JP7250058B2 - Information processing device, information processing method and information processing program - Google Patents

Description

The present invention relates to an information processing device, an information processing method, and an information processing program.

Conventionally, documents (texts) are classified (clustered) based on their written contents. For example, a technique is known in which a question sentence (first text) and an answer sentence (second text) that are associated with each other are obtained, and the answer sentences are classified based on the feature vector of each answer sentence that has been obtained.

JP 2012-3704 A

However, with conventional techniques, it is difficult to appropriately cluster a large number of acquired documents. For example, in a conventional method of clustering question sentences and answer sentences posted on a so-called Q&A (Question Answering) site using only the distance between centroids, it is difficult to appropriately adjust the granularity of clustering. It was difficult to accurately and appropriately grasp the overall picture of the topics in the category. Hierarchical clustering, in which small clusters are started and grouped together, often yields better clustering than clustering only once (only the first clustering) using text expressions. In this case, since the distance between centroids is usually used in the second clustering process after the first clustering process, the present invention further improves the hierarchical clustering.

The present application has been made in view of the above, and an object of the present application is to appropriately cluster texts associated with each other.

An information providing device according to the present application includes an acquisition unit that acquires a first text and a second text that are associated with each other; a generation unit for generating a bipartite graph by connecting a part, a cluster on the first text side formed by clustering, and a cluster on the second text side, and a group of clusters on the first text side in the bipartite graph; or a second clustering unit that re-clusters the cluster group on the second text side.

According to one aspect of the embodiment, each text associated with each other can be appropriately clustered.

FIG. 1 is a diagram showing an example of an information processing apparatus according to this embodiment. FIG. 2 is a diagram showing a configuration example of an information processing apparatus according to this embodiment. FIG. 3 is a schematic diagram showing a state in which question sentences and answer sentences acquired by the information processing apparatus according to the present embodiment are distributed. FIG. 4 is a schematic diagram showing an example of a bipartite graph generated by the information processing apparatus according to this embodiment. FIG. 5 is a flowchart showing an example of the flow of processing according to this embodiment. FIG. 6 is a diagram illustrating an example of a hardware configuration;

Embodiments for implementing an information processing apparatus, an information processing method, and an information processing program according to the present application (hereinafter referred to as "embodiments") will be described in detail below with reference to the drawings. The information processing apparatus, information processing method, and information processing program according to the present application are not limited to this embodiment.

[Embodiment]
[1. Overview of information processing device]
First, an example of an information processing apparatus 10 according to this embodiment will be described using FIG. In the following description, as an example of processing executed by the information processing apparatus 10, processing for clustering a large number of questions and answers posted on a so-called Q&A site will be described. In the description of the present application, the question sentence corresponds to the first text and the answer sentence corresponds to the second text. A question sentence and an answer sentence are associated with each other, and a plurality of answer sentences may be associated with one question sentence. This type of Q&A site allows users (questioners) to post questions and other users (answers) to post answers to the questions, thereby allowing users to share knowledge and wisdom. It is something to do. Further, in the following description, a user who instructs the information processing apparatus 10 to cluster question sentences and answer sentences is referred to as a user.

FIG. 1 is a diagram showing an example of an information processing apparatus according to this embodiment. FIG. 2 is a diagram showing a configuration example of an information processing apparatus according to this embodiment. The information processing device 10 shown in FIG. 1 is a device for clustering question texts (first texts) and answer texts (second texts), and is realized by, for example, a server device, a cloud system, or the like. For example, the information processing device 10 can be connected to various wireless communication networks such as 4G (Generation), 5G, LTE (Long Term Evolution), Wifi (registered trademark) or wireless LAN (Local Area Network), or networks such as various wired communication networks. N (see, for example, FIG. 2), communication is performed between the terminal device 100 used by the user and the web server 200 that manages the Q&A site. In the example of FIG. 1 , the information processing device 10 executes clustering processing of question sentences and answer sentences belonging to a desired range (category) in response to an instruction from the terminal device 100 (user), for example. The terminal device 100 is provided with the result.

