JP7042720B2 - Information processing equipment, information processing methods, and programs - Google Patents

Description

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

When content such as a news article is provided to a user via the Internet or the like, content related to the content may also be provided. In this regard, there is known a technique for estimating the importance of content and preferentially displaying the estimated content with high importance (see, for example, Patent Document 1).

Japanese Unexamined Patent Publication No. 2017-59057

However, in the conventional technology, it may not be possible to provide the user with other time-series-related content together with the target content for the target content that the user has browsed or may browse in the future. rice field.

The present invention has been made in view of the above problems, and an object of the present invention is to provide an information processing device, an information processing method, and a program capable of providing a user with content related to a target content and a time series. It is supposed to be.

One aspect of the present invention is based on an extraction unit that extracts a keyword from a predetermined number of contents having a high degree of similarity to a certain content of interest among a plurality of contents, and the keyword extracted by the extraction unit. It is an information processing device including a selection unit for selecting the content of interest and one or more related contents related to a time series from a plurality of contents.

According to one aspect of the present invention, the target content and the content related to the time series can be provided to the user.

It is a figure which shows an example of the information processing system 1 including the information processing apparatus 100 of an embodiment. It is a figure which shows an example of the content displayed on the screen of the 1st terminal apparatus 10. It is a figure which shows an example of the content displayed on the screen of the 1st terminal apparatus 10. It is a figure which shows an example of the structure of the information processing apparatus 100 in an embodiment. It is a figure which shows an example of content data 132. It is a figure which shows an example of the user log data 134. It is a flowchart which shows the flow of a series of processing of the control unit 110 in an embodiment. It is a figure for demonstrating the selection process of the related content. It is a figure for demonstrating the method of grouping of related contents. It is a figure for demonstrating the method of grouping of related contents. It is a figure for demonstrating the method of grouping of related contents. It is a figure for demonstrating the method of grouping of related contents. It is a figure for demonstrating the method of grouping of related contents. It is a figure for demonstrating the method of grouping of related contents. It is a figure for demonstrating the method of grouping of related contents. It is a figure for demonstrating the selection method of representative-related contents. It is a figure which shows an example of the hardware composition of the information processing apparatus 100 of an embodiment.

Hereinafter, an information processing apparatus, an information processing method, and a program to which the present invention is applied will be described with reference to the drawings.

[Overview]
The information processing device is realized by one or more processors. The information processing device extracts keywords from a predetermined number of high-ranking contents having a high degree of similarity to one content of interest (hereinafter referred to as the content of interest) among a plurality of contents, and a plurality of keywords are extracted based on the extracted keywords. Select one or more related contents related to the content of interest and the time series from the contents of. This makes it possible to provide the user with the target content and the content related to the time series.

The content in the present embodiment is, for example, an article posted on a blog, a website, or the like, and is content including text. This article may be an article that deals with the theme of social events (current affairs) that change from moment to moment, such as news, politics, economy, and sports. In such an article, if there is an article that is the first report on an event and then the event changes over time, an article that conveys the change may be provided as a follow-up report. In the following description, as an example, the content is assumed to be a document such as a news article.

[overall structure]
FIG. 1 is a diagram showing an example of an information processing system 1 including the information processing apparatus 100 of the embodiment. The information processing system 1 in the embodiment includes, for example, one or more first terminal devices 10, one or more second terminal devices 20, and an information processing device 100. These devices are connected to each other, for example, via a network NW.

Each device shown in FIG. 1 transmits and receives various information via the network NW. The network NW includes, for example, the Internet, a WAN (Wide Area Network), a LAN (Local Area Network), a provider terminal, a wireless communication network, a wireless base station, a dedicated line, and the like. It should be noted that not all combinations of the devices shown in FIG. 1 need to be able to communicate with each other, and the network NW may include a local network in part.

