Relationship extraction device and program

Abstract

To automatically extract a hierarchical relationship regarding, for example, educational content.SOLUTION: An educational hierarchical relationship extraction unit takes words included in a text of curriculum guidelines and defined in concept dictionary data, as search object words in the highest hierarchy and performs processing of, with respect to each search object word of the search object words, extracting a hierarchical relationship indicating that the search object word is higher and the most important word included in an educational content text and having the highest degree of importance in the educational content text is lower, and adding the most important word as one of search object words if a condition a) that the most important word is not acquired as a search object word, a condition b) that the degree of importance of the most important word is higher than the degree of importance of the search object word, a condition c) that the degree of importance of the most important word is equal to or higher than a prescribed threshold, and a condition d) that the most important word is a word defined in the concept dictionary data are satisfied.SELECTED DRAWING: Figure 1

Description

Technical Field

[0006] , , , , ,

[0001] The present invention relates to a relation extraction device and a program.

Background Art

[0002] In the field of education, learning services utilizing Internet technology have attracted attention. Furthermore, not only learning services based on a uniform learning order but also learning services tailored to the understanding and progress of learners have been demanded. Therefore, various methods for automatically presenting content (learning content) to users (learners) and instructors have been studied.

[0003] For example, Patent Document 1 discloses a technique for defining a learning area based on the curriculum guidelines and recommending teaching materials to users based on the grade, subject, or textbook unit within that area.

[0004] Note that the information on the curriculum guidelines is publicly available and can be obtained via communication. As an example, the Science Curriculum Guidelines for Lower Secondary Schools (announced in March 2017) can be accessed at the URL [https: / / jp-cos.github.io / LowerSecondary / 2017 / 理科].

[0005] Also, for example, Non-Patent Document 1 discloses a method of describing the relationship between learning content and other (non-learning) video content in the data format of RDF (Resource Description Framework) that supports the Semantic Web, and presenting the relationship between the two in an explicit manner.

Prior Art Documents

Patent Documents

[0006]

Patent Document 1

Non-Patent Documents

[0007] [Non-Patent Document 1] Makoto Urakawa and Hiroshi Fujisawa, "Verification of the Use of Video Content Linked to Educational Applications Using Structured Data," ITE Technical Report, Vol. 42, No. 11, BCT2018-38, pp. 1-4 Summary of the Invention [Problem to be solved by the invention]

[0008] In recent years, the importance of "personalized learning," which provides opportunities for students to engage in learning activities and tasks tailored to their individual interests, has been emphasized. In "personalized learning," if the concepts to be learned, as outlined in the curriculum guidelines, can be identified, the learning materials can be selected in accordance with the student's interests. However, the curriculum guidelines do not specify specific concepts (materials). Therefore, the methods described in Patent Document 1 and Non-Patent Document 1 are unable to clearly indicate specific concepts that will attract students' interest, making them difficult to apply to personalized learning.

[0009] Learning can be done effectively by learning both the core concept to be learned and the more specific concepts that serve as examples of that concept. However, although the relationship between the core concept and the more specific concepts (the relationship between the two is like that between a higher-level concept and a lower-level concept) is important as a guideline for learning, it is not defined in the curriculum guidelines. As a result, in order to clearly indicate the specific concepts that will interest users, the only option is to search for and think about them manually, which is an extremely time-consuming process.

[0010] The present invention has been made in consideration of the above circumstances, and aims to provide a relationship extraction device and program that can automatically extract relationships between higher and lower levels, for example, in relation to educational content. [Means for solving the problem]

[0011] [1] In order to solve the above problem, a relationship extraction device according to one aspect of the present invention includes a learning hierarchical relationship extraction unit that extracts, as a hierarchy relationship associated with a learning course code, a hierarchy relationship between a search target word and other search target words, based on a learning course text associated with a learning course code, concept dictionary data that is data defining a hierarchy relationship between a word and other words, and an educational content text that is a text of educational content associated with the learning course code, and the learning hierarchical relationship extraction unit extracts, for each of the search target words, a hierarchy relationship between a word included in the learning course text and defined in the concept dictionary data, the highest-level search target words. If a) there is no other search target word in the educational content text that is more important than the search target word, and the most important word, which is the word with the highest importance in the educational content text, has not been acquired as a search target word, and b) the importance of the most important word is higher than the importance of the search target word, and c) the importance of the most important word is equal to or greater than a predetermined threshold, and d) the most important word is a word defined in the concept dictionary data, then a process is performed to extract a hierarchical / hierarchical relationship associated with the curriculum code, in which the search target word is higher and the most important word is lower, and to add the most important word as one of the search target words.

[0012] [2] Furthermore, one aspect of the present invention is that in the relationship extraction device of [1] above, the learning hierarchical relationship extraction unit further extracts, when two words contained in one of the educational content texts are defined as having a superordinate-subordinate relationship in the concept dictionary data, the superordinate-subordinate relationship as a superordinate-subordinate relationship associated with the curriculum code.

[0013] [3] Furthermore, in one aspect of the present invention, in the relationship extraction device of [1] or [2] above, the concept dictionary data further defines an alias relationship in which a certain word is an alias of another word, and the learning hierarchical relationship extraction unit, when extracting the hierarchical-subordinate relationship associated with the curriculum code in which the search target word is higher and the most important word is lower, further refers to the concept dictionary data to extract the hierarchical-subordinate relationship associated with the curriculum code in which the word equivalent to the alias of the most important word is the lower word, in which the search target word is higher and the word equivalent to the alias is lower, and performs processing to add the word equivalent to the alias as one of the search target words.

[0014] [4] Furthermore, one aspect of the present invention is that the relationship extraction device of any of [1] to [3] above further comprises a content systematization unit that performs processing to associate content corresponding to the educational content text that contains the most important word when the learning hierarchical relationship extraction unit extracts the superordinate-subordinate relationship associated with the curriculum code with the subordinate word in the superordinate-subordinate relationship.

[0015] [5] Furthermore, one aspect of the present invention is that the relationship extraction device according to any one of [1] to [4] above further comprises a visualization unit that graphically visualizes and displays the set of superordinate and subordinate relationships extracted by the learning hierarchy relationship extraction unit in association with the curriculum code.

