation support program of the present invention includes an interesting word extraction unit that extracts interesting words, which are words that are highly likely to be of interest to a speaker, from a speech record, and A computer implements a common word-of-interest specifying unit that specifies the words of interest, and an information providing unit that provides a user with information related to the common words of interest specified by the common words-of-interest specifying unit, and extracts the words of interest. The unit registers the interesting words extracted from the utterance record in the storage unit for each speaker, and the common interesting word specifying unit stores the words registered for each of the plurality of persons in the storage unit. If the words of interest match or are similar to each other, the words of interest are specified as the common words of interest, and the words of interest extraction unit extracts the words of interest from a list of words for extraction from the transcript. When a word list for extraction contains a morpheme of the statement content contained in the statement record or the original form of this morpheme, the interesting word is specified from a word that is a dependency of the morpheme. .

With this configuration, the computer executing the conversation support program of the present invention can extract the morpheme of the utterance content contained in the utterance record, or, if the root form of this morpheme is contained in the word list for extraction, the morpheme of interest to the speaker. Even if it is not a word that indicates , the interesting word is identified from the word that is the source of the dependency for this morpheme, so that information that is highly likely to be of interest to both the user and the user's counterpart is appropriately provided. be able to.

The conversation support system and conversation support program of the present invention can appropriately provide information that is highly likely to be of interest to both the user and the user's partner.

1 is a block diagram of a system according to one embodiment of the invention; FIG. 2 is a block diagram of the conversation support system shown in FIG. 1 when configured by one computer; FIG. FIG. 3 is a diagram showing an example of an extraction word list shown in FIG. 2; FIG. 2 is a diagram showing an example of information registered in an interest database shown in FIG. 1; FIG. 2 is a block diagram of a user terminal shown in FIG. 1; FIG. 2 is a sequence diagram of the operation of the system shown in FIG. 1 when an interest word is registered in an interest database; FIG. FIG. 7 is a flowchart of an interesting word extraction process shown in FIG. 6; FIG. 8 is a diagram showing an example of a morphological analysis result in the interesting word extraction process shown in FIG. 7; FIG. FIG. 8 is a diagram showing an example of a result of dependency analysis in the interesting word extraction process shown in FIG. 7; 2 is a sequence diagram of the operation of the system shown in FIG. 1 when providing topic candidates to a user; FIG. FIG. 11 is a flowchart of a common-interest-word specifying process shown in FIG. 10; FIG. FIG. 11 is a flowchart of topic candidate selection processing shown in FIG. 10 ; FIG. 11 is a flowchart of notification data generation processing shown in FIG. 10; 6 is a diagram showing an example of a screen displayed on the display unit shown in FIG. 5 by a topic candidate display application based on notification data; FIG. 3 is a flowchart of the operation of the conversation support system shown in FIG. 2 when determining whether or not to start processing for providing topic candidates to the user; 16 is a sequence diagram of the operation of the system when the conversation support system starts the process of providing topic candidates to the user by the operation shown in FIG. 15; FIG. 2 is a sequence diagram of the operation of the system shown in FIG. 1 when providing topic candidates to the user without using the schedule system; FIG. FIG. 18 is a flowchart of a common-interest-word specifying process shown in FIG. 17; FIG.

BEST MODE FOR CARRYING OUT THE INVENTION Hereinafter, embodiments of the present invention will be described with reference to the drawings.

First, the configuration of a system according to one embodiment of the present invention will be described.

FIG. 1 is a block diagram of system 10 according to the present embodiment.

As shown in FIG. 1, the system 10 includes a conversation support system 20 for supporting user conversation. The conversation support system 20 may be composed of one computer or may be composed of a plurality of computers.

The system 10 includes a statement database 30 that manages statement records, such as conversation records that are records of conversations. The remarks managed by the remarks database 30 are, for example, records of one or more speakers and the contents of remarks made by the speakers in meetings, business negotiations, and the like. The speech record database 30 may be composed of one computer, or may be composed of a plurality of computers.

The system 10 has an interest database 40 as a storage unit that manages words that are likely to be of interest to a specific person (hereinafter referred to as "interest words"). The interest database 40 may be composed of one computer or may be composed of a plurality of computers.

The system 10 includes a schedule system 50 that manages the user's schedule. In the schedule system 50, a schedule application for managing the user's schedule is executed. The schedule system 50 may be composed of one computer or may be composed of a plurality of computers.

The system 10 has a customer information database 60 that manages customer information of users. The customer information database 60 may be composed of one computer, or may be composed of a plurality of computers. The information managed by the customer information database 60 is, for example, the company location and company name of the user's customer.

The system 10 includes an information providing site 70 such as a news site that provides information. The information providing site 70 may be composed of one computer, or may be composed of a plurality of computers. The system 10 can have at least one information providing site that provides information other than the information providing site 70 .

The system 10 has a user terminal 80 used by a user. In addition to the user terminal 80, the system 10 can include at least one user terminal having the same configuration as the user terminal 80. FIG.

Conversation support system 20, statement database 30, interest database 40, schedule system 50, customer information database 60, information providing site or user terminal communicate via a network such as a LAN (Local Area Network) or the Internet. It is possible.

FIG. 2 is a block diagram of conversation support system 20 when configured by a single computer.

