JP7152437B2 - Summary generation program, summary generation device and summary generation method - Google Patents

Description

The present invention relates to a summary generation program, a summary generation device, and a summary generation method.

2. Description of the Related Art In recent years, with the rapid spread of the Internet, various contents are being actively provided via networks. Techniques for providing content more suitable for users are also known.

For example, Patent Literature 1 proposes a technique for personalizing the display style of arbitrary content provided from a server according to user preferences.

JP 2013-77240 A

However, with the conventional technology described above, it is not always possible to generate a summary that reflects the user's preferences. For example, in the above conventional technology, the content requested by the terminal is acquired from the server, the portion corresponding to the preference data is personalized for the acquired content, and the personalized content and the acquired content are switched to switch between the personalized content and the acquired content. Send to

As described above, the above-described conventional technology merely personalizes content according to the user's taste, and does not generate a summary sentence from character information. Therefore, with the conventional technology described above, it is not always possible to generate a summary that reflects the user's preferences.

The present application has been made in view of the above, and an object thereof is to generate a summary that reflects the user's taste.

A summary generating program according to the present application includes an acquisition procedure for acquiring user information indicating a user's interest and a content to be processed, A computer is caused to execute an extraction procedure for extracting a target sentence, which is a topical sentence, and a generation procedure for generating a summary sentence in which the contents of the content to be processed are summarized based on the target sentence.

According to one aspect of the embodiment, it is possible to generate a summary that reflects the user's taste.

FIG. 1 is a diagram showing an overview of summary generation processing according to the embodiment. FIG. 2 is a diagram illustrating an example of a summary generation process performed by the summary generation device according to the embodiment; FIG. 3 is a diagram showing an example of summary generation processing using a tree structure. FIG. 4 is a diagram illustrating a configuration example of a summary generating device according to the embodiment; FIG. 5 is a diagram showing an example of a positive/negative map according to the embodiment. FIG. 6 is a flowchart showing a summary generation processing procedure according to the embodiment. FIG. 7 is a hardware configuration diagram showing an example of a computer that implements the functions of the abstract generation device.

Embodiments (hereinafter referred to as "embodiments") for implementing the abstract generating program, abstract generating apparatus, and abstract generating method according to the present application will be described in detail below with reference to the drawings. Note that the abstract generating program, abstract generating apparatus, and abstract generating method according to the present application are not limited to this embodiment. Also, in each of the following embodiments, the same parts are denoted by the same reference numerals, and overlapping descriptions are omitted.

[1. Overview of summary generation processing according to the embodiment]
At present, AI (executive intelligence) is being installed in various devices such as robots, and it is expected to realize dialogs, recommendations, and sentence summaries that are adapted to individual users.

Also, for example, if it is possible to summarize contents (for example, article contents and emails) for a predetermined period according to individual tastes, the user can easily find out only the contents that interest him/herself. Convenient.

In addition, for example, if it is possible to summarize distribution candidate content (for example, article content) according to the tastes of individual users to whom it is distributed, for example, it is possible to further personalize the distribution surface on which the content is displayed. , it will be possible to realize a delivery aspect that can effectively increase the interest of users.

Therefore, based on the above premise, the summary generation process according to the embodiment identifies, from among the sentences, sentences in which the subject of interest of the user is a topic, and using the specified sentences, A summary sentence reflecting this user's preference is generated.

Specifically, the summary generating process according to the embodiment acquires user information indicating the user's interest and content to be processed, and extracts sentences included in the acquired content to be processed from the interest indicated by the user information. extracts a target sentence, which is a topical sentence. Then, based on the extracted target sentence, a summary sentence that summarizes the contents of the content to be processed is generated.

Also, as will be described in detail with reference to FIGS. 1 and 2, in the summary generation process according to the embodiment, target sentences are extracted as follows. In the summary generation process according to the embodiment, as a target of interest of the user, a target indicated by a proper noun included in the content to be processed and a target in which the user is interested is extracted as a target sentence. do. More specifically, in the summary generation process according to the embodiment, among a plurality of proper nouns with different expressions, synonymous proper nouns are unified into one formal proper noun (for example, name identification), and this formal proper noun indicates A target (for example, a team, a person, etc.) and a target that the user is interested in is extracted as a target sentence.

In addition, in the summary generation process according to the embodiment, the subject is estimated for sentences that do not contain a subject among the sentences included in the content to be processed. Then, among the sentences whose subjects have been estimated, sentences whose subject is the object indicated by the synonymous proper noun (one formal proper noun) are extracted as target sentences.

[2. Summary generating program and summary generating device]
The abstract generating program according to the embodiment is a program for causing a computer to execute the abstract generating process according to the embodiment described above. Specifically, the summary generation program according to the embodiment includes an acquisition procedure for acquiring user information indicating the user's interest and content to be processed; A computer is caused to execute an extraction procedure for extracting a target sentence, which is a sentence in which an object of interest is a topic, and a generation procedure for generating a summary sentence in which the contents of the content to be processed are summarized based on the target sentence. It is a program for

Also, in the present embodiment, the summary generation device 100 is taken as an example of such a computer. That is, the summary generating apparatus 100 acquires user information indicating the user's interest and content to be processed, under the control of the summary generating program according to the embodiment, and extracts sentences included in the acquired content to be processed by the user. An object of interest indicated by information is extracted as a sentence that is a topic. Then, based on the extracted target sentence, the summary generating apparatus 100 generates a summary sentence that summarizes the contents of the content to be processed.

Also, in the present embodiment, the abstract generation device 100 is assumed to be realized by a server device, a cloud system, or the like. On the other hand, the computer on which the abstract generation program is executed may be, for example, an information processing terminal used by the end user. (Personal Computer), desktop PC, mobile phone, PDA (Personal Digital Assistant), and the like.

[3. Overview of summary generation processing]
From now on, an overview of the summary generation process performed by the summary generation device 100 according to the embodiment will be described with reference to FIG. FIG. 1 is a diagram showing an overview of summary generation processing according to the embodiment. Also, in the present embodiment, the content to be processed is assumed to be "article content based on sports articles". Then, in this embodiment, the user is assumed to be a fan of a team (or a fan of a specific athlete) in a specific sport (for example, baseball), and a summary reflecting the interests and tastes of the user is generated. Take the scene as an example.

On the other hand, the summary generation process according to the embodiment is not limited to sports articles, and can be applied to any articles such as news articles and blogs. Further, the summary generating process according to the embodiment is not limited to articles, and can be applied to e-mail contents (summaries of e-mail texts) and books (summaries of book texts), for example.

First, the summary generation device 100 acquires user information indicating the user's interest. For example, the summary generation device 100 can determine what kind of things (people, things, events, jobs, etc.) it is interested in, what things are important to it, what things it likes, User information indicating the user's interest is acquired by receiving registration of user information indicating information about the interest. In the example of FIG. 1, the summary generating apparatus 100 acquires user information indicating the interest of the user U1 from the user U1 by accepting registration of user information from the user U1.

Specifically, in the example of FIG. 1, the summary generation device 100 acquires user information such as fans of "team Ta" (interested in "team Ta" or likes "team Ta") from user U1. is doing. In addition, in the example of FIG. 1, the summary generation device 100 acquires user information such as fans of "PY1 player" (interested in "PY1 player" or likes "PY1 player") from user U1. .