In addition, the question sentence and the answer sentence (document) in this embodiment shall be text data containing one or more sentences (sentence). Also, a sentence is assumed to be a range of text data delimited by periods, exclamation marks, question marks, blanks, and the like. A question sentence is a document processed by the information processing apparatus 10 as if it was created with the intention of asking a question. An answer sentence is a document processed by the information processing apparatus 10 as if it was created with the intention of answering a question sentence.

The terminal device 100 is implemented by a smart device such as a PC (Personal Computer), a server device, a smart television, a smartphone, a tablet, or the like, and can communicate with the information processing device 10 via the network N. It is a mobile terminal device. In addition, the terminal device 100 has a screen such as a liquid crystal display, which has a touch panel function. It may have a function capable of accepting various operations on the content to be received.

Note that the terminal device 100 is used by the user U in the example shown in FIG. In addition, although one user U is described in the example shown in FIG. 1, the present invention is not limited to this. Since an arbitrary number of users can connect to the information processing apparatus 10 via respective terminal devices, each user can obtain the results of the clustering process.

The web server 200 is an information processing device that manages the Q&A site, and is realized by, for example, a server device, a cloud system, or the like. The web server 200 receives questions and answers posted by users, and stores them in association with answers corresponding to the questions. The web server 200 also provides users with questions and answers found on the Q&A site. In the present embodiment, the web server 200 is described as being separate from the information processing apparatus 10, but they may be configured integrally.

[2. About an example of processing]
Processing performed by the information processing apparatus 10 will be described. First, as shown in FIG. 1, the user U designates a desired category via the terminal device 100 and instructs clustering processing of question sentences and answer sentences (step S1). In this embodiment, for example, the user indicates the clustering results of question sentences and answer sentences belonging to a category for which the user desires the processing result. It may be a sentence and an answer sentence.

Subsequently, the information processing device 10 acquires question texts and answer texts belonging to the specified category from the web server 200 (step S2). For example, when the user U designates the “hobby” category, the information processing apparatus 10 acquires from the web server 200 question sentences and answer sentences belonging to the “hobby” category. Here, the question sentences and answer sentences are converted (vectorized) into N-dimensional vector data in advance by machine learning using a predetermined learning model (for example, an autoencoder). Therefore, the information processing apparatus 10 can extract and acquire all question sentences and answer sentences belonging to the designated category based on the feature amounts of these vector data. Further, the question sentence and the answer sentence corresponding to the question sentence are associated with each other by, for example, being linked by a common ID (question ID) or the like. Therefore, the information processing apparatus 10 can acquire a plurality of sets of question texts and answer texts that belong to the specified category and are associated with each other.

FIG. 3 is a schematic diagram showing a state in which question sentences and answer sentences acquired by the information processing apparatus according to the present embodiment are distributed. In FIG. 3, symbols Q1, Q2, . . . indicate question sentences, and symbols A1, A2, . Groups (pairs) with the same numbers, such as question sentence Q1 and answer sentence A1, are question sentences and answer sentences that are associated with each other. Also, symbols L1, L2, . In the example of FIG. 3, for example, two answer sentences A6 and A6' are associated with the question sentence Q6.

[2-1. An example of the first clustering process]
Next, the first clustering process will be explained. The information processing apparatus 10 executes a first clustering process of clustering the obtained plural question sentences and plural answer sentences with a predetermined granularity (step S3). In the present embodiment, the first clustering process is a process for classifying into clusters with a relatively fine granularity, and the granularity of the clusters is finer (smaller) than the clusters formed in the second clustering process described later. It's becoming

The information processing apparatus 10 classifies question sentences and answer sentences based on expressions generated by expression learning using, for example, similarity models (for example, DSSM (Deep Structured Semantic Models)). In this case, the information processing apparatus 10 arranges question sentences and answer sentences that are close in expression, that is, have a high degree of similarity, closer to each other to form clusters, so clustering with fine granularity is possible. Note that the technique of the first clustering process is not limited to this, and other techniques may be used as long as they are clustered more finely than the clusters generated in the second clustering process.

In the example of FIG. 3, the information processing apparatus 10 has a question-side cluster QC1 including question sentences (Q1 to Q8) with high similarity, and an answer-side cluster AC1 including answer sentences (A2 to A7, A9) with high similarity. and Although one question-side cluster and one answer-side cluster are shown in FIG. 3, it is assumed that there are a plurality (a large number) of these clusters. In the example of FIG. 3, each cluster is formed by including eight to nine question sentences and answer sentences. make it possible.