The conversation support system 20 shown in FIG. 2 includes an operation unit 21, which is an operation device such as a keyboard and a mouse for inputting various operations, and a display device such as an LCD (Liquid Crystal Display) for displaying various information. A certain display unit 22, a communication unit 23 which is a communication device that communicates with an external device via a network such as a LAN or the Internet, or directly by wire or wirelessly without a network, and stores various information. A non-volatile storage device 24 such as a semiconductor memory, HDD (Hard Disk Drive), or the like, and a control unit 25 for controlling the conversation support system 20 as a whole.

The storage device 24 stores a conversation support program 24a for supporting user's conversation. For example, the conversation support program 24a may be installed in the conversation support system 20 at the manufacturing stage of the conversation support system 20, or may be stored in a CD (Compact Disk), a DVD (Digital Versatile Disk), a USB (Universal Serial Bus) memory, or the like. It may be additionally installed in the conversation support system 20 from an external storage medium, or may be additionally installed in the conversation support system 20 from the network.

The storage device 24 stores a list (hereinafter referred to as an "extraction word list") 24b of words that are likely to be of interest to a person, that is, words for extracting interesting words (hereinafter referred to as "extraction words"). Remember.

FIG. 3 is a diagram showing an example of the extraction word list 24b.

As shown in FIG. 3, the extraction word list 24b includes the original form of the extraction word for each representative word representing a plurality of extraction words. For example, in FIG. 3, "Interest" indicating "interest", "LivingPlace" indicating "place of residence", "BirthPlace" indicating "hometown", and "Visit" indicating "visited place" are representative words. is. "Interest", "Interest", "Distraction", "Attention", "Interest", "Entertainment", "Interest", and "Like" are each the archetype of the word for extraction represented by "Interest". . "To live", "to live", and "to live" are the original forms of words for extraction represented by "LivingPlace". "Birthplace", "born", "source", "place of birth", "origin", "hometown", "hometown", and "hometown" are each the original form of the word for extraction represented by "BirthPlace". be. "Visit", "business trip", and "trip" are the original forms of extraction words represented by "Visit".

As shown in FIG. 2, the storage device 24 stores a synonym dictionary 24c showing synonyms. The synonym dictionary 24c contains representative words representing a plurality of synonyms. The control unit 25 determines one or more words selected by the user from a list of words such as the synonym dictionary 24c and a synonym dictionary provided in a system external to the conversation support system 20 as extraction words, Synonyms of the determined extraction words are automatically determined as extraction words based on the synonym dictionary 24c and a synonym dictionary provided in a system external to the conversation support system 20, and representative words representing the determined extraction words are selected. The extraction word list 24b may be generated by automatic determination based on the synonym dictionary 24c.

The control unit 25 includes, for example, a CPU (Central Processing Unit), a ROM (Read Only Memory) storing programs and various data, and a RAM (Random Memory) as a memory used as a work area for the CPU of the control unit 25. Access Memory). The CPU of the control unit 25 executes programs stored in the storage device 24 or the ROM of the control unit 25 .

By executing the conversation support program 24a, the control unit 25 extracts words that are highly likely to be of interest to the speaker, that is, interesting words from the speech record, and an interesting word extraction unit 25a that extracts words that are of interest to a plurality of persons. A common interest word specifying unit 25b that specifies common interest words that are highly likely words, and topic candidates are provided to the user as information related to the common interest words specified by the common interest word specifying unit 25b. An information providing unit 25c is realized.

FIG. 4 is a diagram showing an example of information registered in the interest database 40. As shown in FIG.

As shown in FIG. 4, the interest database 40 contains the names of speakers, representative words of extraction words, interesting words corresponding to these representative words, and dates and times when these interesting words were registered (hereinafter referred to as "registration dates and times"). ) are stored in association with each other.

FIG. 5 is a block diagram of the user terminal 80. As shown in FIG.

A user terminal 80 shown in FIG. 5 includes an operation unit 81, which is an operation device such as a keyboard and a mouse, for inputting various operations, and a display unit 82, which is a display device such as an LCD for displaying various information. A communication unit 83, which is a communication device that communicates with an external device via a network such as a LAN, the Internet, or directly by wire or wirelessly without a network; It has a non-volatile storage device 84 such as an HDD, and a control unit 85 that controls the user terminal 80 as a whole.

The storage device 84 stores a program 84a for an application for displaying topic candidates provided by the conversation support system 20 (hereinafter referred to as "topic candidate display application"). Examples of topic candidate display applications include chat tools, message applications, SNSs (Social Networking Services), and e-mails. The topic candidate display application may be a web application.

The control unit 85 includes, for example, a CPU, a ROM that stores programs and various data, and a RAM that serves as a work area for the CPU of the control unit 85 . The CPU of the control unit 85 executes programs stored in the storage device 84 or the ROM of the control unit 85 .

The control unit 85 can request the conversation support system 20 to provide topic candidates via the topic candidate display application.

The operation of system 10 will now be described.

First, the operation of the system 10 when an interest word is registered in the interest database 40 will be described.

FIG. 6 is a sequence diagram of the operation of system 10 when an interest word is registered in interest database 40. As shown in FIG.

A user can register a statement in the statement database 30 via a computer such as the user terminal 80, for example. The utterance record registered in the utterance record database 30 may be one automatically transcribed by a speech recognition system capable of converting voice into text, or it may be the text of the speaker and the utterance content of the speaker. may be manually input by the user. When a new utterance record is registered, the utterance record database 30 notifies the conversation support system 20 that the new utterance record has been registered (S101). This notification includes the newly registered transcript itself.