The first terminal device 10 is a terminal device including an input device, a display device, a communication device, a storage device, and an arithmetic unit, such as a mobile phone such as a smartphone, a tablet terminal, and various personal computers. The communication device includes a network card such as a NIC (Network Interface Card), a wireless communication module, and the like. In the first terminal device 10, a UA (User Agent) such as a web browser or an application program is activated, and a request according to a user's input content is transmitted to the information processing device 100. Further, the first terminal device 10 in which the UA is activated causes the display device to display various images based on the information acquired from the information processing device 100.

The second terminal device 20 is a terminal device including an input device, a display device, a communication device, a storage device, and an arithmetic unit, such as a mobile phone such as a smartphone, a tablet terminal, and various personal computers. The communication device includes a network card such as a NIC, a wireless communication module, and the like. For example, an employee of a company or business such as mass media, or an individual such as a journalist or a blogger operates the second terminal device 20 to upload contents such as news articles to the information processing device 100.

The information processing device 100 may be, for example, a web server that provides a web page to the first terminal device 10 in response to a request from a web browser (for example, an HTTP (Hypertext Transfer Protocol) request). Web pages include content such as news articles. Further, the information processing device 100 may be an application server that provides contents to the first terminal device 10 in response to a request from an application program.

2 and 3 are diagrams showing an example of the content displayed on the screen of the first terminal device 10. The example of FIG. 2 represents a web page in which the content is posted in a list format. Such web pages display representative images contained in each content, snippets (summary) of each content, hyperlinks LK to other web pages containing detailed information of each content, and the like. To. For example, in the web page illustrated in FIG. 2, when the hyperlink LK ₁ of the content CT ₁ in the uppermost row is selected, the screen transitions to the web page illustrated in FIG. On this web page, for example, along with the content CT ₁ selected by the user, images, snippets, hyperlinks LK, and the like of the content CT ₁ and other contents related to the time series are displayed. Thereby, for example, when the user browses a certain content CT ₁ , the topic handled by the content CT ₁ is dealt with at a past time point before the browsing time or at a future time point after the browsing time. Content CT _X , CT _Y , CT _Z can be provided to the user. The details of the method of providing such contents will be described below.

[Information processing device configuration]
FIG. 4 is a diagram showing an example of the configuration of the information processing apparatus 100 in the embodiment. As shown in the figure, the information processing apparatus 100 includes, for example, a communication unit 102, a control unit 110, and a storage unit 130.

The communication unit 102 includes, for example, a communication interface such as a NIC. The communication unit 102 communicates with the first terminal device 10, the second terminal device 20, and the like via the network NW. For example, the communication unit 102 may communicate with the first terminal device 10 and receive an HTTP request or the like. Further, for example, the communication unit 102 may communicate with the second terminal device 20 and receive the content from the second terminal device 20. When the communication unit 102 receives the content, the communication unit 102 stores the received content in the storage unit 130 as the content data 132 described later.

The control unit 110 includes, for example, a preprocessing unit 112, an extraction unit 114, a selection unit 116, a classification unit 118, and a provision unit 120. These components are realized, for example, by a processor (or processor circuit) such as a CPU (Central Processing Unit) or a GPU (Graphics Processing Unit) executing a program (software) stored in the storage unit 130. .. Further, some or all of the components of the control unit 110 are hardware (circuit unit: circuitry) such as LSI (Large Scale Integration), ASIC (Application Specific Integrated Circuit), FPGA (Field-Programmable Gate Array), and the like. ), Or it may be realized by the cooperation of software and hardware. Further, the program referenced by the processor may be stored in the storage unit 130 in advance, or is stored in a removable storage medium such as a DVD or a CD-ROM, and the storage medium is the drive of the information processing apparatus 100. It may be installed in the storage unit 130 from the storage medium by being attached to the device.

The storage unit 130 is realized by, for example, a storage device such as an HDD (Hard Disc Drive), a flash memory, an EEPROM (Electrically Erasable Programmable Read Only Memory), a ROM (Read Only Memory), and a RAM (Random Access Memory). In addition to various programs such as firmware and application programs, the storage unit 130 stores content data 132, user log data 134, keyword dictionary 136, and the like.