Note that the summary generating apparatus 100 does not acquire user information through user registration, but rather uses behavior information indicating user behavior (for example, user location information, user purchase history, user browsing history), etc. to obtain user information. An interest may be estimated and the estimated information may be acquired as user information indicating the user's interest. Further, in order to make such an estimation, the summary generating apparatus 100 may have a storage unit (not shown) in which user behavior information (behavior history) is stored. may be obtained.

Next, the summary generation device 100 acquires the content to be processed. For example, the summary generating apparatus 100 acquires an article content group for a predetermined period (for example, past one month) as content to be processed. For example, if “team Ta” is a baseball team and “PY1 player” is a baseball player, the summary generation device 100 acquires a content group of sports article content in which articles about baseball are published. Further, the summary generating apparatus 100 may acquire newspaper articles for a predetermined period of time, or may acquire web content distributed over the Internet for a predetermined period of time. In the example of FIG. 1, the summary generation device 100 acquires such content to be processed from the external server SV.

Note that it may be arbitrary where the summary generation device 100 acquires the content to be processed. For example, when the user desires a summary of mail texts (an example of content to be processed) for a predetermined period, the summary generating apparatus 100 may accept registration of text information indicating the mail texts for the predetermined period from the user. good. Further, when the user desires a summary of the mail text for a predetermined period (an example of the content to be processed), the summary generating apparatus 100 acquires text information indicating the mail text for the predetermined period from a predetermined mail server. may

In this way, the summary generation device 100 starts the summary generation process according to the embodiment after acquiring the user information indicating the user's interest and the content to be processed. First, based on the user information of user U1, summary generating apparatus 100 acquires article content related to user U1's interest from among the acquired article content group, thereby narrowing down the content to be processed (step S1). ). For example, the summary generation device 100 acquires article content related to user U1's interest from the article content group. According to the example of FIG. 1, user U1's interests are "team Ta" and "player PY1". Also, "team Ta" and "PY1 players" are both proper nouns.

Therefore, the summary generating device 100 acquires the article content regarding "team Ta" and "PY1 player" from the article content group. For example, the summary generation device 100 acquires article content including "team Ta" and "PY1 player" in the headline (heading) or text from the article content group. As a simple example in FIG. 1, it is assumed that the summary generating apparatus 100 acquires article content C1, which is one piece of article content. Therefore, at this point, the content to be processed is narrowed down to the article content C1. Note that when it is determined that the target of interest is not a proper noun but a general term, for example, the summary generating apparatus 100 may estimate a proper noun closest to the determined general term, or register the proper noun. User may be warned.

Also, as will be described later, the article content C1 includes a plurality of sentences forming the text, but also includes information unnecessary for summarizing (for example, photographs). Therefore, next, the summary generating apparatus 100 removes unnecessary information from the article content C1 by performing cleansing processing (data cleansing) on the article content C1 (step S2). The summary generator 100 can employ any conventional technique for such cleansing processing.

Next, the summary generation device 100 merges proper nouns for sentences included in the article content C1 (step S3). Here, taking the person “PY1” as an example, the person “PY1” may be described in various expressions in the article content C1. For example, the subject of one sentence in the article content C1 may be written as "PY1 player of team Ta", while the subject of another sentence in the article content C1 may be written as "PY1 player". In addition, when "PY1 player" is called by a specific nickname by fans, this nickname may be used in the subject of a certain sentence in the article content C1.

In this way, the person "PY1" may be described by a plurality of different expressions, but all of them can be grouped into one person. For example, "PY1 player of team Ta" and "PY1 player" can be merged into one person "PY1". In other words, "PY1 player of team Ta" and "PY1 player" are synonymous, although the expressions are different in that they both refer to one person "PY1".

For this reason, the summary generating apparatus 100 determines whether or not the proper noun indicating the target of user U1's interest is a proper noun that may be written with a plurality of different proper nouns. Then, when determining that there is a case where the summary generating device 100 is described by a plurality of different proper nouns, the summary generating device 100 identifies sentences having each of the plurality of proper nouns as a subject among the sentences included in the article content C1. Then, the summary generation device 100 merges the most formal proper nouns (for example, real names) with respect to a plurality of different proper nouns for the specified subject of the sentence. In the example of FIG. 1, the summary generation device 100 merges "PY1" into the subject of a sentence having "PY1 player of team Ta" as the subject. In addition, in the example of FIG. 1, the summary generation device 100 merges "PY1" into the subject of a sentence having "PY1 player" as the subject.

In addition, what other expressions may be used for proper nouns that indicate the object of interest of the user? Information about what kind of proper nouns can be unified from synonymous proper nouns is shown in a dedicated proper noun dictionary. Therefore, the summary generating apparatus 100 can formally merge proper nouns by comparing this proper noun dictionary with proper nouns that indicate objects of interest to the user. In addition, by performing such merging of proper nouns, the summary generating apparatus 100 makes it possible for the user's interest in the sentence to be a topic. You can prevent being excluded. As a result, the summary generating apparatus 100 can generate a summary that effectively reflects the user's preferences.

Next, the summary generation device 100 estimates the subject of sentences that do not contain the subject among the sentences included in the article content C1 (step S4). In Japanese sentences, sentences that omit the subject are sometimes used when the subject can be clearly identified from the context. In this case, the omitted subject is called the ``implicit subject''. In step S4, the summary generation device 100 estimates this implicit subject for sentences that do not contain a subject among the sentences included in the article content C1.

For example, the summary generating apparatus 100 determines whether or not there is a sentence that does not contain a subject among the sentences included in the article content C1. Infer the implicit subject of For example, when the summary generating apparatus 100 can determine a sentence that does not contain a subject, it estimates the implicit subject based on the subject of the sentence preceding this sentence (the subject after proper noun merging). Such tacit estimation of the subject makes it possible to prevent the text from being excluded from the summary target even though the target of the user's interest is the topical text.

Next, the summary generation device 100 extracts, from among the sentences included in the article content C1, a target sentence, which is a sentence in which the target of user U1's interest is a topic (step S5). For example, the summary generation device 100 extracts sentences whose subject is the object of interest of the user U1 as target sentences from sentences included in the article content C1. For example, the summary generation device 100 extracts, as target sentences, sentences in which proper nouns have been merged in step S3 from sentences included in the article content C1. Further, for example, the summary generation device 100 extracts, as target sentences, sentences in which the merged proper noun is estimated as the implicit subject in step S3 from the sentences in which the implicit subject was estimated in step S4.

Next, the summary generation device 100 obtains the tree structure of the target sentence by parsing the target sentence according to an arbitrary algorithm (step S6). Such a tree structure is the result of syntactic analysis. Moreover, such a tree structure is a tree structure whose nodes are clauses that form the target sentence. An example of the tree structure will be described with reference to FIG.

The summary generation device 100 also generates a summary sentence summarizing the content of the article content C1 based on the clauses (nodes) included in the tree structure (step S7). For example, the summary generation device 100 generates a summary sentence summarizing the content of the article content C1 by connecting clauses along the tree structure so that the number of branches from the clause is equal to or less than a predetermined number.

Here, whether a topic or article gives a positive impression (that is, changes to positive emotions) or a negative impression (that is, changes to negative emotions) to an individual depends on each individual. Much depends on individual interests and tastes (preferences). For example, an article about a team winning is an article that makes the fans of that team feel positive, but it makes the fans of the opposing team feel negative.

