JP2017059077A - Information provision device, information provision method, and information provision program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To allow for outputting information that assists creativity of a user.SOLUTION: An information provision device of the present invention includes; a reception unit for receiving input information; a pattern identification unit configured to extract syntax and a group of words embedded in the syntax from the input information; a similarity computation unit configured to compute similarity using a distributed representation of the identified group of words included in an identified pattern; and an output unit configured to output the extracted information based on the similarity computed by the similarity computation unit.SELECTED DRAWING: Figure 1

Description

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

  2. Description of the Related Art Conventionally, a technique for searching or generating information related to input information based on an analysis result of input information and outputting the searched or generated information as a response is known. As an example of such a technology, words, sentences, and contexts contained in the input text are converted into multidimensional vectors and analyzed, and based on the analysis result, text similar to the input text or input A natural language processing technique for analogizing text following text and outputting an analogy result is known.

JP 2006-127077 A

“Natural Language Processing with word2vec”, Yasukazu Nishio, May 2014, ISBN978-4-87311-683-9 "Research on hypothetical knowledge generation support for creative design", The Japan Society of Mechanical Engineers, No03-27, 13th Design Engineering and System Division Lecture Proceedings

  However, in the above-described conventional technology, there are cases where information for assisting the creation of the user cannot be output. For example, in the above-described prior art, only information that can be predicted by the user, such as text similar to the input text or text following the input text, is output, and is related to the input text. However, it is difficult to output information that the user cannot predict. For this reason, the above-described conventional technology cannot provide information that gives inspiration to the user.

  The present application has been made in view of the above, and an object thereof is to provide an information providing apparatus, an information providing method, and an information providing program capable of outputting information for assisting creation of a user.

  The information providing apparatus according to the present application includes a receiving unit that receives input information, a pattern specifying unit that extracts a specific syntax and a group of words embedded in the specific syntax from the input information, and the specified pattern included in the specified pattern. A similarity calculation unit that calculates similarity using a distributed representation of a word group, and an output unit that outputs information extracted based on the similarity calculated by the similarity calculation unit.

  According to one aspect of the embodiment, it is possible to output information that assists the creation of the user.

FIG. 1 is a diagram illustrating an example of an information providing apparatus according to the embodiment. FIG. 2 is a diagram showing an equivalent conversion theory which is one of the invention excavation methods. FIG. 3 is a diagram illustrating a first example of input and output variations of the information providing apparatus according to the embodiment. FIG. 4 is a diagram illustrating a second example of input and output variations of the information providing apparatus according to the embodiment. FIG. 5 is a diagram illustrating a third example of input and output variations of the information providing apparatus according to the embodiment. FIG. 6 is a diagram illustrating a fourth example of input and output variations of the information providing apparatus according to the embodiment. FIG. 7 is a diagram illustrating an example of a functional configuration of the information providing apparatus according to the embodiment. FIG. 8 is a diagram for explaining an example of words extracted by the information providing apparatus according to the embodiment. FIG. 9 is a diagram illustrating an example of a word set extracted by the information providing apparatus according to the embodiment. FIG. 10 is a diagram for explaining an example of processing for extending the distributed representation space executed by the information providing apparatus according to the embodiment. FIG. 11 is a diagram for explaining the limitation and expansion of the distributed representation space executed by the information providing apparatus according to the embodiment. FIG. 12 is a diagram illustrating an example of processing for extracting similar words by the information providing apparatus according to the embodiment. FIG. 13 is a diagram for explaining an example of the relationship between word vectors. FIG. 14 is a diagram for explaining an example of the concept of the word proposed by the information providing apparatus. FIG. 15 is a flowchart illustrating a flow of extraction processing executed by the information providing apparatus according to the embodiment. FIG. 16 is a flowchart for explaining a specific processing flow of the extraction processing executed by the information providing apparatus according to the embodiment. FIG. 17 is a flowchart for explaining a specific processing flow of processing for selecting a word group that is likely to cause serendipity by the information providing apparatus according to the embodiment. FIG. 18 is a diagram illustrating an example of processing for outputting information related to information in an input field and information in a different field. FIG. 19 is a hardware configuration diagram illustrating an example of a computer that implements extraction processing.