[0016] [6] Furthermore, one aspect of the present invention includes a learning hierarchical relationship extraction unit that extracts, as the hierarchy relationship associated with the learning course code, a hierarchy relationship between a search target word and other search target words, based on a learning course text associated with a learning course code, concept dictionary data that is data defining the hierarchy relationship between a word and other words, and an educational content text that is the text of educational content associated with the learning course code, and the learning hierarchical relationship extraction unit extracts, for each search target word, a hierarchy relationship between a word that is included in the learning course text and that is defined in the concept dictionary data, and the highest hierarchy relationship is the search target word. a) there are no other search target words in the educational content text that are more important than the search target word, and the most important word, which is the word with the highest importance in the educational content text, has not yet been acquired as a search target word; and b) the importance of the most important word is higher than the importance of the search target word; c) the importance of the most important word is equal to or greater than a predetermined threshold; and d) the most important word is a word defined in the concept dictionary data, then the program causes a computer to function as a relationship extraction device that extracts a hierarchical / hierarchical relationship associated with the curriculum code, in which the search target word is higher and the most important word is lower, and performs a process of adding the most important word as one of the search target words.

[0017] [Reference Aspect 1] Also, a relationship extraction method according to the reference aspect is as follows: That is, the relationship extraction device includes a learning hierarchical relationship extraction unit. As the relationship extraction method, the learning hierarchical relationship extraction unit extracts, as a hierarchy relationship associated with a curriculum guideline code, a hierarchy relationship between a search target word and another search target word, based on a curriculum guideline text associated with the curriculum guideline code, concept dictionary data which is data defining a hierarchy relationship between a word and another word, and an educational content text which is a text of educational content associated with the curriculum guideline code. In addition, at this time, the learning hierarchy relationship extraction unit sets a word that is included in the curriculum guideline text and that is defined in the concept dictionary data as the top-level search target word, and for each of the search target words, if: a) the word is included in the educational content text, there is no other search target word in the educational content text that is more important than the search target word, and the most important word that is the word with the highest importance in the educational content text has not yet been obtained as a search target word, b) and the importance of the most important word is higher than the importance of the search target word, c) and the importance of the most important word is equal to or greater than a predetermined threshold, and d) and the most important word is a word defined in the concept dictionary data, then it extracts a hierarchy relationship associated with the curriculum guideline code in which the search target word is higher and the most important word is lower, and performs processing to add the most important word as one of the search target words.

[0018] [Reference Aspect 2] In one reference aspect, in the relationship extraction method of the above [Reference Aspect 1], the learning hierarchy relationship extraction unit further extracts, when two words contained in one of the educational content texts are defined as having a superordinate-subordinate relationship in the concept dictionary data, the superordinate-subordinate relationship as a superordinate-subordinate relationship associated with the curriculum code.

[0019] [Reference Mode 3] In one reference mode, in the relationship extraction method of the above [Reference Mode 1] or [Reference Mode 2], the concept dictionary data further defines an alias relationship in which a certain word is an alias for another word, and when the learning hierarchy relationship extraction unit extracts a superordinate-subordinate relationship associated with the curriculum code in which the search target word is a superior and the most important word is a subordinate, the learning hierarchy relationship extraction unit further refers to the concept dictionary data to extract a superordinate-subordinate relationship associated with the curriculum code in which the search target word is a superior and the word corresponding to the alias is a subordinate, with the word corresponding to the alias of the most important word being the subordinate word, and performs a process of adding the word corresponding to the alias as one of the search target words.

[0020] [Reference Aspect 4] In one reference aspect, in the relationship extraction method of any one of [Reference Aspect 1] to [Reference Aspect 3], the relationship extraction device includes a content systematization unit that associates content corresponding to the educational content text that contains the most important word when the learning hierarchical relationship extraction unit extracts the superordinate-subordinate relationship associated with the curriculum code with the subordinate word in the superordinate-subordinate relationship.

[0021] [Reference Aspect 5] In one reference aspect, in the relationship extraction method of any one of [Reference Aspect 1] to [Reference Aspect 4], the relationship extraction device includes a visualization unit that graphically visualizes and displays a set of superordinate and subordinate relationships that the learning hierarchy relationship extraction unit has extracted in association with the curriculum guideline code. [Effects of the Invention]

[0022] According to the present invention, the relationship extraction device can automatically extract the hierarchical relationships of words related to a specific curriculum code based on the curriculum text and the educational content text. [Brief explanation of the drawings]

[0023] [Figure 1] 1 is a block diagram showing a schematic functional configuration of a relationship extraction device according to an embodiment of the present invention. [Figure 2] 3 is a schematic diagram showing an example of the configuration of content data acquired by a content data acquisition unit according to the embodiment. FIG. [Figure 3] FIG. 2 is a schematic diagram showing a specific example of content data (educational content data) used in the embodiment. [Figure 4] FIG. 2 is a schematic diagram showing a specific example of curriculum guideline data used in the embodiment. [Figure 5] 10 is a schematic diagram showing an example of a hierarchy obtained from a plurality of contents associated with a specific curriculum code in the embodiment. FIG. [Figure 6] 10 is a schematic diagram showing an example of text data of educational content that is used as a basis when the learning hierarchical relationship extraction unit according to the embodiment extracts superordinate and subordinate relationships. FIG. [Figure 7] FIG. 7 is a schematic diagram showing a list of pairs of words extracted from text data such as that shown in FIG. 6 and their importance levels in the embodiment. [Figure 8] FIG. 2 is a schematic diagram showing an example of a higher-level / lower-level relationship associated with a curriculum guideline code in the embodiment. [Figure 9] 10 is a schematic diagram showing an example of a user interface generated by a visualization unit according to the embodiment. FIG. [Figure 10] 10 is a schematic diagram showing another example of a user interface generated by the visualization unit according to the embodiment. FIG. [Figure 11] 10 is a flowchart (1 / 2) showing the procedure of processing executed by the relationship extraction device according to the embodiment. [Figure 12] 10 is a flowchart (2 / 2) showing the procedure of the process executed by the relationship extraction device according to the embodiment. [Figure 13] FIG. 10 is a block diagram showing an example of the internal configuration of a relationship extraction device realized using a computer. DETAILED DESCRIPTION OF THE INVENTION

[0024] Next, an embodiment of the present invention will be described with reference to the drawings.