A more detailed example is given. For example, an article about "Team Ta" winning causes positive feelings toward fans of "Team Ta", and an article about "Team Ta" losing causes positive feelings toward anti-"Team Ta". . On the other hand, an article about "team Ta" being defeated makes fans of "team Ta" feel negative feelings, and an article about "team Ta" winning makes people feel negative feelings about anti-"team Ta". .

For this reason, if it is possible to effectively determine whether the information provided to the user gives the user a positive or negative feeling, the following will be possible in the future. is considered possible. For example, the information provided to the user can be classified into information that causes positive emotions and information that causes negative emotions. It becomes possible to easily distinguish only high-level information. In addition, in an individual-adaptive interactive personal agent system, it becomes possible to realize a dialogue that is more individualized.

Based on this premise, the summary generating apparatus 100 generates the summary sentence generated in step S7 for each type of emotion (positive emotion, negative emotion) based on user U1's preference for the object of interest of user U1. An emotion value indicating the degree of emotion for the content of is estimated (step S8). For example, based on user U1's preferences, summary generating apparatus 100 estimates an emotion value indicating positive emotion and an emotion value indicating negative emotion for the content of the abstract generated in step S7. Hereinafter, an emotion value indicating a positive emotion and an emotion value indicating a negative emotion may be abbreviated as "PN value".

In addition, the summary generating apparatus 100 uses a model that has learned the relationship between the sentence and the PN value, such as what kind of expression the sentence makes the user feel positive or negative. The PN value for the content of the abstract is estimated, the details of which will be described later.

Then, the summary generation device 100 outputs a summary sentence with a PN value to the user U1 (step S9). For example, the summary generating apparatus 100 transmits the PN-added summary to the terminal device of the user U1 so that the PN-added summary is displayed on the terminal device.

Now, as has been described with reference to FIG. 1, the summary generation device 100 acquires user information indicating the user's interest and content to be processed according to the summary generation program according to the embodiment. Among the sentences included in the content to be processed, a target sentence that is a topic of interest indicated by the user information is extracted. Then, based on the extracted target sentence, the summary generating apparatus 100 generates a summary sentence that summarizes the contents of the content to be processed.

As a result, the summary generation program according to the embodiment can generate a summary that reflects the user's preferences. As a result, the summary generation program according to the embodiment can generate summary sentences that can effectively increase the user's interest.

[4. Example of summary generation processing]
From here, an example of a summary generation process performed by the summary generation device 100 according to the embodiment will be described with reference to FIG. 2 . Specifically, FIG. 2 shows an example of a summary generation process according to the flow shown in FIG. 1 using a specific example of sentences included in article content C1. Note that explanations overlapping those in FIG. 1 are simplified or omitted. Also, the step numbers (such as S1) shown in FIG. 2 correspond to the step numbers shown in FIG.

First, FIG. 2(a) shows the original text, which is the text included in the article content C1 and has not been processed. As shown in FIG. 2(a), article content C1 includes text CSN11, text CSN12, text CSN13, text CSN14 and text CSN15 in the original state.

In such a state, the summary generation device 100 performs cleansing processing on the original text (step S2). The text CSN15 is excluded by the cleansing process, and as a result, the article content C1 after cleansing becomes as shown in FIG. 2(b).

Next, the summary generator 100 merges proper nouns in the state of FIG. 2(b) (step S3). Here, the target of user U1's interest is, for example, the person "PY1". ” is used. In addition, "PY1 player of team Ta" and "PY1 player" are synonymous, although the expressions are different in that both refer to one person "PY1". For this reason, "PY1 player of team Ta" and "PY1 player" can be unified into the person "PY1" who is the target of user U1's interest.

Therefore, the summary generation device 100 merges the proper noun "PY1" with respect to each of "PY1 player of team Ta" and "PY1 player". As a result, the merged article content C1 becomes as shown in FIG. 2(c).

Next, in the state of FIG. 2(c), the summary generation device 100 determines whether or not there is a sentence that does not contain a subject among the sentences included in the article content C1, and can be determined, the implicit subject of this sentence is inferred. In the state of FIG. 2(c), the summary generation device 100 determines the sentence CSN12 as a sentence containing no subject, and estimates the implicit subject of the sentence CSN12. In such an example, the summary generation device 100 estimates the implicit subject "PY1" based on the subject of the sentence before the sentence CSN12 (the subject after proper noun merging). Also, the summary generator 100 inserts the estimated implicit subject into the sentence CSN12. As a result, the article content C1 after implicit subject estimation is as shown in FIG. 2(d). Thus, the subject (implicit subject) of sentence CSN12 is a proper noun indicating the object of interest of user U1, and corresponds to the proper noun merged in step S4.

In addition, in the state of FIG. 2D, the summary generating device 100 extracts, as target sentences, sentences whose subject is the target of user U1's interest, among the sentences included in the article content C1 (step S5). For example, the summary generation device 100 extracts, as target sentences, sentences in which proper nouns have been merged in step S3 from sentences included in the article content C1. Further, for example, the summary generation device 100 extracts, as target sentences, sentences in which the merged proper noun is estimated as the implicit subject in step S3 from the sentences in which the implicit subject was estimated in step S4.

The summary generation device 100 may extract all of the text CSN11, the text CSM12, and the text CSN13, or may extract any one of them. For example, assuming that a rule has been set in advance to extract sentences in which an implicit subject is estimated, the summary generation device 100 extracts sentences CSN11, CSM12, and CSN13, as shown in FIG. 2(e). , the sentence CSM12 is extracted as a target sentence.

Next, the summary generation device 100 obtains the tree structure of the sentence CSN12 by parsing the sentence CSN12, which is the target sentence, according to an arbitrary algorithm (step S6). An example of a tree structure corresponding to the text CSN12 and an example of summary generation processing using this tree structure will be described with reference to FIG.

FIG. 3 is a diagram showing an example of summary generation processing using a tree structure. According to the example of FIG. 3, the summary generating apparatus 100 obtains a tree structure in which 15 clauses (chunks) such as ch00 to ch14 are nodes by parsing the target sentence CSN12. In the example of FIG. 3, the clause ch00 corresponds to "tying", the clause ch01 corresponds to "in the sixth inning", the clause ch02 corresponds to "going ahead", and the clause ch03 corresponds to "double on the left wing line." , the clause ch04 corresponds to ``hattaga'', and the clause ch14 corresponds to ``changed''. Since the other clauses are also as shown in FIG. 3, description thereof will be omitted.

In addition, in this embodiment, sentences corresponding to a tree structure can be classified according to the number of branches branching from clauses. For example, a simple sentence without branches is defined as type 1, a sentence with two branches is type 2, a sentence with three branches is type 3, and a sentence with more than three branches and a complex branch is defined as nxm type. be able to. According to such rules, the text corresponding to the tree structure shown in FIG. 3 (text CSN12) can be defined as 2-3-3-3 type. Specifically, the tree structure shown in FIG. 3 branches into "2" at clause ch04, "3" at clause ch08, "3" at clause ch12, and " 3" branched. Therefore, the text corresponding to the tree structure (text CSN12) can be defined as 2-3-3-3 type.

In such a state, the summary generating apparatus 100 selects phrases that include words that match the interest of the user U1 and the user U1's preference for the interest. Then, the summary generation device 100 connects the selected clauses along the tree structure so that the number of branches branching from the clause is equal to or less than a predetermined number, thereby generating a summary sentence summarizing the contents of the article content C1. do.