Upon receiving the notification in S101, the interesting word extraction unit 25a of the conversation support system 20 analyzes the statement included in the notification in S101 to identify the speaker described in the statement and the statement of the speaker. content for each speaker (S102). Any method can be adopted as the acquisition method in S102. For example, the conversation support system 20 is made to read a format in which the name of the speaker and the contents of the statement of the speaker are described in which part of the statement, and based on this format, the interesting word extractor. A method of obtaining the speaker and the utterance content of the speaker from the utterance record for each speaker by 25a may be used. Also, for example, "one or more kanji characters before 'san' or 'sama' at the beginning of each sentence is the speaker's name, and after the opening 'san' or 'sama' is the speaker's statement. A method in which a user sets a rule such as "is the content" in the conversation support system 20, and according to the rule, the interesting word extraction unit 25a acquires the speaker and the utterance content of the speaker from the utterance record for each speaker. But it's okay.

After the process of S102, the interesting word extracting unit 25a executes an interesting word extracting process of extracting words that are likely to be of interest to the speaker, that is, interesting words from the utterance content acquired in S102 (S103). .

FIG. 7 is a flowchart of the interesting word extraction process in S103.

As shown in FIG. 7, the interesting word extraction unit 25a targets only one statement content that has not yet been targeted in the processing shown in FIG. 7, among the statement content acquired in S102 (S121).

Next, the interesting word extracting unit 25a selects only one sentence that has not yet been subjected to the processing shown in FIG. 7 among the sentences included in the current target statement content (S122).

Next, the interesting word extraction unit 25a performs morphological analysis, which is generalized as a natural language processing technology, on the sentence of the current target, and converts the morpheme, which is the smallest unit having meaning in language, into the current target sentence. Obtain the morphemes of the current target sentence by dividing the sentence (S123). For example, if the current target sentence is "I am from Hokkaido.", the interesting word extraction unit 25a can obtain the result shown in FIG. 8 by the processing of S123. In the case of Japanese, for example, Mecab (http://taku910.github.io/mecab/) exists as open source software for morphological analysis.

After the process of S123, the interesting word extraction unit 25a performs dependency analysis, which is a generalized natural language processing technique, on the current target sentence. It is analyzed which of the acquired morphemes is dependent (S124). For example, if the current target sentence is "I am from Hokkaido.", the interesting word extraction unit 25a can obtain the result shown in FIG. 9 by the processing of S124. In the case of Japanese, for example, CaboCha (https://taku910.github.io/cabocha/) exists as open source software for dependency parsing. In FIG. 9, for example, the morpheme "Hokkaido" with the line number "2" is dependent on the morpheme "from" with the line number "4".

After the process of S124, the interesting word extraction unit 25a selects only one morpheme that has not yet been processed in the process shown in FIG. ).

Next, the interesting word extraction unit 25a determines whether or not the original form of the current target morpheme is included in the extraction word list 24b as an extraction word (S126). For example, in the result shown in FIG. 9, "from" is included in the extraction word list 24b.

When the interesting word extracting unit 25a determines in S126 that the original form of the current target morpheme is included in the extraction word list 24b as an extraction word, the interesting word extraction unit 25a extracts a dependency from the current target morpheme other than the current target morpheme. is specified based on the result of the processing in S124 (S127). For example, in the result shown in FIG. 9, words other than "from" that are dependent on "from" are "watashi", "Hokkaido", "desu" and ".".

After the process of S127, the interesting word extracting unit 25a specifies, from among the words specified in S127, words of a part of speech such as a proper noun that clearly indicates what the word indicates, that is, an interesting word (S128). . For example, when the interesting word extracting unit 25a identifies "I", "Hokkaido", "desu" and "." in S127, it identifies the proper noun "Hokkaido" as an interesting word in S128.

If the interesting word extracting unit 25a determines in S126 that the original form of the current target morpheme is not included in the extraction word list 24b as an extraction word, or if the process of S128 is executed, the interesting word extraction unit 25a extracts Among the morphemes acquired in S123, it is determined whether or not there is a morpheme that has not yet been targeted in the processing shown in FIG. 7 (S129).

If the interesting word extraction unit 25a determines in S129 that there is a morpheme that has not yet been targeted in the processing shown in FIG.

If the interesting word extraction unit 25a determines in S129 that there is no morpheme that has not yet been targeted in the processing shown in FIG. 7 among the morphemes acquired in S123 for the current target sentence, It is determined whether or not there is a sentence that has not yet been processed in the processing shown in FIG. 7 (S130).

If the interesting word extraction unit 25a determines in S130 that there is a sentence that has not yet been targeted in the processing shown in FIG.

If the interesting word extraction unit 25a determines in S130 that there is no sentence included in the current target statement content that has not yet been targeted in the processing shown in FIG. It is determined whether or not there is statement content that has not yet been targeted in the processing shown in FIG. 7 (S131).

If the interesting word extracting unit 25a determines in S131 that there is statement content that has not yet been targeted in the processing shown in FIG. 7 among the statement content acquired in S102, it executes the processing of S121.

If the interesting word extraction unit 25a determines in S131 that there is no statement content that has not yet been targeted in the processing shown in FIG. 7 among the statement content acquired in S102, the processing shown in FIG. 7 ends.