[0025] FIG. 1 is a block diagram showing a schematic functional configuration of a relationship extraction device according to this embodiment. As shown in the figure, the relationship extraction device 1 includes a content data acquisition unit 101, an external data acquisition unit 102, a learning hierarchy relationship extraction unit 103, a content systematization unit 104, and a visualization unit 105. Each of these functional units can be realized, for example, by a computer and a program. Each functional unit also has a storage unit as needed. The storage unit is, for example, a program variable or memory allocated by program execution. Non-volatile storage units such as a magnetic hard disk drive or a solid-state drive (SSD) may also be used as needed. At least some of the functions of each functional unit may also be realized as a dedicated electronic circuit rather than a program. The functions of each unit are described below.

[0026] The content data acquisition unit 101 acquires content data from an external source. The content data is metadata related to the content. The structure of the content data will be described later. The content handled by the relationship extraction device 1 includes educational content and general content. The educational content is content that is associated with a curriculum code. The general content is content that is not associated with a curriculum code.

[0027] The content data acquisition unit 101 transfers the content data to a different destination depending on whether the acquired content data is data related to educational content or data related to general content. Specifically, the content data acquisition unit 101 transfers content data of educational content associated with a curriculum guideline code to the learning hierarchy relation extraction unit 103. In addition, the content data acquisition unit 101 transfers content data of general content not associated with a curriculum guideline code to the content systematization unit 104.

[0028] The external data acquisition unit 102 acquires concept dictionary data and curriculum guideline data, and passes the concept dictionary data and curriculum guideline data to the learning hierarchy relation extraction unit 103.

[0029] Concept dictionary data is data that represents the relationships between general concepts. For example, with respect to concepts A and B, the concept dictionary data represents relationships such as "A is a type of B" or "A is an instance of B." In other words, concept dictionary data can represent the hierarchical relationship between two concepts A and B. On the other hand, curriculum guideline data is data that represents the curriculum guidelines associated with each curriculum guideline code. Curriculum guideline data is data that represents, for example, school types, subjects, etc.

[0030] The learning hierarchical relationship extraction unit 103 extracts the superordinate-subordinate relationships between concepts contained in the educational content based on the content data of the educational content, the concept dictionary data, and the curriculum guideline data. These superordinate-subordinate relationships are called learning hierarchical relationships. In other words, a learning hierarchical relationship is a hierarchical relationship between a certain word (superordinate concept) and a more specific word (subordinate concept) that can be an example of that word.

[0031] The learning hierarchy relation extraction unit 103 acquires content data from the content data acquisition unit 101. The learning hierarchy relation extraction unit 103 also acquires concept dictionary data and curriculum guideline data from the external data acquisition unit .

[0032] The content systematization unit 104 systematizes the general content acquired from the content data acquisition unit 101 in association with the hierarchical relationship extracted by the learning hierarchical relationship extraction unit 103. Specifically, the content systematization unit 104 links the content to a specific word in the hierarchical relationship extracted by the learning hierarchical relationship extraction unit 103. Here, the word to link the content may be the lowest-level word in the superordinate-subordinate relationship between words.

[0033] The visualization unit 105 has a function of visualizing the hierarchical relationship extracted by the learning hierarchical relationship extraction unit 103 and presenting it to the user. At this time, the visualization unit 105 may visualize the content associated with the hierarchical relationship within the hierarchical relationship. In other words, the visualization unit 105 generates a screen for graphical display in a user interface (UI).

[0034] Here, the technical features of each part constituting the relationship extraction device 1 will be further explained.

[0035] The learning hierarchical relationship extraction unit 103 extracts the superordinate and subordinate relationships between the search target word and other search target words as the superordinate and subordinate relationships associated with the course of study code based on the course of study text, concept dictionary data, and educational content text.

[0036] Here, the curriculum guideline text is a text associated with the curriculum guideline code. The curriculum guideline text may be, for example, an explanatory text about the curriculum guideline code. The concept dictionary data is data that defines the hierarchical and subordinate relationships between words and other words. The educational content text is a text of educational content associated with the curriculum guideline code.

[0037] The learning hierarchy relation extraction unit 103 sets the words included in the curriculum text and defined in the concept dictionary data as the top-level search target words, and for each of the search target words, a) A word contained in an educational content text, in which there is no other search target word in the educational content text that is more important than the search target word, and the most important word that is the most important word in the educational content text has not been acquired as a search target word; b) And the importance of the most important word is higher than the importance of the search target word; c) And the importance of the most important word is equal to or greater than a predetermined threshold, d) And the most important word is a word defined in the concept dictionary data, This may involve extracting a hierarchy associated with the curriculum code, in which the search target word is higher and the most important word is lower, and performing a process to add the most important word as one of the search target words.

[0038] Furthermore, the learning hierarchical relationship extraction unit 103 may further extract, when two words contained in one educational content text are defined as having a superordinate-subordinate relationship in the concept dictionary data, the superordinate-subordinate relationship as a superordinate-subordinate relationship associated with the curriculum code.

[0039] The learning hierarchical relationship extraction unit 103 may also acquire aliases of hyponyms. Here, the concept dictionary data may define an alias relationship in which a word is an alias of another word. When extracting a hyponym-subordinate relationship associated with the curriculum guideline code in which the search target word is hyponym and the most important word is hyponym, the learning hierarchical relationship extraction unit 103 may further refer to the concept dictionary data to extract a hyponym-subordinate relationship associated with the curriculum guideline code in which the search target word is hyponym and the word corresponding to the alias is hyponym, with the word corresponding to the alias being the hyponym, and add the word corresponding to the alias as one of the search target words. This allows aliases defined in the concept dictionary data to be acquired as hyponyms.