A method for selecting clauses will be explained. For example, the summary generation device 100 selects, from among the clauses included in the tree structure, clauses containing words that are presumed to affect changes in user U1's emotions according to user U1's preference for an object of user U1's interest. do. According to the example so far, user U1's interests are "Team Ta" and "PY1". In addition, since user U1 is a fan of these, he likes (emphasizes, is interested in) "Team Ta" and "PY1". For this reason, the win or loss of "team Ta", various results (number of hits, number of home runs, progress to the final league, etc.) regarding "team Ta" and "PY1" can be expressed as positive emotions, or , can be a major factor in changing to negative emotions. Therefore, the summary generating apparatus 100 selects phrases containing words indicating the win or loss of "team Ta" and various results regarding "team Ta" and "PY1".

Note that if "team Ta" is a baseball team and "PY1 player" is a baseball player, it can be said that user U1 likes baseball itself. For users who like baseball in general, numerical information indicating match results and player performance is important information, and this may cause positive emotions or negative emotions. Therefore, from this point of view, the summary generating apparatus 100 may select phrases containing words indicating match results and various achievements.

In addition, the summary generation device 100 has a rule list in which information considered important in the category (for example, results and achievements in the case of sports category) is defined for each category (for example, sports category), The clause may be selected on a rule basis by referring to this list.

In addition, the summary generation device 100 selects clauses that are presumed to change the emotion value indicating the degree of the user's emotion according to the user U1's preference for the object of interest from among the clauses included in the tree structure. You may choose. The emotional value referred to here is an emotional value indicating a positive emotion and an emotional value indicating a negative emotion, that is, a "PN value". For example, each summary sentence that combines clauses to form a simple sentence without branching, that is, a sentence of type 1, is input to a model that outputs a PN value (a model that has learned the relationship between sentences and PN values). do. Then, the summary generation device 100 selects clauses forming a summary sentence whose output PN value is greater than a predetermined value. For example, the summary generating apparatus 100 selects phrases forming a summary sentence in which at least one of the emotion value indicating positive emotion and the emotion value indicating negative emotion is greater than a predetermined value.

In addition, the summary generating apparatus 100 selects words that are presumed to affect changes in user U1's emotion, or words that are presumed to change an emotion value indicating the degree of user U1's emotion, which are related to the current time. Clauses containing words may be preferentially selected. For example, in the case of baseball, during the month (month) when the league championship is decided, there are words specific to this period related to baseball (for example, the name of the team competing for the championship, the name of the player competing for the home run, the name of the leading hitter candidate, etc.) Appear at the top of search rankings. Therefore, the summary generating apparatus 100 may preferentially extract clauses containing words having such a seasonality (seasonality). Note that the summary generating apparatus 100 has a dictionary in which words that appear frequently in each category (for example, the baseball category) at what time period has been learned. You may identify words that have a habit (seasonality) of

In addition, the summary generating apparatus 100 may select a clause containing words (proper nouns) related to the object of user U1's interest and highly related to user U1's preferences. According to the example so far, user U1's interests are "Team Ta" and "PY1". User U1 also likes "team Ta" and "PY1" because he is a fan of them.

Here, taking “team Ta” as an example, the summary generation device 100 generates words that give a positive impression and words that give a negative impression to users who like “team Ta” (fans of “team Ta”). Using a positive-negative map in which the words that give .theta. Specifically, the summary generating apparatus 100 refers to the positive-negative map and identifies words whose inter-word distance from "team Ta" is equal to or less than a predetermined value, thereby selecting clauses containing the identified words. The generation of the positive/negative map based on the distance between words will be described with reference to FIG.

So far, we have explained the method of selecting phrases. After selecting clauses from the tree structure as described above, the summary generation device 100 connects the selected clauses along the tree structure so that the number of branches branching from the clauses is equal to or less than a predetermined number, thereby generating the article content C1. generates a summary sentence that summarizes the contents of For example, the summary generation device 100 generates a summary sentence summarizing the content of the article content C1 by connecting selected phrases so that the number of branches from the phrase is "2" or less. A sentence with a branch number of "2" or less is a simple type 1 sentence with no branch or a simple type 2 sentence with two branches. Here, the summary generating apparatus 100 generates a type 1 sentence by connecting selected phrases, and combines the generated type 1 sentences to form a summary sentence.

Here, for example, words such as "going ahead" and "double" are words that can be a major factor in changing the emotions of user U1 who is interested in "team Ta". In addition, the words indicating that "PY1" has "changed in the middle", that is, "midway" and "change" are words that can be a major factor in changing the emotions of the user who is interested in "PY1". be. For this reason, in the example of FIG. 3, the summary generation device 100 selects the phrase ch02 including "win" and the phrase ch03 including "double hit". Also, in the example of FIG. 3, the summary generation device 100 selects the phrase ch13 including "on the way" and the phrase ch14 including "change".

In this example, the summary generation device 100 selects the phrase ch02 and the phrase ch03, and further selects the phrase ch04 so as to form a type 1 sentence. Then, the summary generating apparatus 100 generates a type 1 sentence CSN121 by connecting clause ch02, clause ch03, clause ch04, clause ch13, and clause ch14 along the tree structure. Specifically, the summary generating apparatus 100 generates a type 1 sentence CSN121 from sentence CSN12, which is the object sentence, "I hit a double on the winning line in a timely manner, but I was replaced midway through."

In addition, in the example of FIG. 3, the summary generating apparatus 100 uses the same method to generate a type 1 sentence CSN111, "Team Ta's PY1 player is No. 30 with 2 runs." there is Further, the summary generating apparatus 100 determines that "PY1 is 'waiting and seeing'" from the sentence CSN13, which is the target sentence. ' is generated as a type 1 sentence CSN131.

Also, according to the examples so far, the sentence CSN14 is not extracted as a target sentence. However, the sentence CSN14 contains information about the game result that may change the user U1's emotions. For this reason, the summary generating apparatus 100 generates a type 1 text CSN141 also from the text CSN14. In this manner, the summary generation apparatus 100 may employ a sentence that is not the target sentence but is used as a sentence that constitutes a summary sentence, depending on the content of the sentence.

Then, the summary generating apparatus 100 combines the sentences CSN111, CSN121, CSN131 and CSN141 of each type 1 to finally generate a summary sentence SM1 as shown in FIG. As shown in FIG. 3, the summary sentence SM1 is a sentence that reflects the preferences based on the object of interest of the user U1. For example, with the conventional summarization technique, a summary sentence that does not include the type 1 sentence CSN131 is generated based on the sentence CSN13 that does not have a subject, or there is no conversion process to a type 1 sentence. There is a possibility that an insufficiently summarized summary sentence will be generated. However, according to the summary generation program according to the embodiment, it is possible to generate a coherent summary sentence that reflects the user's taste.

[5. Configuration of summary generation device]
Next, the summary generation device 100 according to the embodiment will be described with reference to FIG. FIG. 4 is a diagram showing a configuration example of the summary generation device 100 according to the embodiment. As shown in FIG. 4, the summary generation device 100 has a communication section 110, a storage section 120, and a control section .

(Regarding communication unit 110)
The communication unit 110 is realized by, for example, a NIC (Network Interface Card) or the like. The communication unit 110 is connected to the network N by wire or wirelessly, and transmits/receives information to/from a terminal device owned by a user, for example.