Subsequently, the information processing apparatus 10 evaluates the strength of the connection between the formed question-side cluster QC1 and answer-side cluster AC1 (step S4). In this embodiment, a plurality of question-side clusters and a plurality of answer-side clusters formed by the first clustering process are regarded as nodes, and a bipartite graph is generated according to the connections between these nodes. Prior to generating a bipartite graph, the information processing apparatus 10 evaluates the strength of connection between a plurality of question-side clusters and a plurality of answer-side clusters regarded as nodes.

This evaluation is made based on the correspondence between question sentences and answer sentences included in each cluster. That is, since question sentences and answer sentences are originally associated with each other, clusters containing a large number of question sentences and answer sentences that are associated with each other are considered to have strong ties. For this reason, in the present embodiment, the information processing apparatus 10 defines the strength of connection between clusters by the number of question sentences and answer sentences included in each cluster and associated with each other. is equal to or greater than a predetermined value, it is evaluated that the ties between the clusters are strong. In the example of FIG. 3, the information processing apparatus 10 evaluates that the strength of connection is equal to or greater than a predetermined value, and connects the question-side cluster QC1 and the answer-side cluster AC1 with the edge E of the bipartite graph.

[2-2. An example of bipartite graph generation processing]
FIG. 4 is a schematic diagram showing an example of a bipartite graph generated by the information processing apparatus according to this embodiment. In the example of FIG. 4, five question side clusters QC1 to QC5 are shown as the question side cluster group, and six answer side clusters AC1 to AC6 are shown as the answer side cluster group.

Subsequently, the information processing apparatus 10 generates the bipartite graph 300 based on the strength of the connection between clusters (step S5). As described above, the information processing apparatus 10 evaluates the strength of connection between a plurality of question-side clusters and a plurality of answer-side clusters, and connects clusters with strong connections with edges E to form a bipartite graph 300. forming. In the present embodiment, as shown in FIG. 4, if the strength of connection between clusters is equal to or greater than a predetermined value, the information processing apparatus 10 connects one cluster and a plurality of clusters with an edge E to connect two clusters. Generate a partial graph 300 .

[2-3. An example of the second clustering process]
Subsequently, the information processing apparatus 10 executes the second clustering process using the generated bipartite graph 300 (step S6). For example, the information processing device 10 aggregates each cluster based on the connection between the question side clusters QC1 to QC5 and the answer side clusters AC1 to AC6 and performs re-clustering (second clustering process). Specifically, two answer-side clusters AC1 and AC2, which are connected to one question-side cluster QC1 by an edge E, are assumed to be strongly connected via the question-side cluster QC1. For this reason, the information processing apparatus 10 aggregates a plurality of clusters connected to a common cluster, thereby performing clustering into aggregated clusters with a larger granularity.

In the example of FIG. 4, the question side clusters QC1 and QC2 are aggregated to form an aggregated question side cluster SQC1, and the answer side clusters AC1 and AC2 are aggregated to form an aggregate answer side cluster SAC1. Further, the question side clusters QC4 and QC5 are aggregated to form an aggregate question side cluster SQC2, and the answer side clusters AC4, AC5 and AC6 are aggregated to form an aggregate answer side cluster SAC2. In this case, the aggregated question-side cluster SQC1 and aggregated answer-side cluster SAC1 are further aggregated, and the aggregated question-side cluster SQC2 and aggregated answer-side cluster SAC2 are further aggregated to obtain a question-answer group with a high degree of similarity. can also be formed.

Thus, in this embodiment, the strength of the connection between the question-side cluster and the answer-side cluster is evaluated, and based on the strength of this connection, a bipartite graph in which the question-side cluster and the answer-side cluster are regarded as nodes By generating , for example, question sentences or answer sentences can be clustered more appropriately than clustering using only the distance between centroids. Therefore, clustering can be performed with an appropriate granularity that is neither too coarse nor too fine, so that, for example, an overall picture of topics within a desired category on a Q&A site can be accurately and appropriately grasped.