The content data 132 is, for example, data including a plurality of articles (for example, tens of thousands to hundreds of thousands of articles) as contents. FIG. 5 is a diagram showing an example of content data 132. When the content is a document such as a news article, the content data 132 is vectorized with the submission time of the article and the article for the article ID that identifies each article, as shown in the illustrated example. The data may be associated with an article vector (content vector) or the like.

The submission time may be, for example, the time when the article is transmitted from the second terminal device 20 to the information processing device 100, or the time when the article is received by the information processing device 100.

The article vector is one of the content vectors when the content is an article, and is a vector generated from the article by using a method called distributed representation such as word2vec or GloVe. Distributed representation is a method of vectorizing a word or phrase based on the co-occurrence of the word or phrase and the word or phrase that appears before or after the word or phrase, for example, a plurality of prepared words. Based on a corpus that includes words and phrases, it is a method to obtain the appearance probability of words that appear before and after one word of interest in a certain context, and to generate a multidimensional vector with the appearance probability as an element value. Specifically, when the article vector is V, it can be expressed as V = [e1, e2, e3, ...]. The process of generating the article vector from the article may be performed by the preprocessing unit 112, or may be performed by another device other than the information processing device 100.

The user log data 134 is data including the action history of a plurality of users. FIG. 6 is a diagram showing an example of the user log data 134. As shown in the illustrated example, in the user log data 134, the URL (Uniform Resource Locator) of the web page on which the article viewed by the user is posted, the title of the article, and the like are used for the user ID that identifies each user. It is the data associated with the action history.

The keyword dictionary 136 is a dictionary used when a process of extracting a feature word from the content (hereinafter referred to as a feature word extraction process) is performed by using a morphological analysis, a chunking process, or the like. Morphological analysis is a method of dividing a document into morphemes and analyzing them. Chunking processing is a processing in which words divided by morphological analysis etc. are reconnected into a semantic group as necessary in language processing. It is a process of connecting multiple words that are unique phrases. A character string representing a characteristic word is registered in the keyword dictionary 136, and the character string may be a noun such as an organization name, a person name, a place name, or a unique object name. The characteristic word (noun) may be one word or a phrase consisting of a plurality of words.

[Processing flow]
Hereinafter, the flow of a series of processes of the control unit 110 in the embodiment will be described according to a flowchart. FIG. 7 is a flowchart showing a flow of a series of processes of the control unit 110 in the embodiment. The processing of this flowchart may be repeated, for example, at a predetermined cycle.

First, the preprocessing unit 112 determines the content of interest from the plurality of contents included in the content data 132 (S100). For example, the preprocessing unit 112 may determine the article having the latest submission time (new article submitted most recently) among the plurality of articles included in the content data 132 as the content of interest. Further, when the preprocessing unit 112 receives an HTTP request or the like from the first terminal device 10 by the communication unit 102, the preprocessing unit 112 refers to the user log data 134 and makes a request from among a plurality of contents included in the content data 132. The content viewed in the past by the user who uses the first terminal device 10 which is the transmission source of the above (for example, the content which the user last viewed) may be specified, and the specified content may be determined as the content of interest.

Next, the preprocessing unit 112 derives the similarity between the determined content of interest and each of the plurality of contents included in the content data 132 (S102). For example, the preprocessing unit 112 derives the cosine similarity between the content vector corresponding to the content of interest and each content vector corresponding to each of the plurality of contents as the similarity between the contents.

Next, the extraction unit 114 selects and selects the content of the upper predetermined number N having a high degree of similarity with the content of interest from the plurality of contents whose similarity with the content of interest is derived by the preprocessing unit 112. Keywords are extracted from a set of N contents (S104). For example, the extraction unit 114 performs a feature word extraction process on each of the content of interest and each of the N contents, and extracts the character string (noun) registered in the keyword dictionary 136 as the first keyword. Further, the extraction unit 114 performs a unique expression extraction process for each of the content of interest and each of the N contents, and the organization name, the person name, the place name, the date expression, the time expression, the amount expression, the ratio expression, and the unique object name. A character string classified into a predetermined named entity class such as is extracted as a second keyword.