As shown in FIG. 6, after the interesting word extraction process of S103, the interesting word extraction unit 25a performs are associated with the representative word set in the extraction word list 24b, the name of the speaker whose speech content includes this interesting word, and the current date and time as the registration date and time, and specified in the interesting word extraction process of S103. Each interesting word is registered in the interest database 40 (S104). For example, if the sentence “I am from Hokkaido” is uttered by a speaker named “Taro Yamada”, the interesting word extraction unit 25a extracts “Hokkaido” as the interesting word and “Hokkaido” as the representative word. "BirthPlace", "Taro Yamada" as the name of the speaker, and the current date and time as the registration date and time are associated with each other and registered in the interest database 40. FIG.

Note that the interesting word extraction unit 25a may switch between additionally registering the interesting word and overwriting the registered interesting word depending on the type of the representative word. For example, when the interesting word extraction unit 25a registers interesting words associated with representative words "LivngPlace" and "Interest" in the interest database 40 a plurality of times for the same speaker's name, each of the interesting words can be additionally registered. On the other hand, when the interesting word extraction unit 25a registers the interesting word associated with the representative word "BirthPlace" for the same speaker's name a plurality of times in the interest database 40, The interesting word and registration date and time previously registered in association may be overwritten with the interesting word and registration date and time associated with "BirthPlace" as the representative word and registered later.

In the above description, the operation of the system 10 when an interest word is registered in the interest database 40 when an utterance record is registered in the utterance record database 30 has been described. However, interesting words can be registered in the interest database 40 even in cases other than when the statement is registered in the statement database 30 . For example, the user can directly register information including interesting words in the interest database 40 via a computer such as the user terminal 80 . The user can also change or delete information registered in the interest database 40 via a computer such as the user terminal 80 .

Next, the operation of the system 10 when providing topic candidates to the user will be described.

FIG. 10 is a sequence diagram of the operation of the system 10 when providing topic candidates to the user.

When the user has logged in to the conversation support system 20 via the user terminal 80, as shown in FIG. A request can be made to the system 20 (S151).

Upon receiving the request in S151, the common-interest-word specifying unit 25b of the conversation support system 20 determines the current status of the user who made the request in S151 (hereinafter referred to as "target user" in the description of the operation shown in FIG. 10). A request is made to the schedule system 50 for information on a schedule that is later than the time of (S152) and that includes a participant other than the target user, that is, the target user's partner (S152). Here, the schedule with which the target user has a partner includes, for example, a schedule of a meeting in which the target user participates, a schedule of a visit by the target user, and the like. Further, the scheduled information requested in S152 includes information on the venue of this scheduled event and information on the other party of this scheduled target user. Further, the information on the venue is, for example, information on the venue of the conference in the case of the schedule of the conference, and information on the visited place in the case of the schedule of the visit.

When the schedule table system 50 receives the request in S152, it replies to the conversation support system 20 with the information requested in S152 (S153).

Upon receiving the answer in S153, the common interest word specifying unit 25b of the conversation support system 20 stores the information of the target user and the information of the other party in the schedule answered in S153, for each schedule answered in S153. A request is made to the database 40 (S154). Here, the information requested in S154 is a combination of the target user, the interesting word associated with the target user's partner, the representative word associated with this interesting word, and the registration date and time of this interesting word. information.

Upon receiving the request at S154, the interest database 40 replies to the conversation support system 20 with the information requested at S154 (S155).

Upon receiving the answer in S155, the common-interest-word identifying unit 25b of the conversation support system 20 identifies words that are highly likely to be of interest to both the target user and the target user's partner, that is, common-interest words. common interest word identification processing is executed (S156). Here, the common word-of-interest identification unit 25b executes the common word-of-interest identification process for each schedule answered in S153.

FIG. 11 is a flowchart of the common-interest-word specifying process in S156.

As shown in FIG. 11, the common-interest-word specifying unit 25b selects the target user and the current common-interest word for the words of interest associated with "Interest", which is the representative word indicating "interest", among the answers in S155. It is determined whether or not there is a match or similarity with the other party in the target schedule of word identification processing (hereinafter simply referred to as "target schedule" in the description of the operation shown in FIG. 11) ( S181). Here, the similar words include words indicated as synonyms in the synonym dictionary 24c, a synonym dictionary provided in a system external to the conversation support system 20, or the like. For example, when the words to be compared are place names, the common-interest-word identifying unit 25b may compare place names that have an inclusive relationship, such as "Hiroshima Prefecture" and "Kure City" in Hiroshima Prefecture, to You may judge as a similar word.

When the common interest word identifying unit 25b determines in S181 that there is a matching or similar word, it weights each of the words determined in S181 as matching or similar with a specific value (S182). The value used in the weighting in S182 may be set in advance for the word, or the conversation support system 20 may cause the user terminal 80 to display this word on the user terminal 80 so that it can be weighted each time via the user terminal 80. , may be set by the target user. The common-interest-word specifying unit 25b may change the value used for weighting in S182 according to the registration period calculated from the registration date and time associated with the word. For example, the common-interest-word specifying unit 25b changes the value used for weighting in S182 so that the larger the value used for weighting in S182, the shorter the registration period calculated from the registration dates and times associated with the word. You can let me. Here, the common word-of-interest specifying unit 25b uses the average of the registration periods of matching or similar words as the registration period calculated from the registration dates and times associated with the words. Similar words are weighted with the same value in S182.