(Regarding storage unit 120)
The storage unit 120 is realized by, for example, a RAM (Random Access Memory), a semiconductor memory device such as a flash memory, or a storage device such as a hard disk or an optical disk. The storage unit 120 has a dictionary database 121 , a user information storage unit 122 , a content storage unit 123 and a summary information storage unit 124 .

(Regarding dictionary database 121)
The dictionary database 121 corresponds to, for example, a general term dictionary, a thesaurus dictionary, and a proper noun dictionary. Further, for example, in a proper noun dictionary, proper nouns with different expressions are associated with formal proper nouns. Further, the summary generation device 100 may learn terms that are influenced by the time period based on the user's behavior, and associate the learned terms with the time period.

(Regarding the user information storage unit 122)
The user information storage unit 122 stores user information indicating user interests. For example, when receiving registration of such user information from a user, the receiving unit 131, which will be described later, associates the received user information with identification information (user ID) that identifies this user, and registers the user information. Stored in the storage unit 122 .

(Regarding the content storage unit 123)
The content storage unit 123 stores content to be processed. For example, when the acquisition unit 132 to be described later acquires the content to be processed, it stores the acquired content to be processed in the content storage unit 123 . For example, when the reception unit 131 receives registration of content to be processed, the acquisition unit 132, which will be described later, registers the received content with respect to identification information (user ID) that identifies the user of the registration source. They are stored in the content storage unit 123 in association with each other.

(Regarding summary information storage unit 124)
The summary information storage unit 124 stores text information indicating a summary of the content to be processed. For example, when generating a summary sentence for a user in response to registration of user information by the user, the generation unit 134, which will be described later, generates A summary sentence is associated and stored in the summary information storage unit 124 .

Although not shown in FIG. 4, the summary generating apparatus 100 may further include a storage unit for storing models and positive/negative maps.

(Regarding the control unit 130)
The control unit 130 executes various programs (for example, a summary generation program according to the embodiment) stored in a storage device inside the summary generation device 100 by a CPU (Central Processing Unit), MPU (Micro Processing Unit), or the like. It is realized by being executed as a work area. Also, the control unit 130 is implemented by an integrated circuit such as an ASIC (Application Specific Integrated Circuit) or an FPGA (Field Programmable Gate Array).

As shown in FIG. 4, the control unit 130 includes a reception unit 131, an acquisition unit 132, an extraction unit 133, a generation unit 134, an estimation unit 135, an output unit 136, a first control unit 137, a 2 control unit 138, and implements or executes the information processing functions and actions described below. Note that the internal configuration of the control unit 130 is not limited to the configuration shown in FIG. 4, and may be another configuration as long as it performs information processing to be described later. Moreover, the connection relationship of each processing unit of the control unit 130 is not limited to the connection relationship shown in FIG. 4, and may be another connection relationship.

(Regarding the reception unit 131)
The accepting unit 131 accepts registration of user information indicating the user's interest. For example, the reception unit 131 receives user information transmitted from the user's terminal device in response to a user operation, and completes reception of user information registration when the user information is received. The reception unit 131 can also receive registration of content to be processed from the user.

(Regarding the acquisition unit 132)
The acquisition unit 132 acquires user information indicating the user's interest and content to be processed. In the example of FIG. 1, the acquiring unit 132 obtains user information indicating that the user information corresponding to the user U1 is a fan of "team Ta" (interested in "team Ta" or likes "team Ta"). getting information. Further, in the example of FIG. 1, the acquiring unit 132 obtains, as user information corresponding to user U1, a fan of "PY1 player" (interested in "PY1 player" or likes "PY1 player"). The user information shown is acquired. Further, in the example of FIG. 1, the acquisition unit 132 acquires a group of article contents for one month as content to be processed. Further, in the example of FIG. 1, the acquiring unit 132 acquires the article content C1 as the content to be processed after narrowing down the article content group for one month.

(Regarding the extraction unit 133)
The extraction unit 133 extracts a target sentence, which is a sentence in which the object of interest indicated by the user information is a topic, from sentences included in the content to be processed. For example, the extracting unit 133 extracts, as a target sentence, a sentence in which a target of interest to the user is a topic indicated by a proper noun in sentences included in the content to be processed, as a target of interest. .

More specifically, the extracting unit 133 selects a target indicated by a synonymous proper noun among a plurality of different proper nouns in a sentence included in the content to be processed, and a target in which the user is interested. Sentences are extracted as target sentences. In addition, the extraction unit 133 estimates the subject of sentences that do not contain a subject among the sentences included in the content to be processed, and targets the sentences that the estimated subject is the object indicated by the above-mentioned synonymous proper noun. Extract as sentences.

In addition, for this reason, the extraction unit 133 performs a series of extraction processes as described in steps S1 to S5 in FIGS. 1 and 2. FIG.

(Regarding the generation unit 134)
Based on the target sentence extracted by the extracting unit 133, the generating unit 134 generates a summary sentence that summarizes the contents of the content to be processed. For example, the generating unit 134 generates a summary sentence that summarizes the contents of the content to be processed based on the tree structure that is the result of parsing the target sentence. For example, the generating unit 134 generates a summary sentence that summarizes the contents of the content to be processed based on clauses that are nodes constituting a tree structure that is the result of parsing the target sentence.

As an example, the generation unit 134 may generate a tree structure based on a clause that includes a word that is presumed to affect changes in the user's emotion according to the user's preference for the target of interest of the user, among the clauses that make up the tree structure. Generate a summary sentence that summarizes the contents of the content to be processed. For more detailed information, the generation unit 134 generates clauses containing words that are presumed to change the emotional value indicating the degree of the user's emotion according to the user's preference for the object of interest of the user. Based on this, a summary sentence that summarizes the contents of the content to be processed is generated.

In addition, the generation unit 134 may generate a summary sentence that summarizes the contents of the content to be processed, preferentially using phrases containing words related to the current time.

In addition, the generation unit 134 preferentially uses phrases that include words that are related to the target of interest of the user and that are highly related to the user's preference, so that the contents of the processing target content are A condensed summary sentence may be generated.

In addition, the generation unit 134 connects the selected clauses so that the number of branches branching from the clause is equal to or less than a predetermined number along the tree structure that is the result of syntactic analysis of the target sentence. Generate a summary sentence that summarizes the contents of the content.

In addition, for this reason, the generation unit 134 performs a series of generation processes as described with reference to steps S6 and S7 in FIG. 1 and FIG.

(Regarding the estimation unit 135)
The estimation unit 135 estimates an emotion value (PN value) indicating the degree of each emotion with respect to the contents of the abstract for each emotion type, based on the user's preference for the user's interest target. For example, the estimation unit 135 estimates an emotion value indicating the degree of emotion with respect to the content of the summary for each of positive emotion (positive emotion) and negative emotion (negative emotion) as the type of user's emotion. do.

In addition, for example, the estimation unit 135 may obtain an arbitrary sentence from which proper nouns are excluded and an emotion value indicating the degree of emotion with respect to the content of the arbitrary sentence, which is a value evaluated by the user for the arbitrary sentence. Based on the model that has learned the relationship between and, an emotion value that indicates the degree of emotion toward the content of the summary is estimated. For example, the estimating unit 135, as such a model, outputs a higher emotion value for an emotion that corresponds to the relationship between the user's preference and the content of the abstract, out of the positive emotions and the negative emotions. Based on the learned model, an emotion value indicating the degree of emotion towards the content of the summary is estimated.