  Hereinafter, a mode for implementing an information providing apparatus, an information providing method, and an information providing program according to the present application (hereinafter referred to as “embodiment”) will be described in detail with reference to the drawings. Note that the information providing apparatus, the information providing method, and the information providing program according to the present application are not limited by this embodiment. Moreover, in the following embodiment, the same code | symbol is attached | subjected to the same site | part and the overlapping description is abbreviate | omitted.

[1. Example of information providing device)
First, an example of processing executed by the information providing apparatus 10 will be described with reference to FIG. FIG. 1 is a diagram illustrating an example of an information providing apparatus according to the embodiment. For example, in the example illustrated in FIG. 1, the information providing apparatus 10 acquires a user's remarks as input information when a plurality of users are holding a conference such as brainstorming, and based on the acquired input information. An example of processing that supports the user's thought by generating a speech that does not appear in normal thinking, converting the generated speech to speech and outputting it to a robot or the like has been described.

  The information providing apparatus 10 illustrated in FIG. 1 is realized by an information processing apparatus such as a server apparatus. Note that the information providing apparatus 10 may be realized as a single information processing apparatus. For example, a plurality of information processing apparatuses existing on the cloud network may be realized in cooperation. Such an information providing apparatus 10 converts a user's speech into text data and analyzes the text data by natural language processing. And the information provision apparatus 10 produces | generates the speech which supports a meeting and a user's thought based on an analysis result, and outputs the produced | generated speech.

  Here, in the prior art, text similar to the input text or text following the input text is inferred using a distributed expression represented by a multi-dimensional word vector constituting the input text. However, the prior art merely outputs text whose distributed representation is similar to the input text, that is, text that can be predicted by the user. For this reason, the conventional technology provides information that the user has never thought of, information that gives the user a new inspiration, that is, information that can cause the user's serendipity (inspiration, awareness, surprise). I could not. Also, for example, if text that does not resemble a simple input text is output, text that is not related to the user's thought will be output, which may hinder the user's thought. is there.

  Therefore, the information providing apparatus 10 executes the following process. First, the information processing apparatus 10 receives a user's remarks as input information. When the information processing apparatus 10 receives the input information, the information processing apparatus 10 identifies a specific syntax from the input information and extracts a word group embedded in the identified specific syntax. In addition, the information processing apparatus 10 calculates the degree of similarity with the distributed expression of other words using the distributed expression of the word group included in the specific syntax. Then, the information providing apparatus 10 outputs information extracted based on the calculated similarity.

  Specifically, the information providing apparatus 10 specifies the field to which the input information belongs. Further, the information providing apparatus 10 specifies a plurality of pieces of information having a predetermined relationship from information belonging to the specified field. And the information provision apparatus 10 extracts the information which has the concept similar to the concept produced by the specified several information from the information which belongs to the field | area different from the specified field | area.

  Here, the information providing apparatus 10 uses the distributed representation of a plurality of pieces of information having a predetermined relevance to input information having an attribute whose similarity with the attribute generated by the pieces of information satisfies a predetermined condition as input information Extraction is made from information belonging to a field in which the similarity to the field to which it belongs is equal to or less than a predetermined threshold. For example, the information providing apparatus 10 identifies a plurality of words having a predetermined relationship from documents in the field to which the input information belongs, and identifies a field having the same attribute as the attribute generated by the identified plurality of words. Are extracted from documents belonging to the field in which the degree of similarity is equal to or less than a predetermined threshold. Then, the information providing apparatus 10 outputs the extracted information.

  Here, the concept generated by a plurality of pieces of information is, for example, an attribute of each piece of information. More specific examples will be described. When a word is adopted as information, the attribute of a plurality of words includes a set of attributes such as how many words each word has and the meaning of each word. In other words, the concept is a word set. In the following description, the information providing apparatus 10 describes the attribute of a word, biological information, content, etc. handled as information as a “concept” of the information, and includes a plurality of words, a plurality of biological information, and a plurality of contents. Are described as “concepts” of the plurality of pieces of information.