Next, the selection unit 116 is related to the content of interest and the time series from among the plurality of contents included in the content data 132, subject to the keywords (first keyword and second keyword) extracted by the extraction unit 114. Select the content to be used (hereinafter referred to as related content) (S106). The related content is, for example, content that deals with the same event as the event of interest, and if the content of interest deals with an event at a certain first time, the second content is before or after the first time. Content that deals with events at time. Specifically, if the content of interest is a news article about "Typhoon No. 3" that occurred on August 1, 2020, the news article about "Typhoon No. 3" as of August 2, 2020 or later will be Selected as related content.

For example, the selection unit 116 is the first one or more of the plurality of first keywords extracted from both the content of interest and the N contents by the extraction unit 114 performing the feature word extraction process, and the accuracy is equal to or higher than the threshold value. Content containing a predetermined percentage or more of keywords is selected as provisional related content from a plurality of contents. The accuracy is an index value indicating the degree of certainty that the first keyword is a keyword that simply represents the event of the content of interest, and is, for example, a keyword that appears in the content of interest and N contents. It may be expressed by frequency (number of appearances). In this case, the accuracy increases as the first keyword appears more in the content of interest and the N contents.

When the selection unit 116 selects the content including the first keyword whose accuracy is equal to or higher than the threshold value as one or more provisional related contents from the plurality of contents, the selection unit 116 extracts from one or more provisional related contents. The content including the second keyword extracted from the content of interest is selected as the related content by the 114 performing the named entity extraction process.

FIG. 8 is a diagram for explaining a selection process of related contents. In the figure, CT ₁ represents the content of interest, and each of CT ₂ to CT ₄ represents the content to be selected for the related content (a plurality of contents included in the content data 132). For example, from the content CT ₁ of interest, the character strings "Tokyo", "fireworks", "Minato Ward", "festival", and "○○ park" are extracted as the first keyword whose accuracy is equal to or higher than the threshold value, and the second As keywords, the character string "Tokyo" classified into the class "place name" and the character string "fireworks display" classified into the class "unique object name" are extracted. From the content CT ₂ , the character strings "Tokyo", "fireworks", "Minato Ward", "festival", and "○○ park" are extracted as the first keyword, and the class "place name" is extracted as the second keyword. The character string "Tokyo" classified in "Tokyo" and the character string "Fireworks display" classified in the class "unique object name" are extracted. From the content CT ₃ , the character strings "Tokyo", "fireworks", "Shinjuku Ward", "○○ stadium", and "watching baseball" are extracted as the first keyword, and the second keyword is "place name". The character string "Tokyo" classified into the class and the character string "Fireworks display" classified into the class "unique object name" are extracted. From the content CT ₄ , the character strings "Kanagawa", "Fireworks", "Yokohama City", "Festival", and "△△ Park" are extracted as the first keyword, and the class "Place Name" is extracted as the second keyword. The character string "Kanagawa Prefecture" classified into "Kanagawa Prefecture" and the character string "Fireworks display" classified into the class "Unique object name" are extracted.

For example, when the predetermined ratio is 80 [%], the content CT ₂ includes the first keyword whose accuracy is equal to or higher than the threshold value at a ratio of 100 [%] and includes all the second keywords, so that the selection unit 116 Selects the content CT ₂ as the related content of the content CT ₁ of interest. On the other hand, although the content CT ₃ includes all the second keywords, the content CT 3 contains the first keyword whose accuracy is equal to or higher than the threshold value at a ratio of 40 [%], which is less than a predetermined ratio. CT ₃ is not selected as the related content of the content CT ₁ of interest. Also. Since the content CT ₄ does not include a part of the second keyword, contains the first keyword having an accuracy of the threshold value or more at a ratio of 40 [%], and is less than a predetermined ratio, the selection unit 116 selects the content. CT ₄ is not selected as the related content of the content CT ₁ of interest.