[0040] As a systematization method, the content systematization unit 104 may perform a process of associating content corresponding to the educational content text that contains the most important word when the learning hierarchical relationship extraction unit 103 extracts the superordinate-subordinate relationship associated with the curriculum code with the subordinate word in the superordinate-subordinate relationship.

[0041] The visualization unit 105 may be capable of graphically visualizing and displaying a set of superordinate and subordinate relationships that the learning hierarchical relationship extraction unit 103 has extracted in association with the curriculum guideline code.

[0042] 2 is a schematic diagram showing an example of the configuration of content data acquired by the content data acquisition unit 101. The content data includes data items such as a content ID, title, summary, stream URL, subtitle data, and whether or not a curriculum code is included. The content data can be described as text data using a format such as RDF (Resource Description Framework). The meaning of each data item is as follows:

[0043] The content ID is identification information for uniquely identifying the content.

[0044] The title is the title of the content.

[0045] The summary is explanatory text that gives an overview of the content.

[0046] A stream URL is a string of information that indicates the location of content (such as a video). URL stands for "Uniform Resource Locator." For example, a user can use this URL to obtain the content stream.

[0047] The subtitle data is the text of the subtitles included in the content.

[0048] The presence or absence of a curriculum code is information indicating whether the content is associated with a curriculum code. If the curriculum code is "present," the content data further has a curriculum code.

[0049] FIG. 3 is a schematic diagram showing a specific example of content data (content data of educational content). However, in the content data in the example of this figure, data items such as the content ID, stream URL, and subtitle data are omitted. In the example shown, the title of the content is "What is Finance?" The summary is text such as "Japan's first bank, the First National Bank. Understand the mechanisms of finance, including indirect finance and direct finance. In a building in Kabutocho, Nihonbashi, Chuo-ku, Tokyo... (omitted below)." The curriculum code is "8323233211300000."

[0050] FIG. 4 is a schematic diagram showing a specific example of curriculum guideline data. The illustrated example is information corresponding to one curriculum guideline code. As shown in the figure, curriculum guideline data has data items such as curriculum guideline code, school type, subject, field, subject, classification, and text. In the illustrated example, the curriculum guideline code is "8323233211300000." The school type is "junior high school." The subject is "social studies." The field, subject, and classification is civics. The text describes the objectives of the curriculum guideline code, such as "to understand the mechanisms and functions of modern production and finance."

[0051] Note that the curriculum code is the same for both the data illustrated in Figure 3 and the data illustrated in Figure 4: "8323233211300000." In other words, in the example shown here, the content data related to the educational content associated with the curriculum code defined in Figure 4 is the data shown in Figure 3.

[0052] Figure 5 is a schematic diagram showing examples of superordinate and subordinate relationships obtained from multiple content items associated with a specific curriculum code. As shown, superordinate and subordinate relationships can be expressed as pairs of superordinate words (concepts) and subordinate words (concepts). Each row in the diagram represents one superordinate and subordinate relationship. The first superordinate and subordinate relationship shown is "financial institution" (superordinate) - "bank" (subordinate). This indicates that a bank is a type of financial institution. The second superordinate and subordinate relationship is "bank" (superordinate) - "national bank" (subordinate). This indicates that a national bank is a type of bank. This is similar to the following, and the final superordinate and subordinate relationship shown is "company" (superordinate) - "beauty salon" (subordinate). This indicates that a beauty salon is a type of company. Note that, as shown in the example, superordinate and subordinate relationships can be multi-hierarchical.

[0053] Here, the details of the process by which the learning hierarchical relation extraction unit 103 extracts the hierarchical relations shown in FIG. 5 will be described.

[0054] The learning hierarchy relation extraction unit 103 first extracts nouns from the text data of the educational content acquired by the content data acquisition unit 101 and the text data of the curriculum guidelines acquired by the external data acquisition unit 102. At this time, the learning hierarchy relation extraction unit 103 may also extract compound nouns as a type of noun. Since it can be said that there are more specific concepts in compound nouns than in nouns that are not compound nouns, it is appropriate for the learning hierarchy relation extraction unit 103 to extract nouns that include compound nouns. Hereinafter, the nouns will be described as including compound nouns. The process of extracting nouns (including compound nouns) from text is realized, for example, by performing morphological analysis or the like.

[0055] The concept dictionary data includes information that indicates the hierarchical and hierarchical relationships between words (nouns (including compound nouns)).

[0056] Next, the learning hierarchical relationship extraction unit 103 references the concept dictionary data for each noun extracted from the text data of the content to acquire word pairs that have a hierarchical relationship within the same content. The learning hierarchical relationship extraction unit 103 holds information about the acquired word pairs and information identifying the source content from which the word pairs were acquired.

[0057] At this time, the learning hierarchical relation extraction unit 103 can use the premise that the contexts of concepts (words) included in the same content match.

[0058] Here, the constraint that concepts contained in the same content have the same context can be utilized. In other words, the learning hierarchical relationship extraction unit 103 may acquire the superordinate-subordinate relationship on the assumption that concepts contained in the same content have the same context. For example, the word (noun) written as "Thailand" may represent a country name or a type of fish. In other words, the concept dictionary data includes the superordinate-subordinate relationship of country (superordinate) → Thailand (subordinate) and the superordinate-subordinate relationship of fish (superordinate) → Thailand (subordinate). In other words, the learning hierarchical relationship extraction unit 103 acquires word pairs that have a superordinate-subordinate relationship within the same content as a hierarchical relationship. In other words, the country name "Thailand" and the fish species "Thailand" are written the same but have different meanings (homonyms). For example, the word "Thailand" that appears in a country-related context is more likely to co-occur with the word "country" than with the word "fish." That is, the learning hierarchical relationship extraction unit 103 can acquire only relationships that match the context by acquiring word pairs that have a hierarchical relationship within the same content.

[0059] Next, if the noun extracted from the text of the curriculum guideline data is a word that exists in the concept dictionary data, the learning hierarchical relation extraction unit 103 acquires the hierarchical relation by using the noun as a search target word for the top-level word candidate of the learning hierarchical relation. Note that if a hyponym of the top-level word candidate has already been acquired from the concept dictionary data, this hyponym is also used as a search target word.