  That is, the information providing apparatus 10 identifies a concept formed by information having a predetermined relationship in the category to which the input information belongs, and information that can form a common concept from information belonging to a category different from the input information. Extract. As a result of such processing, the information providing apparatus 10 maintains the implicit connection with the input information, but explicitly has discontinuous information, that is, as a result of different categories, it seems that there is no relationship at first glance. Visible information can be provided.

  Here, information that has an implicit discontinuous relationship with the input information, that is, information that seems to have no relationship at first glance as a result of different categories, is assumed by the user. It can be information that did not exist or information that gives the user a new inspiration.

  For example, the information providing apparatus 10 identifies a plurality of pieces of information indicating points of invention or idea in the field from the information of the field to which the input information belongs, and indicates the concept indicated by the identified information, that is, the point of invention or idea. Identify the concept. Then, the information providing apparatus 10 extracts information similar to the identified concept from information in other fields. The information extracted by such processing suggests to the user information that satisfies the concept indicating the point of invention or idea in a field different from the field to which the input information belongs, that is, the point of invention or idea in another field. It can be information that can cause serendipity of users. As a result, the information providing apparatus 10 can assist the creation of the user.

For example, FIG. 2 is a diagram showing an equivalent conversion theory that is one of the invention excavation techniques. For example, as shown in FIG. 2A, Ao, which is the original specific event (for example, the original invention), is a viewpoint vi that suits the development purpose as shown in FIG. It consists of. Further, Ao includes a special condition Σa possessed by Ao, such as a technical attribute to which Ao belongs, as shown in FIG. Therefore, as shown in FIG. 2D, the concept cε obtained by removing Σa from Ao can be a concept that forms the core of Ao. Then, as shown in FIG. 2E, when a special condition Σb different from Σa is considered in the concept cε, as shown in FIG. _It is considered that _τ can be derived.

Therefore, the information providing apparatus 10 extracts the concept of the idea in the field to which the input information belongs, that is, the concept cε that forms the core of Ao by extraction processing described later, and applies the extracted concept cε to the condition Σb of the other field. Thus, information that can derive a new idea _Bτ is provided to the user.

  Hereinafter, an example of a specific extraction process executed by the information providing apparatus 10 will be described using the example of FIG. First, the information providing apparatus 10 receives the user's speech A and speech B in the conference as input (step S1). For example, the information providing apparatus 10 converts the utterance A uttered by the user into text data, and acquires the converted text data as input information.

  In this case, the information providing apparatus 10 shows a concept that is a word that belongs to a field different from the field to which the input information belongs, and that is formed by a plurality of words having a predetermined relationship in the field to which the input information belongs. An extraction process for extracting a word is executed (step S2). Hereinafter, the flow of the generation process executed by the information providing apparatus 10 will be described in steps S3 to S8.

  First, the information providing apparatus 10 specifies the field to which the input information belongs (step S3). For example, the information providing apparatus 10 analyzes text data of a user's utterance and identifies in which field the word included in the text data is used. For example, when the information providing apparatus 10 includes words such as “glasses type”, “watch type”, “display”, “smart device”, etc., the field to which the input information belongs is “wearable device”. Note that the information providing apparatus 10 stores, for example, a word that will be included in the input information and a field to which the input information including the word belongs in advance, and is associated with the word included in the input information. May be specified as a field to which the input information belongs. For example, using a search service on the web or the like, a field related to a word included in the input information is searched and a field obtained as a search result is input. It may be a field to which information belongs.

  Subsequently, the information providing apparatus 10 extracts a word set including a plurality of words having a predetermined relationship from the document data in the specified field (step S4). For example, the information providing apparatus 10 acquires document data in a specified field from a document database that describes various technologies such as a published patent gazette and a patent gazette. Subsequently, the information providing apparatus 10 performs morphological analysis on the acquired document data, and extracts a plurality of word sets that appear in a predetermined order from the document data. More specifically, the information providing apparatus 10 identifies a sentence having a structure that is used when explaining the idea and features of the invention, and extracts a word set included in the identified sentence.