When the selection unit 116 selects the related content from the plurality of contents included in the content data 132, the selection unit 116 selects the content whose similarity with the content of interest (for example, the cosine similarity) is less than the threshold value of the related content. It may be excluded from the selection.

Returning to the description of FIG. 7, the classification unit 118 then determines whether or not the number of related contents selected by the selection unit 116 exceeds a predetermined number (S108). The predetermined number is, for example, the number of related contents that can be displayed in the remaining area of the web page when the content of interest is displayed on the web page. A predetermined number may be treated as a hyperparameter that can be arbitrarily determined by a web page designer or the like.

When the classification unit 118 determines that the number of related contents is equal to or less than a predetermined number, the providing unit 120 controls the communication unit 102 to provide (transmit) the related contents to the first terminal device 10 (S110). .. For example, the providing unit 120 may provide the related content to the first terminal device 10 by posting the related content on a web page as illustrated in FIG.

On the other hand, when the classification unit 118 determines that the number of related contents exceeds a predetermined number, a plurality of related contents are changed based on changes in the relative contents of the content of interest and the plurality of related contents related to the time series. Each of the related contents is classified into one of a plurality of groups (clusters) (S112).

9 to 15 are diagrams for explaining a method of grouping related contents. In the figure, V1 to V9 represent the content vector of the related content, the vertical axis represents the content of the related content, and the horizontal axis represents the time when the content was uploaded (for example, the submission time). The content of the content is, for example, a plurality of elements included in the content vector. In the illustrated example, the content is represented in one dimension for the sake of brevity, but it may be multidimensional.

First, the classification unit 118 derives the center of gravity _GR of V9 from the content vector V1 distributed in a multidimensional space with a spread. In the example of FIG. 9, the content vectors V1 to V3, V4 to V6, and V7 to V9 are distributed on the same axis, respectively, and when viewed from the content vectors V4 to V6, V1 to V3 and V7 to V9. Are distributed equidistantly from each other. Therefore, as illustrated in FIG. 10, the center of gravity _GR appears on the axis passing through V4 to V6.

When the center of gravity _GR is derived, the classification unit 118 sets a boundary line between the content vectors, which is the boundary line of the group, in order to determine where the groups are separated from the content vectors V1 to V9, which are discretely distributed in the time axis direction. Set to. As shown in FIG. 11, for example, the boundary line that becomes the boundary of the group is time t _1-2 , t _2-3 , t _3-4 , t _4-5 , t _5-6 , t _6-7 , t. It is provided at one or more of _7-8 and _t8-9 . Time t _1-2 is the time between the upload time t ₁ corresponding to the content vector V1 and the upload time t ₂ corresponding to the content vector V2, and time t _2-3 is the upload corresponding to the content vector V2. It is a time between the time t ₂ and the upload time t ₃ corresponding to the content vector V3. The same applies to other times t _3-4 , t _4-5 , t _5-6 , t _6-7 , t _7-8 , and t _8-9 . Since the vertical axis is one-dimensional for convenience, the boundary of the group is described as a one-dimensional boundary line. However, as described above, when the vertical axis is multidimensional, the boundary of the group is also a multidimensional space ( For example, it may be a plane).

For example, the classification unit 118 has a plurality of time points t _1-2 , t _2-3 , t _3-4 , t _4-5 , t _5-6 , t _6-7 , t _7-8 , which are candidates for the boundary line. , T _8-9 , by setting the boundary line at the time when the error when approximating the content vector in the group by the center of gravity is the smallest, a plurality of multidimensional spaces in which the content vectors V1 to V9 are distributed can be created. Separate into groups.