If the common interesting word identifying unit 25b determines in S181 that there is no matching or similar word, or executes the process of S182, the common interesting word specifying unit 25b determines that the interesting word associated with the representative word other than "Interest" among the answers in S155 is , whether or not there is a match or similarity between the target user and the other party in the target schedule (S183). For example, the interesting word associated with the representative word "BirthPlace" for the target user is "Hokkaido", and the interesting word associated with the representative word "Visit" for the target user is "Hokkaido". , the common interesting word specifying unit 25b determines in S183 that there is a matching interesting word "Hokkaido". It should be noted that the judgment criteria for word similarity in S183 are the same as those in S181.

When the common interest word identifying unit 25b determines in S183 that there is a matching or similar word, it weights each of the words determined in S183 as matching or similar with a specific value (S184). The value used in the weighting in S184 may be set in advance for the word, or the conversation support system 20 causes the user terminal 80 to display the word so that the user terminal 80 can weigh the word each time. , may be set by the target user. The common-interest-word specifying unit 25b may change the value used for weighting in S184 according to the registration period calculated from the registration date and time associated with the word. For example, the common-interest-word identifying unit 25b changes the value used for weighting in S184 so that the larger the value used for weighting in S184, the shorter the registration period calculated from the registration dates and times associated with the word. You can let me. Here, the common word-of-interest specifying unit 25b uses the average of the registration periods of matching or similar words as the registration period calculated from the registration dates and times associated with the words. Similar words are weighted with the same value in S184.

When the common interest word identification unit 25b determines in S183 that there is no matching or similar word, or executes the processing of S184, the information on the venue answered in S153 for the target schedule includes the area name and company name. It is determined whether or not a word indicating a specific venue such as is included (S185). It should be noted that the word indicating the target scheduled venue is a word that is highly likely to be of interest to both the target user and the target user's partner.

When the common interest word specifying unit 25b determines in S185 that the word indicating the specific venue is not included, it determines whether the target schedule is a visit schedule (S186).

When the common interest word identifying unit 25b determines in S186 that the target schedule is a visit schedule, the common interest word identifying unit 25b selects a word indicating a specific location such as the location or name of the company of the target user in the target schedule. It is obtained from the customer information database 60 (S187).

When the common interest word identification unit 25b determines in S185 that a word indicating a specific venue is included, or when the process of S187 ends, the common interest word identification unit 25b determines that the word is included in the venue information answered in S153 for the target schedule. A specific value is used to weight the words indicating the specific venue that are stored or the words indicating the specific venue acquired in S187 (S188). The value used in the weighting in S188 may be set in advance for the word, or the conversation support system 20 may cause the user terminal 80 to display this word on the user terminal 80 so that it can be weighted each time via the user terminal 80. , may be set by the target user.

When the common interest word identification unit 25b determines in S186 that the target schedule is not a visit schedule, or when the process of S188 ends, the common interest word identification unit 25b determines that the words weighted in S182, S184, and S188 have a specific value or more. The words weighted by the values are specified as common interest words (S189), and the processing shown in FIG. 11 is terminated. Note that the common word-of-interest specifying unit 25b may specify all the words weighted in S182, S184, and S188 as common words of interest.

As shown in FIG. 10, when the common word-of-interest identifying process in S156 ends, the information providing unit 25c selects the common-interest words identified in the common-interest-word identifying process in S156 singly or in a plurality of combinations. By using it as a search word, one or more information providing sites such as the information providing site 70 are requested for search results related to the common interest word (S157).

The information providing site receives the request in S157 and provides one or more search results to the conversation support system 20 (S158). Here, the search results include sentences (hereinafter referred to as "summary sentences") indicating the outline of the search results.

The information providing unit 25c of the conversation support system 20 executes topic candidate selection processing for selecting topic candidates to be provided to the user from the search results provided in S158 (S159).

FIG. 12 is a flowchart of topic candidate selection processing in S159.

As shown in FIG. 12, the information providing unit 25c targets only one of the search results provided in S158 that has not yet been targeted in the processing shown in FIG. 12 (S201).

Next, the information providing unit 25c performs a positive/negative analysis on the summary sentences included in the search result of the current target (S202). Here, there are various existing techniques for positive/negative analysis of sentences. For example, prepare a dictionary (hereinafter referred to as a "positive/negative dictionary") that manually or mechanically digitizes the degree of positive/negative for each word, and for each word in the result of morphological analysis of the sentence, in the positive/negative dictionary There is a method of obtaining the positive/negative degree of a document by obtaining the associated positive/negative degrees and totaling all the obtained positive/negative degrees.

After the process of S202, the information providing unit 25c acquires a weighting value (hereinafter referred to as "weighting value for multiplication") for multiplying the positive/negative degree, which is the result of the positive/negative analysis performed in S202. (S203). That is, if there is one common word of interest as a search word in S157 for the current target search result, the information providing unit 25c sets The weighted value is acquired in S203 as a weighted value for multiplication. In addition, if there are a plurality of common words of interest that are used as search words in S157 for the current target search result, the information providing unit 25c performs An average value of the weighted values is obtained as a weighted value for multiplication in S203.

After the process of S203, the information providing unit 25c multiplies the positive/negative degree, which is the result of the positive/negative analysis performed in S202, by the weighting value for multiplication acquired in S203, and provides the user with A value for selecting a topic candidate (hereinafter referred to as "selection value") is calculated (S204).

After the processing of S204, the information providing unit 25c determines whether or not there are any search results that have not yet been targeted in the processing shown in FIG. 12 among the search results provided in S158 (S205).