  To explain specific examples, for example, a sentence that may explain a feature of an idea or an invention, that is, a feature of an idea is a noun indicating a target of processing, a state expression word or an adverb phrase indicating a target state indicated by the noun , And a verb indicating the content of the process. Also, in such a sentence, a part of speech that can exist between words such as particles is enclosed in curly brackets, and "(noun) {"} (state expression word, adverb phrase) {, by}} (verb ) {Yes, I can do} ". Hereinafter, such a syntax structure may be referred to as a cε dictionary method.

  Therefore, the information providing apparatus 10 extracts the sentence having the structure described above from the literature data of the specified field, and three words included in the extracted sentence, that is, a noun, a state expression word or an adverb phrase, and a verb, Extracted as a word set indicating an idea. As described above, the information providing apparatus 10 extracts the specific syntax from the input information, and extracts a word group embedded in the extracted specific syntax. Here, the specific syntax is a syntax having a predetermined pattern, that is, a structure of cε dictionary. Since the word set extracted in this way can reproduce a sentence showing an idea such as an invention or idea only by adding a predetermined case particle, the concept indicated by the sentence from which the word set is extracted, That is, the concept of an idea such as an invention or an idea can be shown.

  Here, it is predicted that the document data belonging to a certain field includes a large number of word sets indicating concepts that are commonly used in the field. For example, it is predicted that the literature data in the program field will contain many sentences with the concept of “realizing” an arbitrary idea by “creating” a “program”. For this reason, it can be said that the word set including words such as “program, creation, realization” and the like among the word sets extracted from the literature data in the program field includes a concept that can cause the serendipity of the user. Absent.

  Therefore, the information providing apparatus 10 selects a word group that can cause serendipity from the extracted word group (step S5). That is, the information providing apparatus 10 selects a word set that indicates a concept that is difficult for the user to conceive and is easy to understand when presented. Specifically, the information providing apparatus 10 can generate serendipity based on the number of word pairs indicating the same concept among the other extracted word sets and the relationship between words included in the word set. Select a pair.

  For example, the information providing apparatus 10 selects a word group in which the number of word groups indicating the same concept is equal to or less than a predetermined number in order to select a word group indicating a concept that is harder for the user to conceive. In addition, the information providing apparatus 10 selects a word set indicating a concept that is easy to understand when presented, so that a word set in which words included in the word set are related words from the selected word set, When a related word is traced from a word included in the word group, a word group in which the number of related words traced until reaching another word included in the same word group is equal to or less than a predetermined threshold is selected.

  Here, even if it is a word set that can cause serendipity, a word set that is used across a plurality of documents is a word set that is easy for a user to predict, and therefore, the probability of causing serendipity is low. Therefore, the information providing apparatus 10 removes the word group used across the selected word group (step S6).

  For example, the information providing apparatus 10 calculates a sentence frequency (DF: Document Frequency) that is a frequency at which the selected word group is included in each document. To explain a more specific example, the information providing apparatus 10 calculates a value obtained by dividing the number of documents in which the word group appears for each word group by the number of all documents. And the information provision apparatus 10 selects the word group from which the calculated value becomes below a predetermined threshold value. Note that the documents used for calculating the sentence frequency may or may not be limited to documents in the field specified in step S3.

  Here, when the sentence frequency is set to a value lower than the predetermined threshold, the number of selected word sets is reduced. Therefore, the information providing apparatus 10 expands the selected word set using verbs indicating various operations performed on the object indicated by the noun (hereinafter, referred to as operational verbs) (step S7). . In other words, the information providing apparatus 10 generates a new set by combining a plurality of pieces of information indicating a predetermined operation with respect to a set including a plurality of pieces of information having a predetermined relevance. Extends the concept of a set containing information on

  For example, the information providing apparatus 10 stores in advance the Koverk & Baknar operational verbs. Here, Koverk & Bakunar operational verbs are, for example, “increase”, “divide”, “remove”, “relieve”, “reverse”, “disconnect”, “replace”, “integrate” , “Twist”, “rotate”, “smooth out”, “squeeze”, “supplement”, “submerge”, “freeze”, “soften”, “inflate”, “detour” "Do", "add", "subtract", "lighten", "repeat", "thicken", "stretch", "push out", "splash", "defend", "pull away", The verbs are “integrate”, “symbolize”, “abstract”, “cut”.