(Regarding the output unit 136)
The output unit 136 outputs a summary sentence summarizing the content to be processed to the user. For example, the output unit 136 outputs a summary sentence to which the PN value estimated by the estimation unit 135 is added as a summary sentence that summarizes the details of the content to be processed. For example, the output unit 136 transmits the PN-added summary to the terminal device of the user so that the PN-added summary is displayed on the user's terminal device.

(Regarding the first control unit 137)
The first control unit 137 learns the relationship between an arbitrary sentence and an emotion value, which is a value evaluated by the user for the arbitrary sentence and indicates the degree of emotion with respect to the contents of the arbitrary sentence. learn the model. Specifically, the first control unit 137 determines an arbitrary sentence from which proper nouns are excluded and a value evaluated by the user for the arbitrary sentence, and determines the degree of emotion with respect to the contents of the arbitrary sentence. Generate a model that has learned the relationship with the indicated emotional value. In addition, the emotion referred to here is a positive emotion (positive) emotion and a negative emotion (negative emotion).

Further, for example, the first control unit 137, as a model, outputs a higher emotion value for the emotion corresponding to the relationship between the user's preference and the content of the summary sentence, out of the positive emotion and the negative emotion. to learn the model.

A more detailed example is given. For example, an article about "Team Ta" winning causes positive feelings toward fans of "Team Ta", and an article about "Team Ta" losing causes positive feelings toward anti-"Team Ta". . On the other hand, an article about "team Ta" being defeated makes fans of "team Ta" feel negative feelings, and an article about "team Ta" winning makes people feel negative feelings about anti-"team Ta". . In this manner, which of the positive emotion and the negative emotion is stronger is determined according to the relationship between the user's preference and the content of the abstract.

Therefore, when using such an example, the first control unit 137 determines whether the summary sentence of the victory of "team Ta" for the fan user of "team Ta" is positive emotion, out of positive emotion and negative emotion. Train the model to increase the emotional value that indicates the degree. In addition, for the anti-“team Ta” user, the first control unit 137 increases the emotion value indicating the degree of the positive emotion, out of the positive emotion and the negative emotion, for the summary of the content of the defeat of “team Ta”. Train the model to improve.

On the other hand, for the fan user of "team Ta", the first control unit 137 sets an emotion value indicating the degree of negative emotion among the positive emotion and the negative emotion for the summary of the content of the defeat of "team Ta". Train the model to do better. For the anti-“team Ta” user, the first control unit 137 sets an emotion value indicating the degree of the negative emotion, out of the positive emotion and the negative emotion, for the summary of the content of the victory of “team Ta”. Train the model to do better.

An example of model learning will now be described in more detail. For example, the first control unit 137 is a sample sentence of type 1 sentences or type 2 sentences clarified in FIG. The relationship between the content of the sentence and the emotion value is learned by using a set of values obtained by the user evaluating the emotion value of the positive emotion or the emotion value of the negative emotion for the sample sentence as a learning data set.

For example, for a sample sentence (Type 1 sentence) that says, "I hit a double on the left wing line of the winning team, but I was replaced halfway through." Emotion value is evaluated for each of the sentences, such as "double hit", "change in the middle", etc., and the emotion value of the sentence as a whole is also evaluated. Therefore, in this example, the first control unit 137 performs learning using combinations of this sample sentence, the emotion value for each general term, and the emotion value of the sentence as a whole as a learning data set. Since the learning accuracy increases as the number of trials increases, it is desirable to use a large number of sample sentences.

In such a state, the first control unit 137 controls the number of branches included in the tree structure corresponding to the sample sentence, the number of clauses included in the tree structure corresponding to the sample sentence, and the number of clauses included in the tree structure corresponding to the sample sentence. A model is generated based on the number of lexical phrases (morphemes) used, the emotional value of a sentence as a whole composed of general terms, and the emotional value based on user evaluation. In addition, if the sentence does not contain proper nouns, it is difficult for the user's individual subjectivity to be reflected in the user evaluation. be able to.

Here, the user-evaluated emotion values may be user-evaluated emotion values for proper nouns instead of user-evaluated emotion values for general terms. However, since there are many proper nouns and the number of proper nouns changes over time, it is difficult for the user to evaluate all proper nouns. The second control unit 138, which will be described below, handles such situations.

(Regarding the second control unit 138)
The second control unit 138 calculates an inter-word distance indicating the relationship (similarity) between words. For example, the second control unit 138 calculates the inter-word distance between proper nouns and general terms. The second control unit 138 also clusters proper nouns and general terms based on general knowledge such as articles. The second control unit 138 then controls such that words within the same cluster automatically have the same semantic orientation. According to such control, as long as the user evaluates the emotion value only for the words in the cluster, the emotion value is dynamically calculated even for proper nouns whose emotion values are unknown because user evaluation has not been performed. will be As a result, the first control unit 137 can obtain an emotion value for a proper noun that has not been evaluated by the user because of its large number and the property that the number changes over time. , a more accurate model can be generated by using this emotion value.

For example, the second control unit 138 controls each word to have a semantic orientation in accordance with changes in individual preferences over time and in the current event according to the following rules. For example, the distance between words includes parameters such as size, field of view, and movement, and the size and field of view require the appearance frequency of words. Based on , the distance between words is calculated, and each word is controlled so that it has a semantic orientation according to changes in individual preferences over time and in the current event. The result of such control can be expressed as a positive-negative map in the form of a three-dimensional torus.

FIG. 5 shows an example of the positive/negative map PN-MP according to the embodiment. In the table TB shown on the left side of FIG. 5, for each "team" (team name), "players" (player names) belonging to the team are clustered. A list of "interword distances" calculated between each "athlete" is displayed.

In the example of the table TB shown in FIG. 5, in the team "Ta", the player "PY1a", the player "PY2a", and the player "PY3a" are clustered as one team "Ta". In the example of the table TB shown in FIG. 5, "74" is calculated as the inter-word distance between the word "Ta" indicating the team name and the word "PY1a" indicating the player name. Further, in the example of the table TB shown in FIG. 5, "64" is calculated as the inter-word distance between the word "Ta" indicating the team name and the word "PY2a" indicating the player name. In the example of the table TB shown in FIG. 5, "157" is calculated as the inter-word distance between the word "Ta" indicating the team name and the word "PY3a" indicating the player name.

Description of team “Tb”, team “Tc”, team “Td”, team “Te”, team “Tg”, and team “Tf” is omitted.

In the positive/negative map PN-MP shown on the right side of FIG. 5, the relationship of each word is shown in a three-dimensional torus format based on the "distance between words" in the table TB. For example, in the positive/negative map PN-MP, as shown in FIG. 5, the word "PY1a" indicating the player name is separated from the sphere CTa corresponding to the word "Ta" indicating the team name by an inter-word distance of "74". are arranged based on Also, the word “PY2a” indicating the player name is arranged based on the inter-word distance “64” with respect to the sphere CTa corresponding to the word “Ta” indicating the team name. Also, the word "PY3a" indicating the player name is arranged based on the inter-word distance "157" with respect to the sphere CTa corresponding to the word "Ta" indicating the team name. The size (diameter) of the sphere shown in the positive/negative map PN-MP corresponds to the appearance frequency of the word corresponding to this sphere.