When the information providing unit 25c determines that there is a search result that has not yet been targeted in the process shown in FIG. 12 among the search results provided in S158, it executes the process of S201.

If the information providing unit 25c determines that there is no search result that has not yet been targeted in the processing shown in FIG. are selected as topic candidates to be provided to the user (S206). Note that the number of search results selected in S206 may be set by the user. For example, the information providing unit 25c causes the user terminal 80 to display the search results provided in S158 in descending order of the selection value calculated in S204, thereby allowing the user to use the number of search results selected in S206. may be set via the user terminal 80.

After the process of S206, the information providing unit 25c ends the process shown in FIG.

As shown in FIG. 10, after the topic candidate selection process in S159, the information providing unit 25c notifies the topic candidate display application that has transmitted the request in S151 of the topic candidates selected in the topic candidate selection process in S159. A notification data generating process for generating data (hereinafter referred to as "notification data") for the notification is executed (S160).

FIG. 13 is a flowchart of notification data generation processing in S160.

As shown in FIG. 13, the information providing unit 25c automatically displays the headline and beginning of the linked article when the topic candidate display application that has sent the request in S151 describes the link (hereinafter referred to as (referred to as "automatic display application")) (S221). For example, LINE (registered trademark) and Slack (registered trademark) are known as automatic display applications. For example, e-mail is known as a topic candidate display application that is not an automatic display application.

If the information provision unit 25c determines in S221 that the topic candidate display application that has transmitted the request in S151 is the automatic display application, the link of the topic candidate selected in the topic candidate selection process in S159 is displayed as the topic candidate in S159. Notification data including each topic candidate selected in the selection process is generated (S222). Here, in the notification data, the information providing unit 25c adds supplementary information such as the degree of recommendation of the article linked to the topic candidate selected in the topic candidate selection process of S159. It may be included for each candidate topic. Supplementary information such as the degree of recommendation of linked articles can be obtained from information providing sites.

When the information providing unit 25c determines in S221 that the topic candidate display application that has sent the request in S151 is not an automatic display application, the link of the topic candidate selected in the topic candidate selection processing in S159 and the link of this link Notification data including the headline and beginning part of the previous article for each topic candidate selected in the topic candidate selection process of S159 is generated (S223). Here, in the notification data, the information providing unit 25c adds supplementary information such as the degree of recommendation of the article linked to the topic candidate selected in the topic candidate selection process of S159. It may be included for each candidate topic. The headline and opening part of the linked article and supplementary information such as the degree of recommendation of the linked article can be obtained from the information providing site. Even if the topic candidate display application is not an automatic display application, the information provision unit 25c includes the headline and beginning part of the article in the notification data, thereby improving the user's convenience.

After the processing of S222 or S223, the information providing unit 25c ends the processing shown in FIG.

As shown in FIG. 10, after the notification data generating process of S160, the information providing unit 25c sends the notification data generated in the notification data generating process of S160 to the topic candidate display application that has transmitted the request in S151. Send (S161).

Therefore, the control unit 85 of the user terminal 80 causes the display unit 82 to perform display based on the notification data transmitted in S161 using the topic candidate display application (S162).

FIG. 14 is a diagram showing an example of a screen 90 displayed on the display unit 82 by the topic candidate display application based on the notification data.

A screen 90 shown in FIG. 14 includes a link 91, an image 92 such as a photograph used in the linked article indicated by the link 91, a headline 93 and a beginning part 94 of the linked article indicated by the link 91. are included for each topic candidate. When the link 91 is pressed via the operation unit 81 , the control unit 85 can acquire the linked article indicated by the link 91 from the Internet and display it on the display unit 82 . The target user can refer to the image 92 , the headline 93 and the beginning part 94 when selecting a topic to actually use from the topic candidates displayed on the screen 90 .

As described above, when the morpheme of the utterance content included in the utterance record or the original form of this morpheme is included in the word list for extraction 24b (YES in S126), the conversation support system 20 converts this morpheme into speech. Even if the word does not indicate the user's interest, the word of interest is specified from the word that is the source of the dependency for this morpheme (S127 and S128), so there is a possibility that both the user and the other party of this user are interested. It is possible to appropriately provide topic candidates with a high

In sales activities through field sales, small talk and ice breakers in meetings and business negotiations are important factors in building trusting relationships with customers. When the person in charge of sales activities is the user, the conversation support system 20 appropriately selects topic candidates that are likely to be of interest to both the person in charge of sales activities and the customer who is the partner of the person in charge. Therefore, it is not necessary to rely on the communication skills and experience of the person in charge of sales activities to decide what topics to talk about in chats and ice breakers in sales activities. Although meetings and negotiations in sales activities have been explained, the same applies to meetings other than these.

The conversation support system 20 provides topic candidates that are highly likely to be of interest to both the user and the user's partner, even if the user's partner is not registered in the conversation support system 20 as a user. can be provided appropriately.

The conversation support system 20 can appropriately provide topic candidates that are highly likely to be of interest to both the user and the other party of the user. Even if the user does not fully understand the other party, communication between the other party and the user can be promoted.

Conversation support system 20 selects topic candidates to be provided to the user based on weighting values performed for each word of common interest related to the candidates (S203, S204 and S206). , it is possible to appropriately provide topic candidates that are highly likely to be of interest to both the user and the other party.