  Subsequently, the information providing apparatus 10 newly generates a word set in which the verb included in the selected word set is replaced with each of the above-described operational verbs. For example, the information providing apparatus 10 newly generates a word set such as “glasses, wear, increase”, “glasses, wear, divide” from a word set such as “glasses, wear, view”. As a result, the information providing apparatus 10 can generate a concept variation that is highly likely to cause serendipity for the user.

  Subsequently, for each word set, the information providing apparatus 10 maps a vector indicating the word set on a space indicating words in various fields as vectors, that is, on a distributed expression space including the words expressed in a distributed manner. . Then, the information providing apparatus 10 extracts, from the distributed expression space, words that belong to a field different from the field to which the word group belongs and that have the same concept as the word group (step S8).

  For example, the information providing apparatus 10 stores in advance a distributed expression space including distributed expressions of words belonging to various fields using W2V (word2vec). In such a distributed expression space, a distributed expression of words included in a document belonging to the field specified in step S3 is also included. Subsequently, the information providing apparatus 10 uses W2V to convert each word included in the word set into a distributed representation, and calculates the sum of the distributed representation of each word. That is, the information providing apparatus 10 calculates the sum of vectors indicating each word included in the word set.

  Here, on the distributed representation space, the direction indicated by the sum of vectors indicating each word included in the word set (hereinafter referred to as word set vector) is considered to suggest the concept indicated by the word set. It is done. For this reason, it is considered that the word indicated by the vector whose direction is similar to that of the word set vector has a concept similar to the concept indicated by the word set. However, even if the word has a concept similar to the concept indicated by the word set, a word in the same field as the field to which the word set belongs is a word that can be predicted by the user, and is less likely to cause serendipity.

  Therefore, the information providing apparatus 10 extracts words that have a concept similar to the concept indicated by the word set and are included in documents in a field different from the field to which the word set belongs. More specifically, the information providing apparatus 10 is a vector whose direction is similar to that of a word set vector, and a word indicated by a vector that is separated from the word set vector in the distributed expression space, that is, the word set. A word indicating a concept similar to the concept indicated by, which belongs to a field different from the word set, is extracted.

  For example, as the directions of two vectors are aligned, the value of the cosine distance between the two vectors increases. Further, as the cosine distance between a certain synthesized vector and another vector increases, the value of the cosine distance between each word vector constituting the synthesized vector and another vector decreases. Therefore, the information providing apparatus 10 calculates the cosine distance between the word set vector and each vector in the distributed expression space. Further, the information providing apparatus 10 calculates a cosine distance between each word vector included in the word set and each vector on the distributed expression space.

  Then, the information providing apparatus 10 has a cosine distance value with a word set vector larger than a predetermined threshold value, and a sum of cosine distances with each word vector included in the word set is smaller than a predetermined threshold value. A vector is extracted, and a word indicated by the extracted vector is specified. In other words, the information providing apparatus 10 extracts words having a concept similar to the concept indicated by the word set and having a field different from the field to which the word set belongs. In other words, the information providing apparatus 10 extracts a word having an analogy similar to that of the word group extracted from the specific syntax and belonging to a field different from the extracted word group based on the equivalent conversion theory.

  And the information provision apparatus 10 outputs the extracted word as a proposal (step S9). For example, when the word “rotation” is extracted by the process shown in step S8 for the word set “graph, collective, add”, the information providing apparatus 10 reads “Rotate?” A sentence that includes the extracted word and suggests the concept indicated by the extracted word is generated, and the generated sentence is read as a utterance C by a robot or the like to make a proposal to the user.

  When such a proposal is made, the user is likely to generate a new inspiration based on the concept of the proposed word. For this reason, the information provision apparatus 10 can raise a user's serendipity.

[2. Variations on concepts provided by information providing devices)
Here, in the example illustrated in FIG. 1, the process of proposing a word indicating a concept that causes the user to generate serendipity from words included in the user's utterance is described. For example, FIG. 3 is a diagram illustrating a first example of input and output variations of the information providing apparatus according to the embodiment. For example, as shown in FIG. 3A, the information providing apparatus 10 maps old words and different fields by mapping words included in a patent specification such as a public gazette as a document to a distributed expression space using W2V. Generate a distributed representation space that contains a common analogy of technology.