[0060] Then, the learning hierarchical relation extraction unit 103 extracts hyponyms for each search target word based on the educational content that includes that search target word. Note that here, if the educational content that includes the search target word also includes other search target words, the educational content that has the highest importance for the target search target word is selected. The conditions for extracting hyponyms are as follows: a) The most important words in the educational content have not yet been acquired as learning hierarchy relationships in the target curriculum code, b) And the importance of the most important word is higher than the importance of the search target word, c) And the importance of the most important word is equal to or greater than a predetermined threshold, d) And the most important word is a word that exists in the concept dictionary data, In this case, the learning hierarchical relationship extraction unit 103 estimates that the most important word is a more specific word than the search target word, and adds the most important word as a subordinate word in the learning hierarchical relationship. The learning hierarchical relationship extraction unit 103 also adds the most important word to the search target word.

[0061] At this time, the learning hierarchical relation extraction unit 103 also holds information on the educational content from which the word pairs of the hypernym and the hyponym are obtained (source content information).

[0062] Furthermore, if a hyponym of the added word has already been acquired from the concept dictionary data, the learning hierarchical relationship extraction unit 103 also adds the hyponym to the learning hierarchical relationship. That is, the hyponym-hyperordinate relationship between the added word and its hyponym is extracted. The learning hierarchical relationship extraction unit 103 also adds the hyponym added here to the search target words.

[0063] If information on an alternative name for a certain hyponym exists in the concept dictionary data, the learning hierarchical relationship extraction unit 103 also adds the alternative name as a hyponym. The learning hierarchical relationship extraction unit 103 also adds the alternative name to the search target words.

[0064] The learning hierarchical relation extraction unit 103 repeats the above process until there are no more unprocessed search target words.

[0065] The importance of a word may be determined, for example, by using the TF-IDF (Term Frequency-Inverse Document Frequency) value of the text for each subject. However, the method of calculating the importance is not limited to this. When using the TF-IDF value as the importance of a word, the threshold value in the above process may be set to 0.3, for example. However, other values ​​may also be used as the threshold value.

[0066] The learning hierarchy relation extraction unit 103 performs the above procedure for all educational contents associated with the curriculum code. The learning hierarchy relation acquired by the learning hierarchy relation extraction unit 103 can be structured and represented in a format such as RDF (Resource Description Framework). However, the method of representing the learning hierarchy relation is not limited to this and is arbitrary.

[0067] FIG. 6 shows an example of text data of educational content that is the source when the learning hierarchical relationship extraction unit 103 extracts hierarchical and subordinate relationships. This text data of educational content is the same as the data shown in FIG. 3. However, in FIG. 6, the words "finance" and "bank" are underlined. The word "finance" is a search target word. In other words, the word "finance" is a hypernym when the learning hierarchical relationship extraction unit 103 extracts hierarchical and subordinate relationships. Furthermore, the word "bank" is its hyponym.

[0068] FIG. 7 is a schematic diagram showing a list of pairs of words extracted from the text data shown in FIG. 6 and their importance (here, TF-IDF values ​​are used). In the example shown, the importance of the word "bank" is 0.65. The importance of the word "finance" is 0.34. The importance of the other words is also calculated by the learning hierarchical relation extraction unit 103.

[0069] Among the words shown in FIG. 7, the word "bank" satisfies the conditions explained in the processing of the learning hierarchical relation extraction unit 103. That is, a) The most important word in the text data of the educational content (here, the word “bank” is the most important word, and its importance is 0.65) has not been acquired as a learning hierarchical relationship of the target curriculum code, b) And the importance of the most important word (the word "bank") (0.65) is higher than the importance of the search target word (the word "finance") (0.34), c) And the importance (0.65) of the most important word (the word "bank") is equal to or greater than a predetermined threshold (e.g., 0.30); d) And the most important word is a word that exists in the concept dictionary data, The above condition is satisfied. When this condition is satisfied, the learning hierarchical relationship extraction unit 103 extracts the hyponym (subordinate concept) "bank" corresponding to the hypernym (superordinate concept) "finance." In other words, the learning hierarchical relationship extraction unit 103 extracts the hyponym-subordinate relationship of "finance"-"bank." Furthermore, the learning hierarchical relationship extraction unit 103 adds the word "bank" as a search target word.

[0070] FIG. 8 is a schematic diagram showing an example of superordinate and subordinate relationships associated with a curriculum guideline code. In the figure, dashed lines represent superordinate and subordinate relationships obtained from the concept dictionary data. Solid lines represent superordinate and subordinate relationships obtained by the learning hierarchical relationship extraction unit 103 based on the text data of the content. In the example shown, the target curriculum guideline code is "8323233211300000." For example, the superordinate and subordinate relationship of "production" - "company" is shown with a solid line. The superordinate and subordinate relationship of "company" - "Antimonopoly Act" is also shown with a solid line. The superordinate and subordinate relationship of "production" - "manufacturing" is also shown with a dashed line. The superordinate and subordinate relationship of "finance" - "banking" is also shown with a solid line (see also FIG. 7). The same applies to the other superordinate and subordinate relationships shown in the figure.

[0071] The learning hierarchical relationship extraction unit 103 passes the content data of the educational content, the source content information from which the word pairs (superordinate-subordinate relationships) were obtained, and structured data (e.g., RDF) representing the obtained learning hierarchical relationships to the content systematization unit 104.

[0072] The details of the method by which the content systematization unit 104 links (systematizes) the contents in a hierarchical relationship are as follows.

[0073] The content systematization unit 104 receives content data of general content from the content data acquisition unit 101. The content systematization unit 104 also receives content data of educational content, source content information from which word pairs (superordinate-subordinate relationships) were acquired, and structured data representing the acquired learning hierarchical relationships from the learning hierarchical relationship extraction unit 103. Based on this information, the content systematization unit 104 associates the general content and educational content with the learning hierarchical relationship data. In other words, the content systematization unit 104 associates each piece of content with a word appearing in the learning hierarchical relationship tree. Specifically, the content systematization unit 104 sets the word located at the lowest level of the learning hierarchical relationship as the link position of the content, so that the user can select content according to their interests and concerns. In this way, the content systematization unit 104 systematizes the content so that it is linked to more specific words.