Subsequently, the information processing apparatus 10 provides the user with the result information re-clustered by the second clustering process (step S7), and ends the process.

[3. About the configuration of the information processing device]
An example of the functional configuration of the information processing apparatus 10 described above will be described below. Note that the following description shows an example of the functional configuration of the information processing apparatus 10 that executes processing for generating a summary of a question on a Q&A site. As shown in FIG. 2 described above, the information processing device 10 has a communication unit 20, a storage unit 30, and a control unit 40. As shown in FIG.

The communication unit 20 is realized by, for example, a NIC (Network Interface Card) or the like. The communication unit 20 is connected to the network N by wire or wirelessly, and transmits and receives information to and from the terminal device 100 .

The storage unit 30 is implemented by, for example, a semiconductor memory device such as a RAM (Random Access Memory) or a flash memory, or a storage device such as a hard disk or an optical disk. The storage unit 30 also has a question/answer storage unit 31 and a learning model storage unit 32 .

The question-and-answer storage unit 31 stores a plurality of sets of corresponding question sentences and answer sentences acquired from the web server 200 . The question sentence and the answer sentence are linked by, for example, a common ID (question ID). A plurality of answer sentences may correspond to the question sentence. Note that the question ID is an identifier for identifying a question sentence.

The learning model storage unit 32 stores a plurality of learning models that are used when executing the various types of processing described above. These learning models can be realized by various classifiers such as DNN (Deep Neural Network). The DNN can employ a neural network having an arbitrary configuration, such as RNN (Recurrent Neural Network), CNN (Convolution Neural Network), LSTM (Long short-term memory), or the like.

The control unit 40 is, for example, a controller, and various programs stored in a storage device inside the information processing apparatus 10 are stored in the RAM as a work area by a CPU (Central Processing Unit), MPU (Micro Processing Unit), or the like. It is realized by executing as Also, the control unit 40 is a controller, and 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. 2, the control unit 40 includes an acquisition unit 41, a first clustering processing unit (first clustering unit) 42, an evaluation unit 43, a generation unit 44, a second clustering processing unit (second clustering unit) 45, and It has a providing unit 46 . The acquisition unit 41 acquires from the web server 200 question texts and answer texts belonging to the specified category. For example, when the user U designates the “hobby” category, the acquisition unit 41 acquires from the web server 200 question texts and answer texts belonging to the “hobby” category. Here, the question sentences and answer sentences are converted (vectorized) into N-dimensional vector data by machine learning using a predetermined learning model in advance. Therefore, the acquiring unit 41 can extract and acquire all the question sentences and answer sentences belonging to the specified category based on the feature amounts of these vector data. Also, the answer sentences corresponding to the question sentences are associated with each other by, for example, being linked by a common ID (question ID) or the like. Therefore, the acquisition unit 41 can acquire a plurality of sets of question texts and answer texts that belong to the designated category and are associated with each other. Acquisition unit 41 may request web server 200 for question texts and answer texts belonging to the “hobbies” category, and acquire these question texts and answer texts from web server 200 .

The first clustering processing unit 42 clusters the obtained plural question sentences and plural answer sentences with a predetermined granularity. In the present embodiment, the first clustering processing unit 42 executes a process of classifying into clusters with relatively fine granularity, and the granularity of the clusters is finer (smaller) than the clusters formed by the second clustering processing unit 45, which will be described later. ) granularity.

The first clustering processing unit 42 classifies the question sentences and the answer sentences based on expressions generated by expression learning using, for example, a similarity model (for example, DSSM (Deep State Space Model). In this case, , the first clustering processing unit arranges the question sentences and the answer sentences closer to each other than the question sentences and the answer sentences that are closer in expression, that is, closer to each other than the similarity, so that fine-grained clustering is possible. The processing technique is not limited to this, and other techniques may be used as long as they are clustered more finely than the clusters generated in the second clustering process.

In the example of FIG. 3, the information processing apparatus 10 has a question-side cluster QC1 including question sentences (Q1 to Q8) with high similarity, and an answer-side cluster AC1 including answer sentences (A2 to A7, A9) with high similarity. and Although one question-side cluster and one answer-side cluster are shown in FIG. 3, it is assumed that there are a plurality (a large number) of these clusters.