In the positive/negative map PN-MP, each word is controlled to have a semantic orientation in accordance with changes in individual tastes. Therefore, for fan users of "team Ta", the color of the sphere CTa corresponding to the word "Ta" indicating the team name is controlled to a specific color (for example, red), and the player names are displayed. The colors of the spheres corresponding to the indicated words "PY1a", "PY2a", and "PY3a" are also controlled to similar colors (for example, red colors). That is, the aspect of the positive/negative map PN-MP is changed so as to have a semantic orientation according to individual preferences.

On the other hand, for the anti-"team Ta", the color of the sphere CTa corresponding to the word "Ta" indicating the team name is controlled to another specific color (for example, blue), and also indicates the player name. The colors of the spheres corresponding to the words "PY1a", "PY2a", and "PY3a" are also controlled to similar colors (for example, blue colors).

In this way, the aspect of the positive-negative map PN-MP can be changed to be semantically oriented according to individual preferences. Therefore, as described with reference to FIG. 3, the summary generating apparatus 100 can identify words whose word-to-word distance from "team Ta" is equal to or less than a predetermined value by referring to the positive/negative map PN-MP. become.

[6. Processing procedure]
Next, the procedure of summary generation processing according to the embodiment will be described with reference to FIG. FIG. 6 is a flowchart showing a summary generation processing procedure according to the embodiment.

First, the reception unit 131 determines whether or not registration of user information has been received from a user (step S101). While the receiving unit 131 determines that registration of user information has not been received (step S101; No), it waits until it can be determined that registration of user information has been received.

On the other hand, when the receiving unit 131 determines that registration of user information has been received (step S101; Yes), the obtaining unit 132 obtains the user information received according to the registration (step S102). For example, the acquisition unit 132 acquires user information from the user information storage unit 122 . The acquisition unit 132 also acquires content to be processed. For example, the acquisition unit 132 acquires content to be processed from the content storage unit 123 . It should be noted that when the content to be processed is also received by the receiving unit 131 due to the registration of the content to be processed by the user, the acquisition unit 132 acquires the content to be processed from the receiving unit 131 .

Next, the extraction unit 133 narrows down the content to be processed based on the user information (step S103). For example, the extraction unit 133 narrows down the content related to the user's interest from the content to be processed. The narrowed down content is also content to be processed. In addition, the extraction unit 133 removes unnecessary data by performing a cleansing process on the content to be processed after narrowing down (step S104).

Next, the extraction unit 133 merges proper nouns with respect to sentences included in the content to be processed after cleansing, based on the target of interest of the user indicated by the user information (step S105). Further, the extraction unit 133 estimates an implicit subject for sentences that do not contain a subject among sentences included in the content to be processed after cleansing, and inserts the estimated subject (step S106).

Next, the extraction unit 133 extracts a target sentence, which is a sentence in which the target of interest indicated by the user information is a hot topic, among the sentences in which the proper nouns have been merged and the sentences in which the implicit subject has been estimated. is extracted (step S107). For example, the extraction unit 133 extracts, as a target sentence, a sentence whose subject is a proper noun (formal proper noun) corresponding to the target of interest indicated by the user information.

Next, the generation unit 134 obtains a tree structure corresponding to the target sentence, which is the analysis result, by performing syntactic analysis on the extracted target sentence (step S108). Then, the generation unit 134 generates a summary based on the tree structure (step S109). For example, the generating unit 134 selects, from among the clauses included in the tree structure, clauses that include words that correspond to the user's preference for the object of interest of the user, and converts the selected sentences into Type 1 sentences. The process of connecting is performed for each sentence. Then, the generation unit 134 generates a summary sentence by combining Type 1 sentences.

Next, the estimation unit 135 estimates an emotion value (PN value) indicating the degree of emotion with respect to the content of the generated summary for each emotion type based on the user's preference (step S110). Then, the output unit 136 outputs the summary sentence to which the emotion value is assigned to the user (step S111).

[7. Hardware configuration]
Also, the summary generating apparatus 100 according to the above embodiment is implemented by a computer 1000 configured as shown in FIG. 7, for example. FIG. 7 is a hardware configuration diagram showing an example of a computer 1000 that implements the functions of the summary generating apparatus 100. As shown in FIG. Computer 1000 has CPU 1100 , RAM 1200 , ROM 1300 , HDD 1400 , communication interface (I/F) 1500 , input/output interface (I/F) 1600 and media interface (I/F) 1700 .

The CPU 1100 operates based on programs stored in the ROM 1300 or HDD 1400 and controls each section. The ROM 1300 stores a boot program executed by the CPU 1100 when the computer 1000 is started up, a program depending on the hardware of the computer 1000, and the like.

HDD 1400 stores programs executed by CPU 1100 and data used by such programs. Communication interface 1500 receives data from other devices via communication network 50 and sends the data to CPU 1100 , and transmits data generated by CPU 1100 to other devices via communication network 50 .

The CPU 1100 controls output devices such as displays and printers, and input devices such as keyboards and mice, through an input/output interface 1600 . CPU 1100 acquires data from an input device via input/output interface 1600 . CPU 1100 also outputs the generated data to an output device via input/output interface 1600 .

Media interface 1700 reads programs or data stored in recording medium 1800 and provides them to CPU 1100 via RAM 1200 . CPU 1100 loads such a program from recording medium 1800 onto RAM 1200 via media interface 1700, and executes the loaded program. The recording medium 1800 is, for example, an optical recording medium such as a DVD (Digital Versatile Disc) or a PD (Phase change rewritable disc), a magneto-optical recording medium such as an MO (Magneto-Optical disk), a tape medium, a magnetic recording medium, or a semiconductor memory. etc.

For example, when the computer 1000 functions as the abstract generation device 100 according to the embodiment, the CPU 1100 of the computer 1000 implements the functions of the control unit 130 by executing the program loaded on the RAM 1200 . In addition, data in storage unit 120 is stored in HDD 1400 . CPU 1100 of computer 1000 reads these programs from recording medium 1800 and executes them, but as another example, these programs may be obtained from another device via communication network 50 .

[8. others〕
Also, each component of each device illustrated is functionally conceptual, and does not necessarily need to be physically configured as illustrated. In other words, the specific form of distribution and integration of each device is not limited to the one shown in the figure, and all or part of them can be functionally or physically distributed and integrated in arbitrary units according to various loads and usage conditions. Can be integrated and configured.

As described above, the embodiments of the present application have been described in detail based on several drawings, but these are examples, and various modifications and It is possible to carry out the invention in other forms with modifications.

Also, the above-mentioned "section, module, unit" can be read as "means" or "circuit". For example, the acquisition unit can be read as acquisition means or an acquisition circuit.