In the example of FIG. 12, the time t _1-2 is the boundary line, and the multidimensional space in which the content vectors V1 to V9 are distributed is the first group including the contents uploaded before the time t _1-2 and the time. It is separated into a second group that includes content uploaded after _t1-2 . In this case, the content vector V1 is classified into the first group, and the content vectors V2 to V9 are classified into the second group.

When the content vector V1 to V9 are classified into either the first group or the second group, the classification unit 118 derives the center of gravity of the content vector for each group, and the error of the content vector with respect to the center of gravity for each group (for example, the minimum). Squared error) is derived.

In the example of FIG. 12, since the center of gravity G1 of the first group is at the same position as the content vector V1, the error of the content vector V1 with respect to the center of gravity G1 is zero. On the other hand, since the center of gravity G2 of the second group is the center of gravity of the content vectors V2 to V9, an error occurs in each content vector from the center of gravity G2.

Next, as in the example of FIGS. 13 to 15, the classification unit 118 sets a boundary line at a time different from the time t _1-2 , separates the multidimensional space into a plurality of groups, and divides the multidimensional space into a plurality of groups, and the center of gravity of each group. Iterates to derive the error of the content vector for. As illustrated in FIG. 14, when the time t _3-4 is set as the boundary line, the error between the first group and the second group is the smallest. Therefore, the classification unit 118 uses the time t _3-4 as the boundary line. Set and search the boundary line after time t _3-4 . In this way, the classification unit 118 searches for the boundary line where the error when approximating the content vector by the center of gravity is the smallest, and separates the multidimensional space in which the content vectors V1 to V9 are distributed into a plurality of groups.

Returning to the description of FIG. 7, when a plurality of related contents are classified into a plurality of groups by the classification unit 118, the providing unit 120 selects one related content (hereinafter referred to as representative related content) as a representative from each group. Select (S114).

FIG. 16 is a diagram for explaining a method of selecting representative related contents. For example, when the classification unit 118 sets a boundary line at each of the time t _3-4 and the time t _6-7 , and separates the multidimensional space in which the content vectors V1 to V9 are distributed into three groups, the content vector 118 is provided. The unit 120 selects representative-related content by selecting one content vector from the content vectors classified into each group. In the illustrated example, the content vector V5 classified into the second group is used as the content vector of the content of interest. For example, when the provision unit 120 emphasizes the connection with the content of interest, as shown in the illustrated example, in each group, the related content that is closest in time to the second group including the content of interest is the representative related content. May be selected as. In this case, as shown in the illustrated example, the related content corresponding to the content vector V3 is selected as the representative related content from the first group, and the related content corresponding to the content vector V7 is selected as the representative related content from the third group. Is selected as.

Further, when the first report is emphasized in a news article or the like, the providing unit 120 corresponds to the related content having the earliest time in each group (in the example of FIG. 16, the related content corresponding to the content vector V1 or the content vector V7). Related content) may be selected as the representative related content.

In the illustrated example, the boundary line is set at two places, but the boundary line is not limited to this. For example, when the number of related contents to be displayed together with the content of interest is two, a boundary line may be set at a certain time in order to set the number of groups to two. Further, when the number of related contents to be displayed together with the content of interest is four or more, a boundary line may be set at three or more times in order to set the number of groups to four or more.

Next, the providing unit 120 provides the representative-related content selected from each group to the first terminal device 10 as the process of S110. This ends the processing of this flowchart.

According to the embodiment described above, the condition is that the extraction unit 114 that extracts keywords from the content of the upper predetermined number N having a high degree of similarity to a certain content of interest among the plurality of contents, and the keywords extracted by the extraction unit 114. In order to provide the selection unit 116 for selecting the content of interest and one or more related contents related to the time series from the plurality of contents, the content of interest and the related contents related to the time series are provided to the user. Can be done.

For example, if the content is a news article, a new article about the event may be submitted each time the event changes within one related event. When new articles that are submitted one after another are provided to the user as content in this way, the user is more likely to follow the article of interest than the article related to the genre (topic) of the article of interest. In some cases, the submitted article or the article corresponding to the first report when the article of interest is a follow-up article is more interesting.