The evaluation unit 43 evaluates the strength of the connection between the formed question-side cluster QC1 and answer-side cluster AC1. In this embodiment, a plurality of question-side clusters and a plurality of answer-side clusters formed by the first clustering process are regarded as nodes, and a bipartite graph is generated according to the connections between these nodes. Prior to generating a bipartite graph, the evaluation unit 43 evaluates the strength of connection between a plurality of question-side clusters and a plurality of answer-side clusters regarded as nodes.

In this embodiment, this evaluation is made based on the correspondence between the question sentences and answer sentences included in each cluster. That is, since question sentences and answer sentences are originally associated with each other, clusters containing a large number of question sentences and answer sentences that are associated with each other are considered to have strong ties. Therefore, the evaluation unit 43 defines the strength of the connection between clusters by the number of question sentences and answer sentences included in each cluster and associated with each other. is evaluated as having strong ties between clusters. Therefore, as shown in FIG. 3, when the evaluation unit 43 evaluates that the strength of connection is equal to or greater than a predetermined value, the question-side cluster QC1 and the answer-side cluster AC1 are connected by the edge E of the bipartite graph. do.

The generator 44 generates a bipartite graph 300 based on the strength of connection between clusters, as shown in FIG. As described above, the generation unit 44 evaluates the strength of connection between a plurality of question-side clusters and a plurality of answer-side clusters, and forms a bipartite graph 300 by connecting clusters with strong connections with edges E. do. If the strength of connection between clusters is equal to or greater than a predetermined value, the evaluation unit 43 determines that one cluster (for example, question-side cluster QC1) and a plurality of clusters (for example, answer-side clusters AC1 and AC2) each have an edge E concatenated with

The second clustering processing unit 45 uses the generated bipartite graph 300 to perform the second clustering processing (re-clustering). For example, the second clustering processing unit 45 performs re-clustering by aggregating each cluster based on the connection between the question side clusters QC1 to QC5 and the answer side clusters AC1 to AC6. Specifically, two answer-side clusters AC1 and AC2 connected by an edge E to one question-side cluster QC1 are strongly connected via the question-side cluster QC1. For this reason, the second clustering processing unit 45 aggregates a plurality of clusters connected to a common cluster, thereby performing clustering into aggregated clusters with greater granularity. In the example of FIG. 4, the second clustering processing unit 45 aggregates the question side clusters QC1 and QC2 to form an aggregated question side cluster SQC1, and aggregates the answer side clusters AC1 and AC2 to form an aggregated answer side cluster SAC1. are doing. The second clustering processing unit 45 aggregates the question side clusters QC4 and QC5 to form an aggregated question side cluster SQC2, and aggregates the answer side clusters AC4, AC5 and AC6 to form an aggregated answer side cluster SAC2. there is

The providing unit 46 provides the user with the result information re-clustered by the second clustering process.

[4. Processing procedure]
Next, an example of the flow of processing executed by the information processing apparatus 10 according to this embodiment will be described with reference to FIG. FIG. 5 is a flowchart showing an example of the flow of processing according to this embodiment. As shown in FIG. 5, the information processing apparatus 10 acquires a question sentence and an answer sentence regarding a designated category (step S101). Subsequently, the information processing apparatus 10 performs the first clustering process on each of the acquired question sentences and good answer sentences (step S102). Subsequently, the information processing apparatus 10 evaluates the strength of connection of each formed cluster (step S103). Subsequently, the information processing apparatus 10 generates a bipartite graph based on the evaluated strength of connection between the clusters (step S104). Subsequently, the information processing apparatus 10 executes the second clustering process using the generated bipartite graph (step S105). Subsequently, the information processing apparatus 10 provides the result information to the user (step S106), and ends the process.

[5. Modification]
The information processing apparatus 10 described above may be embodied in various forms other than the above embodiment. Therefore, other embodiments of the information processing apparatus 10 will be described below.

[5.1. Regarding clustering target]
In the above-described embodiment, as an example of the first text and the second text that are associated with each other, the question text and the answer text posted on the Q&A site are targeted, but the text is not limited to this. For example, the above-described processes may be executed with news posted on a news site as the first text and news comments posted on this news as the second text. Also, a news comment posted on a news posted on a news site may be the first text, and a reply comment to this news comment may be the second text.