  Subsequently, as shown in FIG. 3B, the information providing apparatus 10 specifies a service and a technical field from the input information. And the information provision apparatus 10 is close to the concept which the analogy in the specified field | area has among the analogies of the field | area which is different from the specified field | area by performing the extraction process mentioned above as shown to (C) in FIG. Extract analogies with concepts. As a result, the information providing apparatus 10 can output a hint of a new idea as shown in (D) of FIG.

  The hint output in this way is a concept similar to the concept of the idea in the field to which the input information belongs, and it is predicted that the concept in a different field can be recalled by the user. For this reason, the information providing apparatus 10 utilizes an idea obtained by developing the idea of the old technology in a new environment as shown in (E) of FIG. 3 or a technology in a different field as shown in (F) of FIG. Can help create new technologies.

  However, the embodiment is not limited to this. For example, the information providing apparatus 10 uses a distributed expression space in which a business model and a business field are input instead of words, and the contents of business models and contracts in various industries are replaced with distributed expressions. You may output the information used as the hint which makes a user raise the serendipity for creating a new business model.

  For example, FIG. 4 is a diagram illustrating a second example of input and output variations of the information providing apparatus according to the embodiment. For example, as shown in FIG. 4 (A), the information providing apparatus 10 maps, as a document, features of past business to a distributed representation space by W2V, so that a common analogy of old business and different field technologies can be obtained. Create a distributed representation space containing.

  Subsequently, as shown in FIG. 4B, the information providing apparatus 10 specifies the type of business to which the service or business belongs from the input information. Then, the information providing apparatus 10 extracts features of services and businesses in the identified field, and converts the extracted features into a distributed representation. In addition, the information providing apparatus 10 extracts business features having characteristics similar to the sum of the features converted into the distributed representation from other business fields, and as shown in FIG. The analogies having concepts similar to the concepts possessed by the analogies in the specified industries are extracted from the analogies of different industries. As a result, the information providing apparatus 10 can output a new business hint as shown in FIG.

  The hint output in this way is a concept similar to the business concept in the industry to which the input information belongs, and it is predicted that the concept in a different industry can be recalled by the user. For this reason, as shown in FIG. 4 (E), the information providing apparatus 10 can develop a business model in which an old business idea has been developed in a new environment, or a business model of a different industry as shown in FIG. 4 (F). It can help create new business models.

  Further, for example, the information providing apparatus 10 uses serendipity for creating new content by using arbitrary content such as music, images, advertisements, and the like instead of words as input and using a distributed expression space of the content. Content that causes a person to wake up may be output.

  For example, FIG. 5 is a diagram illustrating a third example of input and output variations of the information providing apparatus according to the embodiment. For example, as illustrated in FIG. 5A, the information providing apparatus 10 maps the features of past contents to the distributed expression space instead of the literature, so that the common analogy of old contents and contents in different fields is mapped. Generate a distributed representation space containing.

  Subsequently, as shown in FIG. 5B, when the information providing apparatus 10 receives content such as sound or image as input information, the information providing apparatus 10 identifies the field to which the content belongs. Further, the information providing apparatus 10 extracts a plurality of features included in the content in the identified field, and converts the extracted plurality of features into a distributed representation. In addition, the information providing apparatus 10 extracts content similar to the sum of features converted into the distributed representation from content in other fields, so that, as shown in FIG. Among the analogies possessed by the content, an analogy having a concept close to the concept possessed by the content analogy in the specified field is extracted. As a result, the information providing apparatus 10 can output a new content hint as shown in FIG.

  The hint output in this manner is a concept similar to the concept of the content in the field to which the input information belongs, and it is predicted that the user can recall the concept of the content in a different field. For this reason, the information providing apparatus 10 utilizes content obtained by developing the idea of old content in a new environment as shown in (E) of FIG. 5 or content in a different field as shown in (F) of FIG. Can help create new content. Further, as shown in FIG. 5G, for example, when the output is a sound, the information providing apparatus 10 can assist in creating a phrase that is a source of composition.