[0074] More specifically, the content systematization unit 104 associates the educational content (source content) used to acquire the learning hierarchical relationship with the word at the lowest level. In other words, the content systematization unit 104 associates the source content, which is the source of acquiring the pair relationship between the word at the lowest level and the word one level higher, with the word at the lowest level.

[0075] Furthermore, for general content, the content systematization unit 104 extracts keywords from the text data of the general content. When extracting these keywords, the content systematization unit 104 uses TF-IDF for each time period of the content distribution date and time, TF-IDF for each genre, etc. However, the content systematization unit 104 may also extract keywords from the text data of the general content based on values ​​other than TF-IDF. Furthermore, if the content data of the general content already has keywords, the content systematization unit 104 may use those keywords instead of extracting keywords from the text data. Then, the content systematization unit 104 searches the learning hierarchical relationship for each of those keywords. If the search results in a word matching the keyword being the lowest word in the hierarchical relationship (which may be an alias), the general content is linked using that word as the link position.

[0076] The content systematization unit 104 may link general content that includes both a word at the lowest level in the hierarchy and a word one level above it to the word at the lowest level. This allows general content that is more in line with the context of the curriculum guidelines to be linked to the hierarchical relationship.

[0077] On the other hand, the content systematization unit 104 may link general content that includes only words at the lowest level in the hierarchy (i.e., general content that does not include words one level higher) to the words at the lowest level. In this case, it is expected that the user will select content that will broaden their interests, starting from the specific word.

[0078] The content systematization unit 104 passes to the visualization unit 105 information on the learning hierarchical relationships and information on the connections (linkages) of the content to the hierarchical relationships.

[0079] The visualization unit 105 visualizes the learning hierarchical relationships passed from the content systematization unit 104 and the connections between the contents linked to the hierarchical relationships. That is, the visualization unit 105 generates a user interface (UI) that shows the hierarchical relationships and the connections between the contents. Examples of the user interface generated by the visualization unit 105 are shown in FIGS. 9 and 10.

[0080] 9 is a schematic diagram showing an example of a user interface generated by the visualization unit 105. As shown in the figure, in this example, nodes in a hierarchical relationship are represented by circles, and the hierarchical relationship (superordinate-subordinate relationship) between nodes is represented by the inclusion relationship of those circles. When a circle (called a first circle) is included in another circle (called a second circle), the first circle corresponds to a superordinate concept (hyperword), and the second circle corresponds to a subordinate concept (hyperword).

[0081] In the example shown, circle 801 is the outermost circle and corresponds to a specific curriculum code. At the level immediately below circle 801, there are circles 802 and 803. That is, circles 802 and 803 are each contained within circle 801. Circle 802 corresponds to the word "production." Circle 803 corresponds to the word "finance." This means that the words "production" and "finance" belong to the curriculum code of circle 801. At the level immediately below circle 802, there are circles 804 and 805. That is, circles 804 and 805 are each contained within circle 802. Circle 804 corresponds to the word "company." Circle 805 corresponds to the word "manufacturing." In other words, a hierarchical relationship between the word "production" and the word "company" and a hierarchical relationship between the word "production" and the word "manufacturing" are expressed. At the level immediately below circle 803, there are circles 806 and 807. In other words, each of circles 806 and 807 is contained within circle 803. Circle 806 corresponds to the word "banknote". Circle 807 corresponds to the word "bank". In other words, a hierarchical relationship between the word "finance" and the word "banknote" and a hierarchical relationship between the word "finance" and the word "bank" are expressed.

[0082] Although a detailed description will be omitted, hereinafter, lower-level words may be represented using circles. Furthermore, each of the lowest-level circles (the innermost circles in this example) may be associated with content. In this example, each piece of content is represented by a rectangle.

[0083] In other words, in the visualization format shown in Figure 9, for any curriculum code, the outer circle corresponds to the core concepts (concepts that should be learned) included in the curriculum. The concepts become more specific as you go inward. The content linked to the specific concepts at the lowest level is displayed.

[0084] FIG. 10 is a schematic diagram showing another example of a user interface generated by the visualization unit 105. As shown in the figure, in this example, hierarchical relationships (superordinate and subordinate relationships) are represented by a graph structure (tree structure). That is, circular nodes in the graph correspond to concepts (words). Furthermore, rectangular nodes correspond to content. Lines (edges) connecting the circular nodes represent hierarchical relationships (superordinate and subordinate relationships) between concepts (words). Furthermore, rectangular nodes are connected to the lowest-level concepts (words). Edges connecting circular nodes with rectangular nodes represent the relationship between the lowest-level concepts (words) and the content linked to those concepts (words).

[0085] In the example shown, node 821 corresponds to the top-level word "production" in this diagram. Node 821 is connected by an edge to node 822 at the immediately lower level. Node 822 corresponds to the word "production." In other words, it represents a hierarchical relationship between the word "production" and the word "production." Node 822 is connected by an edge to each of nodes 823 and 824 at the immediately lower level. Node 823 corresponds to the word "tofu." Node 824 corresponds to the word "strong yen." In other words, it represents a hierarchical relationship between the word "production" and the word "tofu," and a hierarchical relationship between the word "production" and the word "strong yen."

[0086] Node 823 is connected by edges to nodes 831 and 832, which are each represented by a rectangle. Rectangular nodes 831 and 832 each represent content. The content corresponding to nodes 831 and 832 is linked to the word "tofu" at the lowest level, which corresponds to node 823. Node 824 is connected by edges to nodes 833 and 834, which are each represented by a rectangle. Rectangular nodes 833 and 834 each represent content. The content corresponding to nodes 833 and 834 is linked to the word "strong yen" at the lowest level, which corresponds to node 824.