[5.2. Regarding the first clustering process]
Further, as the first clustering process, distance learning may be performed on the acquired question sentences and answer sentences, and the distance-learned question sentences and answer sentences may be clustered. In this distance learning, similar data are learned near the central point of a predetermined semantic space, and dissimilar data are learned far from the central point of the semantic space. The information processing device 10 can execute Mahalanobis distance learning, for example. When executing this Mahalanobis distance learning, a set of similar data and a set of dissimilar data are prepared as an example of user instructions (teaching data). The information processing apparatus 10 learns a predetermined covariance matrix using these similar data sets and dissimilar data sets, and calculates vector data of question sentences and answer sentences using this learned covariance matrix. As a result, the distance between question sentences (answer sentences) having a high degree of similarity can be shortened, so clustering with relatively fine granularity is possible. Moreover, in this configuration, it is preferable to have a function (for example, a user interface) for the user to determine and instruct the granularity from the terminal device 100 .

[5.3. Regarding the second clustering process]
In the above-described embodiment, the first clustering process and the second clustering process are performed in two layers. (n hierarchies in total), and one or more of them may be configured to apply the second clustering process according to the present invention. Here, re-clustering using centroids can be performed on other layers to which the second clustering process has not been applied.

[5-4. Bipartite graph generation processing]
In the above-described embodiment, the edge E of the bipartite graph is connected if the connection between clusters is above a certain level (1 or 0). A configuration (for example, a weighted graph) may be used in which the edge E is weighted.

Moreover, the conventional method in hierarchical clustering that uses the closeness between centroids is not necessarily exclusive with the present invention. It is also possible to make a combination that considers both "connection of

[6. effect〕
As described above, the information processing apparatus 10 includes the acquisition unit 41 that acquires the question sentences and the answer sentences that are associated with each other, and the first clustering process that clusters the acquired plurality of question sentences and the plurality of answer sentences. A generation unit 44 that generates a bipartite graph 300 by connecting a part 42, a question-side cluster QC1 formed by clustering, and an answer-side cluster AC1, question-side clusters QC1, QC2, . . . Alternatively, the second clustering processing unit 45 for re-clustering the answer-side clusters AC1, AC2, . can be clustered. Therefore, clustering can be performed with an appropriate granularity that is neither too coarse nor too fine, so that, for example, an overall picture of topics within a desired category on a Q&A site can be accurately and appropriately grasped.
That is, in the information processing apparatus 10 according to the present embodiment, the question sentence (first text) or the answer sentence (second text) can be clustered more appropriately than the conventional method of clustering using only the distance between centroids. and good clustering performance. In addition, it is more appropriate than the conventional method of clustering using only text expressions for question sentences (first text) or answer sentences (second text), or the conventional method of hierarchical clustering using only the distance between centroids. , and the clustering performance is good.

In addition, as described above, since clustering can be performed with an appropriate granularity, for example, it is possible to grasp topics preferred by users who post on Q&A sites and news sites, and recommend topics that are highly similar to these topics to users. can do. Therefore, it is possible to improve the rate of users' site visits (probability of repeatedly visiting sites).

The information processing device 10 also has an evaluation unit 43 that evaluates the strength of connection between the question-side cluster QC1 and the answer-side cluster AC1. Since the bipartite graph 300 is generated by connecting the clusters, re-clustering using this bipartite graph 300 can be performed appropriately. In addition, the second clustering processing unit 45 aggregates a plurality of answer-side clusters AC1 and AC2 connected to the common question-side cluster QC1 in the bipartite graph 300 to form an aggregate answer-side cluster SAC1. The cluster granularity can be adjusted to an appropriate size.

[7. Hardware configuration]
Also, the information processing apparatus 10 according to the above-described embodiments is implemented by a computer 1000 configured as shown in FIG. 6, for example. FIG. 6 is a hardware configuration diagram showing an example of a computer that implements the functions of the information processing apparatus 1 according to the embodiment. 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.