The conversation support system 20 selects topic candidates to be provided to the user based on the positive/negative degree, which is the result of positive/negative analysis of the synopsis obtained from the information providing site (S202, S204 and S206). Therefore, it is possible to appropriately provide topic candidates with positive content as topic candidates that are highly likely to be of interest to both the user and the user's partner.

In the operation shown in FIG. 10, the conversation support system 20 provides topic candidates to the user in response to a request from the user. However, the conversation support system 20 may automatically provide the user with topic candidates as shown in FIGS. 15 and 16, for example.

When the conversation support system 20 automatically provides the user with topic candidates, the conversation support system 20 always manages the start times of all the user's plans from the schedule system 50 in which the user's partner is present. ing. Any method can be adopted as a method for the conversation support system 20 to always manage the scheduled start time. For example, among the schedules of users of the conversation support system 20, the schedule system 50 may periodically notify the conversation support system 20 of the start times for all schedules in which the user's partner exists. The support system 20 may periodically request the calendar system 50 to provide the start time.

FIG. 15 is a flowchart of the operation of the conversation support system 20 when determining whether or not to start the process of providing topic candidates to the user.

The common interest word identifying unit 25b of the conversation support system 20 performs the operation shown in FIG. 15 when automatically providing topic candidates to the user.

As shown in FIG. 15, the common-interest-word identifying unit 25b selects words that the user of the conversation support system 20 expects to have a partner and has not yet set as an object of provision of topic candidates until the start time. Until it is determined that there is a subject for which the time of is within a specific period of time, among the subject candidates for which the user of the conversation support system 20 is scheduled to have a partner and has not yet been targeted for provision of topic candidates, A determination is made as to whether or not there is a program whose time until the start time is within a specific period of time (S241).

The common interest word specifying unit 25b selects, among the words that the user of the conversation support system 20 plans to have a partner and has not yet set as a topic candidate provision target, the time until the start time is within a specific time period. If it is determined in S241 that there is a candidate for the conversation support system 20, the user of the conversation support system 20 is expected to have a partner and has not yet been targeted for provision of topic candidates. Topic candidates are provided within a specific time period (S242).

Next, the common interest word specifying unit 25b starts the process of providing topic candidates to the user regarding the schedule targeted in S242 (S243), and executes the process of S241.

FIG. 16 is a sequence diagram of the operation of the system 10 when the conversation support system 20 starts the process of providing topic candidates to the user by the process of S243 of the operation shown in FIG.

As shown in FIG. 16, the common-interest-word identifying unit 25b collects information about a target schedule in the current action shown in FIG. , to the schedule system 50 (S261). Here, the scheduled information requested in S261 includes information on the venue of this scheduled event and information on the user's partner in this scheduled event.

The subsequent processing of S262-S271 is the same as the processing of S153-S162 (see FIG. 10). However, the topic candidate display application, which is the target of the notification data generated by the conversation support system 20 in S269 and the destination of the notification data by the conversation support system 20 in S270, is set in advance by the conversation support system 20 for each user. It is a specific topic candidate display application such as a topic candidate display application registered in 20.

In the above, when the conversation support system 20 automatically provides the user with topic candidates, the schedule table system 50 selects all schedules in which the user's partner is present from among the schedules of the user. are always in control. However, when the conversation support system 20 automatically provides the user with topic candidates, the schedule table system 50 selects the next time when the user has a partner among the user's schedule after the current time. A configuration in which the start time is always managed only for the schedule may be used.

As described above, the conversation support system 20 uses the schedule system 50 to provide the user with topic candidates. However, the conversation support system 20 may provide topic candidates to the user without using the schedule system 50, as shown in FIGS. 17 and 18, for example.

FIG. 17 is a sequence diagram of the operation of the system 10 when providing topic candidates to the user without using the schedule system 50 .

When the user has already logged in to the conversation support system 20 via the user terminal 80, as shown in FIG. A request can be made to the conversation support system 20 via the candidate display application (S291).

Upon receiving the request in S291, the common-interest-word identifying unit 25b of the conversation support system 20 receives the request in S291 (hereinafter referred to as "target user" in the description of the operation shown in FIG. 17) and , and the information of the partner specified in the request in S291 from the interest database 40 (S292). Here, the information requested in S292 is a combination of the target user, the interesting word associated with the target user's partner, the representative word associated with this interesting word, and the registration date and time of this interesting word. information.

The subsequent processing of S293-S300 is the same as the processing of S155-S162 (see FIG. 10). However, the common interest word specifying process in S294 is as shown in FIG.

FIG. 18 is a flowchart of the common-interest-word specifying process in S294.

As shown in FIG. 18, the common word-of-interest identifying unit 25b executes the processes of S321 to S324 similar to the processes of S181 to S184 (see FIG. 11).

If the common interest word identifying unit 25b determines in S323 that there is no matching or similar word, or executes the process of S324, it weights the words weighted in S322 and S324 with a value equal to or greater than a specific value. is specified as a common-interest word (S325), and the process shown in FIG. 18 is terminated. Note that the common word-of-interest specifying unit 25b may specify all the words weighted in S322 and S324 as common words of interest.

Conversation support system 20 provides users with information acquired from an information providing site as information related to common interest words in the present embodiment. However, the conversation support system 20 may provide the user with information other than the information acquired from the information providing site as the information related to the common interest words. For example, the conversation support system 20 may provide the common interest word itself to the user as information related to the common interest word.