For example, when providing an article related to a genre to a user as content for an article of interest, whether or not to obtain the similarity between the articles by TF-IDF or the like and provide the article to the user according to the similarity. It can be considered to decide. In this case, although the genres of the articles are similar, the articles may not be temporally related to each other. For example, if there are two articles reporting "heavy rain", one article reports "heavy rain" that occurred in "Tokyo" and the other article reports "heavy rain" that occurred in "Kyushu". In the case of articles to be reported, although the genres of the articles are common to each other due to the natural disaster of "heavy rain", it is not appropriate to associate one article as a follow-up article of the other article because the place of occurrence is different from each other.

On the other hand, in this embodiment, the main proper noun is used as a keyword rather than the closeness of the topics of the articles to each other, and other contents are related to the content of interest according to the commonality of the keywords. In order to determine whether or not to do so, the content of interest and the content related to the time series can be appropriately selected.

Further, according to the above-described embodiment, the classification unit that classifies each of the plurality of related contents into one of the plurality of groups based on the relative change in the contents of the plurality of related contents selected by the selection unit 116. By further providing 118, among a plurality of related contents discretely distributed in the time axis direction, related contents whose contents change little even if the time changes can be grouped into one group. As a result, representative related contents can be left from each group, and other related contents can be thinned out.

In general, in the hierarchical clustering of data, all the data are out of order, and clustering is performed from arbitrary data. On the other hand, in the present embodiment, since a plurality of related contents discretely distributed in the time direction are clustered in the time direction, the related contents can be thinned out while leaving the change in the contents. As a result, it is possible to prevent the user from being provided with content having duplicate contents.

<Modified example of the embodiment>
Hereinafter, a modified example of the above-described embodiment will be described. In the above-described embodiment, the classification unit 118 sets each of the plurality of related contents into a plurality of groups based on the change in the relative contents of the content of interest and the plurality of related contents related to the time series. Although it was explained as being classified as a content, it is not limited to this.

For example, the classification unit 118 may classify a plurality of related contents into groups so that they are evenly spaced in time, or a plurality of related contents so that the number of related contents contained in each group is equal. Content may be grouped.

Further, in the above-described embodiment, the preprocessing unit 112 derives the similarity between each of the plurality of contents and the content of interest, and the extraction unit 114 derives the similarity with the content of interest by the preprocessing unit 112. From the plurality of contents, the top predetermined number N contents having a high degree of similarity to the content of interest are selected, keywords are extracted from the set of the selected N contents, and the selection unit 116 is extracted by the extraction unit 114. After selecting the content of interest and the related content related to the time series from the plurality of contents included in the content data 132 on the condition of the specified keyword, the classification unit 118 classifies the plurality of related contents into groups. I explained it as something to do, but it is not limited to this.

For example, the classification unit 118 treats the content of the upper predetermined number X having a high degree of similarity with the content of interest as the related content from among the plurality of contents whose similarity with the content of interest is derived by the preprocessing unit 112. Related content may be grouped. The predetermined number X may be the same as or different from the predetermined number N described above.

Further, when the genre of the content to be provided to the user is determined in advance, the classification unit 118 may classify only the content corresponding to the genre into a group.

<Hardware configuration>
The information processing apparatus 100 of the above-described embodiment is realized by, for example, a hardware configuration as shown in FIG. FIG. 17 is a diagram showing an example of the hardware configuration of the information processing apparatus 100 of the embodiment.

The information processing device 100 includes NIC100-1, CPU100-2, RAM100-3, ROM100-4, a secondary storage device 100-5 such as a flash memory or an HDD, and a drive device 100-6 on an internal bus or a dedicated communication line. It is configured to be interconnected by. A portable storage medium such as an optical disk is mounted on the drive device 100-6. A program stored in a portable storage medium mounted on the secondary storage device 100-5 or the drive device 100-6 is expanded in the RAM 100-3 by a DMA controller (not shown) or the like, and executed by the CPU 100-2. As a result, the control unit 110 is realized. The program referred to by the control unit 110 may be downloaded from another device via the network NW.