[0087] That is, in the visualization form shown in Fig. 10, the learning hierarchical relationships are displayed as a graph (tree structure). In addition, the connections between the contents linked to the hierarchical relationships are also displayed as part of the graph (tree structure).

[0088] Examples of visualization forms for linking hierarchical relationships and contents have been described above with reference to Figures 9 and 10. However, visualization methods are not limited to these examples, and other methods may also be used.

[0089] 11 and 12 are flowcharts showing the processing steps performed by the relationship extraction device 1. Fig. 11 and Fig. 12 are connected by a connector. In other words, Fig. 11 and Fig. 12 together form a single flowchart. Below, the processing steps for extracting relationships and systematizing content will be explained in accordance with this flowchart.

[0090] First, in step S001, the content data acquisition unit 101 acquires content data.

[0091] Next, in step S002, the external data acquisition unit 102 acquires concept dictionary data and curriculum guideline data.

[0092] [Repeat by curriculum code] Step S003 is the start of the repetitive process. Corresponding to step S003, step S013 (FIG. 12) is the end of the repetitive process. The repetitive process from step S003 to step S013 is executed for each curriculum guideline code in the curriculum guideline data acquired in step S002.

[0093] In step S004, the learning hierarchy relation extraction unit 103 extracts superordinate and subordinate relations from the educational content associated with the current curriculum code. Specifically, in step S004, the learning hierarchy relation extraction unit 103 extracts nouns from the text of the educational content associated with the current curriculum code, and extracts superordinate and subordinate relations based on the concept dictionary data from within the same content, including aliases for each noun.

[0094] Next, in step S005, the learning hierarchy relation extraction unit 103 selects the nouns associated with the current curriculum code that are included in the concept dictionary data as top-level word candidates, and acquires these nouns, including their subordinate words, as a set of search target words.

[0095] [Repeat for each search term] Step S006 is the start of the repetitive process. Corresponding to step S006, step S012 (FIG. 12) is the end of the repetitive process. The repetitive process from step S006 to step S012 is performed for each search target term acquired in step S005 (however, there may be search target terms added in the following processes). The process from step S006 to step S012 is a process for extracting a learning hierarchical relationship.

[0096] In step S007, the learning hierarchical relationship extraction unit 103 acquires, from the unprocessed content, content that does not have any other search target word with a higher importance than the current search target word. However, if there is no unprocessed content, the learning hierarchical relationship extraction unit 103 ends the processing for the search target word and moves on to processing the next search target word.

[0097] In step S008, the learning hierarchical relation extraction unit 103 determines whether the following conditions are true or false. That is, the learning hierarchical relation extraction unit 103 determines whether a) the most important word of the content has not yet been acquired as a learning hierarchical relation of the current curriculum guideline code, b) the importance of the most important word is higher than the importance of the current search target word, c) the importance of the most important word is equal to or greater than a predetermined threshold (for example, the threshold is 0.3), and d) the most important word exists in the concept dictionary data. If the determination result is true (step S008: YES), the process proceeds to the next step S009. If the determination result is false (step S008: NO), the process returns to step S007 without proceeding to step S009 (i.e., attempts to process another content).

[0098] If the process proceeds to step S009, in that step the learning hierarchical relationship extraction unit 103 adds the most important word acquired in step S008 to the hyponyms and also to the search target words. The learning hierarchical relationship extraction unit 103 also acquires aliases of the most important word by referring to the concept dictionary data, and adds these aliases to the hyponyms and also to the search target words.

[0099] 12, in step S010, the learning hierarchical relation extraction unit 103 determines whether or not a hyponym of the word just added has been acquired. If it has been acquired (step S010: YES), the process proceeds to step S011. If it has not been acquired (step S010: NO), the process proceeds to step S012.

[0100] If the process proceeds to step S011, the learning hierarchical relation extraction unit 103 acquires an alias and adds the alias to the hyponym and the search target word. After the process of step S011, the process returns to step S010.

[0101] In step S012, the repeated process that started in step S006 is ended, that is, the process returns to step S006 to process the next search target word. After the process for all search target words has been completed, the process proceeds to step S013.

[0102] In step S013, the repeated process that started in step S003 is ended, that is, the process returns to step S003 to process the next curriculum code. After the process for all search target words has been completed, the process proceeds to step S014.

[0103] In step S014, the content systematization unit 104 links the contents within the constructed hierarchical relationship. Details of how the content systematization unit 104 links the contents have already been described.

[0104] In step S015, the visualization unit 105 visualizes the constructed hierarchical relationship. The visualization unit 105 may also visualize the content linked to the hierarchical relationship. Examples of the visualization format have been described above. The visualization unit 105 displays a visualization screen.

[0105] FIG. 13 is a block diagram showing an example of the internal configuration of the relationship extraction device 1 when it is realized using a computer. The relationship extraction device 1 can be realized using a computer. As shown in the figure, the computer includes a central processing unit 901, a RAM 902, an input / output port 903, input / output devices 904 and 905, etc., and a bus 906. The computer itself can be realized using existing technology. The central processing unit 901 executes instructions contained in a program read from the RAM 902, etc. In accordance with each instruction, the central processing unit 901 writes data to the RAM 902, reads data from the RAM 902, and performs arithmetic and logical operations. The RAM 902 stores data and programs. Each element included in the RAM 902 has an address and can be accessed using the address. RAM is an abbreviation for "random access memory." The input / output port 903 is a port through which the central processing unit 901 exchanges data with external input / output devices, etc. Input / output devices 904 and 905 exchange data with the central processing unit 901 via an input / output port 903. A bus 906 is a common communication path used within the computer. For example, the central processing unit 901 reads and writes data from and to RAM 902 via the bus 906. Also, for example, the central processing unit 901 accesses the input / output port 903 via the bus 906.