In the present embodiment, conversation support system 20 indicates "LivingPlace" indicating "place of residence" and "place of birth" as extraction words represented by representative words other than "Interest" indicating the speaker's interest. The extraction word list 24b includes extraction words represented by representative words indicating places, such as "BirthPlace" and "Visit" indicating a "visited place." However, the conversation support system 20 selects the extraction word represented by the representative word other than the representative word indicating the location as the extraction word represented by the representative word other than "Interest" indicating the interest of the speaker. may be included in list 24b. For example, the conversation support system 20 includes extraction words represented by representative words representing food in the extraction word list 24b as extraction words represented by representative words other than "Interest" representing the speaker's interest. But it's okay.

The statement database 30, the interest database 40, the schedule system 50, and the customer information database 60 are provided separately from the conversation support system 20 in this embodiment. However, conversation support system 20 may comprise at least one of statement database 30 , interest database 40 , schedule system 50 and customer information database 60 .

20 Conversation support system (computer)
24a Conversation support program 24b Word list for extraction 25a Interest word extraction unit 25b Common interest word identification unit 25c Information provision unit 40 Interest database (storage unit)
70 Information site

Claims (9)

an interesting word extraction unit that extracts interesting words, which are words that are likely to be of interest to the speaker, from the transcript;
a common interest word identifying unit that identifies common interest words that are words that are highly likely to be of interest to a plurality of persons;
an information providing unit that provides users with information related to the common words of interest identified by the common words of interest identification unit;
The interesting word extraction unit registers the interesting words extracted from the speech record in a storage unit for each speaker,
When the words of interest registered for each of the plurality of persons in the storage unit match or are similar, the common word-of-interest identification unit identifies the words of interest as the words of common interest,
The interesting word extraction unit adds a morpheme of utterance content included in the utterance record, or a morpheme of this morpheme, to an extraction word list, which is a list of words for extraction, which are words for extracting the interesting word from the utterance record. is included, the interesting word is identified from the word that is the source of dependency for this morpheme. the extraction word list includes the extraction word for each representative word representing a plurality of the extraction words;
The interesting word extraction unit associates the interesting word extracted from the statement record with the representative word set in the extraction word list with the extraction word to which the interesting word depends. , is registered in the storage unit,
2. The conversation support system according to claim 1, wherein the interesting word extracting unit switches between additionally registering the interesting word and overwriting registration of the interesting word depending on the type of the representative word. The common word-of-interest identifying unit, when the words of interest registered for each of the plurality of persons in the storage unit match or are similar to each other, weights the words of interest,
The common interesting word specifying unit changes the value used for weighting the interesting word so that the registration period of the interesting word becomes shorter as the value used for weighting the interesting word increases,
2. The conversation support system according to claim 1, wherein said common word-of-interest identification unit identifies said words of interest weighted by a value equal to or greater than a specific value as said words of common interest. The information providing unit uses the common words of interest identified by the common words of interest identifying unit as search words to provide search results requested to an external information providing site as the information related to the common words of interest. 2. The conversation support system according to claim 1, wherein the conversation support system is provided to the user as. The common-interest-word identifying unit weights each of the common-interest words,
3. The information providing unit selects the information to be provided to the user based on the weighting value executed for each of the common words of interest by the common words of interest identifying unit. 4. The conversation support system according to 4. The information providing unit acquires a summary sentence of the information related to the common interest word identified by the common interest word identification unit from an external information providing site, and performs a positive/negative analysis on the summary sentence. and the information to be provided to the user is selected based on the degree of positive/negative that is the result of the positive/negative analysis. system. The common-interest-word identifying unit weights each of the common-interest words,
The information providing unit acquires a summary sentence of the information related to the common interest word identified by the common interest word identification unit from an external information providing site, and performs a positive/negative analysis on the summary sentence. death,
When the common interest word used as the search word for the search result is one, the information providing unit provides a weighted value for the common interest word and the positive/negative analysis Select the information to be provided to the user based on the value obtained by multiplying the positive/negative degree that is the result of
When there are a plurality of the common words of interest used as the search words for the search results, the information providing unit provides an average value of weighted values of these common words of interest, and the positive 5. The conversation support system according to claim 4, wherein the information to be provided to the user is selected based on a value obtained by multiplying the positive/negative degree, which is the result of the negative analysis. The common word-of-interest identifying unit, when the words of interest registered for each of the plurality of persons in the storage unit match or are similar to each other, weights the words of interest, and determines the words of interest. is identified as the common interest word,
The common interesting word identifying unit changes the value used for weighting the interesting word so that the registration period of the interesting word becomes shorter as the value used for weighting the interesting word increases. A conversation support system according to claim 7. an interesting word extraction unit that extracts interesting words, which are words that are likely to be of interest to the speaker, from the transcript;
a common interest word identifying unit that identifies common interest words that are words that are highly likely to be of interest to a plurality of persons;
an information providing unit that provides a user with information related to the common words of interest identified by the common words of interest identification unit;
is realized on a computer,
The interesting word extraction unit registers the interesting words extracted from the speech record in a storage unit for each speaker,
When the words of interest registered for each of the plurality of persons in the storage unit match or are similar, the common word-of-interest identification unit identifies the words of interest as the words of common interest,
The interesting word extraction unit adds a morpheme of utterance content included in the utterance record, or a morpheme of this morpheme, to an extraction word list, which is a list of words for extraction, which are words for extracting the interesting word from the utterance record. is included, the interesting word is specified from the word that is the source of dependency for this morpheme.

Images