100 summary generation device 120 storage unit 121 dictionary database 122 user information storage unit 123 content storage unit 124 summary information storage unit 130 control unit 131 reception unit 132 acquisition unit 133 extraction unit 134 generation unit 135 estimation unit 136 output unit 137 first control unit 138 second control unit

Claims (20)

an acquisition procedure for acquiring user information indicating a user's interest and content to be processed;
an extracting procedure for extracting target sentences, which are sentences in which an object of interest indicated by the user information is a topic, from sentences included in the content to be processed;
causing a computer to execute a generating procedure for generating a summary sentence summarizing the contents of the content to be processed based on the target sentence ;
The generation procedure influences a change in the user's emotion according to the user's preference for the target of interest among clauses that are nodes constituting a tree structure that is an analysis result of parsing the target sentence. generate a summary sentence summarizing the contents of the content to be processed, based on the clause containing the word that is inferred to be
A summary generator characterized by : The generating procedure performs the processing based on the phrase including, as the word, a word that is presumed to change an emotion value indicating the degree of emotion of the user according to the user's preference for the object of interest. 2. The program for generating a summary according to claim 1 , which generates a summary that summarizes the content of the target content. The generating step generates a summary sentence that summarizes the contents of the content to be processed, by preferentially using the clause containing words related to the current time among the words. 3. The abstract generation program according to 1 or 2 . an acquisition procedure for acquiring user information indicating a user's interest and content to be processed;
an extracting procedure for extracting target sentences, which are sentences in which an object of interest indicated by the user information is a topic, from sentences included in the content to be processed;
a generation procedure for generating a summary sentence summarizing the contents of the content to be processed based on the target sentence;
on the computer, and
The generation procedure generates words related to the object of interest among clauses that are nodes constituting a tree structure that is a result of syntactic analysis of the target sentence, and is more relevant to the user's preference. generate a summary sentence that summarizes the contents of the content to be processed by preferentially using phrases containing words with a high
A summary generator characterized by: The generation procedure generates a summary sentence summarizing the contents of the content to be processed by connecting the clauses along the tree structure so that the number of branches branching from the clause is equal to or less than a predetermined number. 5. The program for generating a summary according to any one of claims 1 to 4 , characterized by: an acquisition procedure for acquiring user information indicating a user's interest and content to be processed;
an extracting procedure for extracting target sentences, which are sentences in which an object of interest indicated by the user information is a topic, from sentences included in the content to be processed;
a generation procedure for generating a summary sentence that summarizes the contents of the content to be processed based on the target sentence;
an estimation step for estimating, for each type of emotion, an emotion value indicating the degree of emotion with respect to the content of the summary based on the user's preference for the object of interest;
A program for generating summaries on a computer. 7. The method according to claim 6 , wherein the estimation procedure estimates an emotion value indicating a degree of emotion with respect to the content of the summary sentence for each of positive emotion and negative emotion as the type of emotion of the user. The described abstract generator. The estimation procedure is a relationship between an arbitrary sentence from which proper nouns are excluded and an emotion value, which is a value evaluated by the user for the arbitrary sentence and indicates the degree of emotion with respect to the content of the arbitrary sentence. 8. The program for generating a summary according to claim 6 or 7 , wherein an emotion value indicating the degree of emotion for the content of the summary is estimated based on the model learned by the program. In the estimation procedure, the model is a model that has been learned to output a higher emotion value for the emotion corresponding to the relationship between the user's preference and the content of the summary sentence, out of the positive emotion and the negative emotion. 9. The program for generating a summary according to claim 8 , wherein an emotion value indicating a degree of emotion for the contents of the summary is estimated based on the above. 10. The method according to any one of claims 6 to 9 , wherein said generating step generates a summary sentence that summarizes the content of said content to be processed based on a tree structure that is a result of syntactic analysis of said target sentence. A summary generator according to any one of the preceding claims. 11. The program for generating a summary according to claim 10 , wherein the generation step generates a summary sentence summarizing the contents of the content to be processed based on clauses that are nodes constituting the tree structure. The extraction procedure is characterized in that, as the object of interest, a sentence, which is a target indicated by a proper noun included in the sentence and in which the target of interest indicated by the user information is a topic, is extracted as the target sentence. 12. The program for generating a summary according to any one of claims 1 to 11 . The extraction procedure extracts, as the target sentence, a sentence that is pointed to by synonymous proper nouns among a plurality of different proper nouns included in the sentence and that the target of interest indicated by the user information is a topic. 13. The program for generating a summary according to claim 12 , characterized by: The extraction procedure estimates a subject for sentences that do not contain a subject among sentences included in the content to be processed, and extracts sentences that have the estimated subject as a target indicated by the synonymous proper noun. 14. The program for generating a summary according to claim 13 , extracting as a sentence. an acquisition unit that acquires user information indicating a user's interest and content to be processed;
an extraction unit that extracts a target sentence, which is a sentence in which an object of interest indicated by the user information is a topic, from among the sentences included in the content to be processed;
a generation unit that generates a summary sentence that summarizes the contents of the content to be processed based on the target sentence ;
The generation unit influences a change in the user's emotion according to the user's preference for the target of interest among clauses that are nodes constituting a tree structure that is an analysis result of parsing the target sentence. generate a summary sentence summarizing the contents of the content to be processed, based on the clause containing the word that is inferred to be
A summary generation device characterized by: an acquisition unit that acquires user information indicating a user's interest and content to be processed;
an extraction unit that extracts a target sentence, which is a sentence in which an object of interest indicated by the user information is a topic, from among the sentences included in the content to be processed;
a generation unit that generates a summary sentence that summarizes the contents of the content to be processed based on the target sentence;
has
The generating unit selects words related to the object of interest among clauses that are nodes constituting a tree structure that is an analysis result of parsing the target sentence, and is more relevant to the user's preference. generate a summary sentence that summarizes the contents of the content to be processed by preferentially using phrases containing words with a high
A summary generation device characterized by: an acquisition unit that acquires user information indicating a user's interest and content to be processed;
an extraction unit that extracts a target sentence, which is a sentence in which an object of interest indicated by the user information is a topic, from among the sentences included in the content to be processed;
a generation unit that generates a summary sentence that summarizes the contents of the content to be processed based on the target sentence;
an estimation unit for estimating, for each type of emotion, an emotion value indicating the degree of emotion with respect to the content of the summary based on the user's preference for the object of interest;
A summary generation device characterized by comprising: A summary generation method executed by a summary generation device,
an acquisition step of acquiring user information indicating the user's interest and content to be processed;
an extracting step of extracting a target sentence, which is a sentence in which an object of interest indicated by the user information is a topic, from among the sentences included in the content to be processed;
a generation step of generating a summary sentence that summarizes the contents of the content to be processed, based on the target sentence ;
In the generation step, among clauses that are nodes constituting a tree structure that is an analysis result of parsing the target sentence, the user's preference for the target of interest influences fluctuations in the user's emotion. generate a summary sentence summarizing the contents of the content to be processed, based on the clause containing the word that is inferred to be
A summary generation method characterized by: A summary generation method executed by a summary generation device,
an acquisition step of acquiring user information indicating the user's interest and content to be processed;
an extracting step of extracting a target sentence, which is a sentence in which an object of interest indicated by the user information is a topic, from sentences included in the content to be processed;
a generation step of generating a summary sentence summarizing the contents of the content to be processed based on the target sentence;
including
The generating step generates words related to the target of interest among clauses that are nodes constituting a tree structure that is the result of syntactic analysis of the target sentence, and is more relevant to the user's preference. generate a summary sentence that summarizes the contents of the content to be processed by preferentially using phrases containing words with a high
A summary generation method characterized by: A summary generation method executed by a summary generation device,
an acquisition step of acquiring user information indicating the user's interest and content to be processed;
an extracting step of extracting a target sentence, which is a sentence in which an object of interest indicated by the user information is a topic, from sentences included in the content to be processed;
a generation step of generating a summary sentence summarizing the contents of the content to be processed based on the target sentence;
an estimation step for estimating, for each type of emotion, an emotion value indicating the degree of emotion with respect to the content of the summary based on the user's preference for the object of interest;
A method for generating a summary, comprising:

Images