  In addition, for example, the information providing apparatus 10 causes the user to generate serendipity by using biometric information such as the five senses acquired from the user instead of words and using a distributed expression space in which various biometric information is expressed in a distributed manner. Information that serves as a hint may be output.

  For example, FIG. 6 is a diagram illustrating a fourth example of input and output variations of the information providing apparatus according to the embodiment. For example, as shown in FIG. 6A, the information providing apparatus 10 maps old biometric information or different types of information by mapping features of biometric information acquired in the past to a distributed expression space instead of the literature. A distributed expression space including a common analogy (for example, a pattern) included in the biometric information is generated.

  Subsequently, as shown in FIG. 6B, the information providing apparatus 10 specifies the type of the biological information of the user from the input information. In addition, the information providing apparatus 10 extracts features included in a plurality of specified types of biological information, and converts the extracted features into a distributed representation. Further, the information providing apparatus 10 extracts biometric information having features similar to the sum of the features converted into the distributed representation from other types of biometric information, and as shown in FIG. The analogy having a concept close to the concept of the analogy of the specified type of biological information is extracted from the analogies of the different types of biological information. As a result, the information providing apparatus 10 can output a hint for inducing new biological information, as shown in FIG.

  As shown in FIG. 6F, the hint output in this way can be used, for example, for specifying a comfortable sensation for the user who is the source of the biological information. For example, the information providing apparatus 10 can output a color, a pitch, or the like that the user likes when the biological information acquired from the user has a taste that the user likes.

  In order to execute the above-described processing, it is necessary to map content other than words on the distributed expression space. Therefore, the information providing apparatus 10 extracts features of contracts, business books, various contents, and biometric information using a technique such as neural network or deep learning, and distributes the vectors representing the extracted features. What is necessary is just to map on space. That is, the information providing apparatus 10 applies the above-described extraction process to arbitrary information as long as the information can be mapped onto the distributed expression space and can be classified into a plurality of types. be able to.

[3. Configuration of information providing device]
Next, the configuration of the information providing apparatus 10 that executes the extraction process illustrated in FIG. 1 will be described with reference to FIG. In the following description, words included in a plurality of documents are mapped on the distributed expression space, and are words in a field different from the field of technology to which the user's remark belongs, and in the field to which the user remark belongs. An example of the information providing apparatus 10 that outputs a word having a concept similar to the idea concept as a hint will be described.

  FIG. 7 is a diagram illustrating an example of a functional configuration of the information providing apparatus according to the embodiment. As shown in FIG. 7, the information providing device 10 is connected to an input device 30 and an output device 31. In addition, the information providing apparatus 10 includes a communication unit 11, a storage unit 12, and a control unit 16.

  The communication unit 11 is realized by, for example, a NIC (Network Interface Card). The communication unit 11 is connected to an input device 30 such as a microphone and a keyboard and an output device 31 such as a monitor, a printer, and a robot that can utter voice, and transmits and receives various types of information.

  The storage unit 12 is realized by, for example, a semiconductor memory device such as a RAM (Random Access Memory) or a flash memory, or a storage device such as a hard disk or an optical disk. The storage unit 12 includes a document database 13, a distributed expression space database 14, and an extended word database 15.

  Documents belonging to various fields are registered in the document database 13. For example, in the document database 13, documents including concepts of ideas made in the past, such as published patent publications, published patent publications, patent publications, and papers, are registered.

  In the distributed expression space database 14, distributed expressions of words belonging to various fields are registered. For example, in the distributed expression space database 14, vectors indicating words included in documents registered in the document database 13 are registered. In addition, since the distributed representation of each word is generated based on the relationship between each word, the direction and distance between each word correspond to the concept and co-occurrence similarity of each word. Become.

  The extended word database 15 is registered with operational verbs used when expanding word groups. For example, in the extended word database 15, an operational verb list including the operational verbs of Kobak & Baknar is registered in advance.

  The control unit 16 is realized by, for example, a CPU (Central Processing Unit), an MPU (Micro Processing Unit), or the like executing various programs stored in the storage device inside the information providing apparatus 10 using the RAM as a work area. Is done. Moreover, the control part 16 is implement | achieved by integrated circuits, such as ASIC (Application Specific Integrated Circuit) and FPGA (Field Programmable Gate Array), for example.