[0106] At least some of the functions of the relationship extraction device 1 in the above-described embodiment can be implemented by a computer and a program. In this case, the functions can be implemented by recording a program for implementing the functions on a computer-readable recording medium and loading and executing the program recorded on the recording medium into a computer system. Note that the term "computer system" as used herein includes hardware such as an OS and peripheral devices. Furthermore, the term "computer-readable recording medium" refers to portable media such as flexible disks, optical magnetic disks, ROMs, CD-ROMs, DVD-ROMs, and USB memory, as well as storage devices such as hard disks built into computer systems. In other words, a "computer-readable recording medium" may be a non-transitory computer-readable recording medium. Furthermore, the term "computer-readable recording medium" may also include media that temporarily and dynamically store programs, such as communication lines used when transmitting programs via networks such as the Internet or telephone lines, or media that store programs for a certain period of time, such as volatile memory within a computer system that serves as a server or client in such cases. The program may also be designed to implement some of the functions described above, or may be capable of implementing the functions described above in combination with a program already stored in the computer system.

[0107] Although the embodiment has been described above, the present invention can also be carried out in the following modified examples.

[0108] As one modified example, for example, the relationship extraction device 1 may not have the visualization unit 105. Even in this case, the learning hierarchical relationship extraction unit 103 can extract the learning hierarchical relationships. Also, the content systematization unit 104 can systematize the content.

[0109] As a modified example, the relationship extraction device 1 may not have the content systematization unit 104. Even in this case, the learning hierarchical relationship extraction unit 103 can extract the learning hierarchical relationship. Furthermore, the visualization unit 105 can visualize the hierarchical relationship.

[0110] It should be noted that multiple modified examples may be combined to the extent possible.

[0111] Although an embodiment of the present invention has been described in detail above with reference to the drawings, the specific configuration is not limited to this embodiment, and includes designs within the scope of the gist of the present invention.

[0112] As described above, according to this embodiment (including the modified examples), it is possible to extract the hierarchical relationships of words based on content. Furthermore, according to this embodiment, it is possible to associate (link) content with the hierarchical relationships of the extracted words. Furthermore, according to this embodiment, it is possible to visualize the hierarchical relationships of words. Furthermore, it is possible to visualize linked content along with the hierarchical relationships of words.

[0113] That is, according to this embodiment, it is possible to systematize content such as broadcast content and online distribution content (for example, video content) from the perspective of learning.

[0114] That is, according to this embodiment, it is possible to select content that matches the user's interests and concerns from among a variety of content related to the core concepts to be learned. [Industrial Applicability]

[0115] The present invention can be used in the content distribution business, for example, but the scope of use of the present invention is not limited to the examples given here. [Explanation of symbols]

[0116] 1. Relationship Extraction Device 101 Content data acquisition unit 102 External data acquisition unit 103 Learning Hierarchy Relation Extraction Unit 104 Content Systematization Department 105 Visualization section 901 Central Processing Unit 902 RAM 903 Input / Output Ports 904,905 Input / Output Devices 906 Bus

Claims

1. A curriculum text associated with the curriculum code; Concept dictionary data is data that defines the hierarchical and subordinate relationships between words and other words; An educational content text that is a text of educational content associated with the curriculum code; a learning hierarchy relation extraction unit that extracts, based on the above, a hierarchy relation between the search target word and another search target word as a hierarchy relation associated with the learning guideline code; Equipped with The learning hierarchical relationship extraction unit The words included in the curriculum text and defined in the concept dictionary data are used as the top search target words, For each search term, a) a word contained in an educational content text, in which there is no other search target word in the educational content text that is more important than the search target word, and the most important word that is the most important word in the educational content text has not been acquired as a search target word; b) The importance of the most important word is higher than the importance of the search target word; c) and the importance of the most important word is equal to or greater than a predetermined threshold; d) And the most important word is a word defined in the concept dictionary data, In the step 1, a hierarchy relationship associated with the curriculum code is extracted, in which the search target word is higher and the most important word is lower, and the most important word is added as one of the search target words. Relationship extractor.

2. The learning hierarchical relationship extraction unit further extracts, when two words included in one of the educational content texts are defined as having a superordinate-subordinate relationship in the concept dictionary data, the superordinate-subordinate relationship as a superordinate-subordinate relationship associated with the curriculum guideline code. The relationship extraction device according to claim 1 .

3. The concept dictionary data further defines an alias relationship in which a certain word is an alias of another word, When extracting the hierarchy / subordinate relationship associated with the course of study code, in which the search target word is at a higher level and the most important word is at a lower level, the learning hierarchical relationship extraction unit further refers to the concept dictionary data to extract the hierarchy / subordinate relationship associated with the course of study code, in which the search target word is at a higher level and the word corresponding to the alias is at a lower level, with a word corresponding to the alias being the lower word, and performs a process of adding the word corresponding to the alias as one of the search target words. The relationship extraction device according to claim 1 .

4. a content systematization unit that performs processing to associate content corresponding to the educational content text that includes the most important word when the learning hierarchical relationship extraction unit extracts the superordinate / subordinate relationship associated with the curriculum guideline code with the subordinate word of the superordinate / subordinate relationship; The relationship extraction device according to claim 1 , further comprising:

5. a visualization unit that graphically visualizes and displays a set of superordinate and subordinate relationships that the learning hierarchical relationship extraction unit has extracted in association with the curriculum guideline code; The relationship extraction device according to claim 1 , further comprising:

6. A curriculum text associated with the curriculum code; Concept dictionary data is data that defines the hierarchical and subordinate relationships between words and other words; An educational content text that is a text of educational content associated with the curriculum code; a learning hierarchy relation extraction unit that extracts, based on the above, a hierarchy relation between the search target word and another search target word as a hierarchy relation associated with the learning guideline code; Equipped with The learning hierarchical relationship extraction unit The words included in the curriculum text and defined in the concept dictionary data are used as the top search target words, For each search term, a) a word contained in an educational content text, in which there is no other search target word in the educational content text that is more important than the search target word, and the most important word that is the most important word in the educational content text has not been acquired as a search target word; b) The importance of the most important word is higher than the importance of the search target word; c) and the importance of the most important word is equal to or greater than a predetermined threshold; d) And the most important word is a word defined in the concept dictionary data, In the step 1, a hierarchy relationship associated with the curriculum code is extracted, in which the search target word is higher and the most important word is lower, and the most important word is added as one of the search target words. A program that enables a computer to function as a relationship extraction device.

Images