The HDD 1400 stores programs executed by the CPU 1100, data used by the programs, and the like. Communication interface 1500 receives data from other devices via network (communication network) N, sends the data to CPU 1100, and transmits data generated by CPU 1100 to other devices via network N. FIG.

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 information processing apparatus 10 according to the embodiment, the CPU 1100 of the computer 1000 implements the functions of the control section 40 by executing programs loaded on the RAM 1200 . The CPU 1100 of the computer 1000 reads these programs from the recording medium 1800 and executes them, but as another example, these programs may be acquired via the network N from another device.

As described above, some of the embodiments of the present application have been described in detail based on the drawings. It is possible to carry out the invention in other forms with modifications.

[8. others〕
Further, among the processes described in the above embodiments and modifications, all or part of the processes described as being performed automatically can be performed manually, or described as being performed manually. All or part of the processing can also be performed automatically by known methods. In addition, information including processing procedures, specific names, various data and parameters shown in the above documents and drawings can be arbitrarily changed unless otherwise specified. For example, the various information shown in each drawing is not limited to the illustrated information.

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. For example, the acquisition unit 41, the first clustering processing unit 42, the evaluation unit 43, the generation unit 44, the second clustering processing unit 45, or the provision unit 46 may be connected via a network as external devices of the information processing apparatus 10. . In addition, different devices each have the acquisition unit 41, the first clustering processing unit 42, the evaluation unit 43, the generation unit 44, the second clustering processing unit 45, or the provision unit 46, and are connected to a network to cooperate. The functions of the information processing apparatus 10 described above may be realized. That is, the information processing apparatus 10 may be realized by a plurality of server computers, and depending on the function, an external platform or the like may be called using an API (Application Programming Interface), network computing, or the like. can be changed to

Also, the above-described embodiments and modifications can be appropriately combined within a range that does not contradict the processing content.

Also, the above-mentioned "section, module, unit" can be read as "means" or "circuit". For example, the acquisition unit 41 can be read as acquisition means or an acquisition circuit.

REFERENCE SIGNS LIST 10 information processing device 20 communication unit 30 storage unit 31 question and answer storage unit 32 learning model storage unit 40 control unit 41 acquisition unit 42 first clustering processing unit (first clustering unit)
43 evaluation unit 44 generation unit 45 second clustering processing unit (second clustering unit)
46 providing unit 100 terminal device 300 bipartite graph Q1 question sentence (first text)
A1 Answer sentence (second text)
QC1 question side cluster (first text side cluster)
AC1 Answer side cluster (Second text side cluster)

Claims (5)

an acquisition unit that acquires the first text and the second text that are associated with each other;
a first clustering unit that clusters the plurality of first texts and the plurality of second texts obtained;
a generation unit that generates a bipartite graph by connecting clusters on the first text side formed by clustering and clusters on the second text side;
a second clustering unit that re-clusters the cluster group on the first text side or the cluster group on the second text side in the bipartite graph;
An information processing device comprising: an evaluation unit that evaluates the strength of connection between the clusters on the first text side and the clusters on the second text side;
2. The information processing apparatus according to claim 1, wherein the generation unit generates the bipartite graph by connecting clusters whose evaluated connection strength is equal to or greater than a predetermined value. 3. The information processing apparatus according to claim 1, wherein said second clustering unit aggregates a plurality of clusters connected to a common cluster in said bipartite graph. An information processing method executed by an information processing device,
an obtaining step of obtaining a first text and a second text that are associated with each other;
a first clustering step of respectively clustering the obtained plurality of first texts and the plurality of second texts;
a generation step of generating a bipartite graph by connecting clusters on the first text side formed by clustering and clusters on the second text side;
a second clustering step of re-clustering the cluster group on the first text side or the cluster group on the second text side in the bipartite graph;
An information processing method comprising: an obtaining step of obtaining a first text and a second text that are associated with each other;
a first clustering step of respectively clustering the obtained plurality of first texts and the plurality of second texts;
a generation step of generating a bipartite graph by connecting clusters on the first text side formed by clustering and clusters on the second text side;
a second clustering step of re-clustering the cluster group on the first text side or the cluster group on the second text side in the bipartite graph;
An information processing program characterized by causing a computer to execute

Images