Although the embodiments for carrying out the present invention have been described above using the embodiments, the present invention is not limited to these embodiments, and various modifications and substitutions are made without departing from the gist of the present invention. Can be added.

1 ... Information processing system, 10 ... First terminal device, 20 ... Second terminal device, 100 ... Information processing device, 102 ... Communication unit, 110 ... Control unit, 112 ... Preprocessing unit, 114 ... Extraction unit, 116 ... Selection Department, 118 ... Classification department, 120 ... Providing department, 130 ... Storage department

Claims (9)

An extraction unit that extracts keywords from a predetermined number of high-ranking contents that have a high degree of similarity to a certain content of interest among multiple contents.
A selection unit for selecting one or more related contents related to the content of interest and the time series from the plurality of contents based on the keyword extracted by the extraction unit is provided.
The extraction unit extracts a first keyword representing a feature word and a second keyword representing a named entity from each of the content of interest and the upper predetermined number of contents.
The selection unit has the first keyword and the second keyword in common with the content of interest in the upper predetermined number of contents, and is before or after the first time associated with the content of interest. Select the content associated with the second time as the related content.
Information processing equipment. The selection unit selects, as the related content, content containing the second keyword and containing the plurality of the first keywords in a predetermined ratio or more from the plurality of contents.
The information processing apparatus according to claim 1 . A preprocessing unit for excluding content having a similarity less than a threshold value with the content of interest from the plurality of contents is further provided.
The information processing apparatus according to claim 1 or 2 . When a plurality of related contents are selected by the selection unit, the plurality of related contents are classified into a plurality of groups based on the relative change of the content vector obtained by vectorizing each of the plurality of related contents. With more parts,
The information processing apparatus according to any one of claims 1 to 3 . The classification unit
The content vectors corresponding to each of the plurality of related contents are discretely distributed on the feature space.
The centers of gravity of the plurality of content vectors distributed discretely on the feature space are derived, and the centers of gravity are derived.
Based on the derived center of gravity, the plurality of related contents are classified into a plurality of groups.
The information processing apparatus according to claim 4. The classification unit
The center of gravity of the content vector corresponding to each of the related contents included in each group is derived for each group.
The plurality of related contents are classified into a plurality of groups so that the error of the center of gravity derived for each group is minimized.
The information processing apparatus according to claim 5. A communication unit that communicates with a terminal device that can be used by the user,
Further includes a providing unit that selects at least one related content from each of the plurality of groups, controls the communication unit, and provides the related content selected from each of the groups to the terminal device. ,
The information processing apparatus according to any one of claims 4 to 6 . The computer
Of multiple contents, keywords are extracted from the top predetermined number of contents that have a high degree of similarity to a certain content of interest.
Based on the extracted keywords, one or more related contents related to the content of interest and the time series are selected from the plurality of contents.
A first keyword representing a feature word and a second keyword representing a named entity are extracted from each of the content of interest and the upper predetermined number of contents.
Among the upper predetermined number of contents, the first keyword and the second keyword are common to the attention content, and the second time before or after the first time associated with the attention content corresponds to. Select the attached content as the related content,
Information processing method. On the computer
Of multiple contents, the process of extracting keywords from the top predetermined number of contents that have a high degree of similarity to a certain content of interest, and
A process of selecting one or more related contents related to the content of interest and the time series from the plurality of contents based on the extracted keywords.
A process of extracting a first keyword representing a feature word and a second keyword representing a named entity from each of the content of interest and the upper predetermined number of contents.
Among the upper predetermined number of contents, the first keyword and the second keyword are common to the attention content, and the second time before or after the first time associated with the attention content corresponds to. The process of selecting the attached content as the related content, and
A program to execute.

Images