  As shown in FIG. 7, the control unit 16 includes a reception unit 17, a field identification unit 18, a pattern extraction unit 19, a related word extraction unit 20, a word space limitation unit 21, a word space expansion unit 22, a suggested word extraction unit 23, An output unit 24 and a learning unit 25 are included.

  The reception unit 17 receives a user's remarks as input information. For example, the reception unit 17 acquires a user's remarks from the input device 30 realized by a microphone, a keyboard, or the like. In such a case, the reception unit 17 converts the received user's remarks into text data. Then, the accepting unit 17 outputs the converted text data to the field specifying unit 18.

  The field specifying unit 18 specifies the field to which the input information belongs. For example, the field specifying unit 18 performs a morphological analysis of the text data received from the receiving unit 17 and specifies in which field the word included in the text data is used. Then, the field specifying unit 18 notifies the pattern extracting unit 19 of the specified field.

  The pattern extraction unit 19 extracts a word set including a plurality of words having a predetermined relationship in the specified field. For example, when receiving the notification of the field from the field specifying unit 18, the pattern extracting unit 19 extracts a document belonging to the notified field from the document database 13. Then, the pattern extraction unit 19 identifies a sentence having a predetermined structure from the extracted document, extracts a plurality of words included in the specified sentence in a predetermined order, and extracts a word set including the extracted plurality of words. Output to the related word extraction unit 20.

  For example, FIG. 8 is a diagram for explaining an example of words extracted by the information providing apparatus according to the embodiment. As shown in FIG. 8A, the sentence indicating the feature of the idea is “(noun) {is} (state expression word, adverb phrase) {, by} (verb) {do, can} It is predicted to have a structure of “ Therefore, the pattern extraction unit 19 extracts a document belonging to the notified field from the document database 13 and uses a technique such as morphological analysis, and “(noun) {a} (state expression word, adverb phrase) { Then, a sentence having a structure of “} (verb) {do, can do}” is extracted. Then, as shown in FIG. 8B, the pattern extraction unit 19 sets the noun included in the extracted sentence as the word # 1, the state expression word or adverb phrase as the word # 2, and the verb as the word # 3. Extraction is performed, and a word set including the extracted words # 1 to # 3 is generated. As shown in FIG. 8C, the word set generated in this way includes the feature of the idea indicated by the extracted sentence, that is, the concept cε.

  Returning to FIG. 7, the description will be continued. The related word extraction unit 20 selects a word set that is a concept that is more difficult for the user to think out of the word sets extracted by the pattern extraction unit 19 and that is easy to understand when presented. For example, FIG. 9 is a diagram illustrating an example of a word set extracted by the information providing apparatus according to the embodiment. In the example shown in FIG. 9, a plurality of published patent publications are subjected to morphological analysis, and word sets extracted from sentences having a predetermined structure, that is, word sets including a plurality of words having a predetermined relationship, In order of appearance.

  For example, as shown in FIG. 9A, the literature contains many word sets such as “program creation realization”, “trellis structure degeneration calculation”, and the like. However, since such a word set is a word set that is often used in a specified field or a stable word set, the idea concept of such a word set is commonly used in that field. Is a concept. As a result, word groups with high appearance frequency have low serendipity.

  On the other hand, as shown in FIG. 9B, a word set such as “wearing and wearing glasses” is not so frequently used in the specified field. It is predicted that the word set is highly likely to cause serendipity. In addition, as shown in FIG. 9C, a word set with an appearance frequency that is too low, that is, a word set with an appearance frequency lower than a predetermined threshold, is difficult to understand, and causes serendipity to the user. The possibility of making it low.

  Therefore, the related word extraction unit 20 includes the number of word sets indicating the same concept in the documents belonging to the specified field among the word sets extracted by the pattern extraction unit 19 below the first threshold value, and the second threshold value. The word group which is the above is extracted. As a result, the related word extraction unit 20 can extract a word set including a concept that is likely to cause serendipity in the specified field, that is, the field to which the user's speech belongs.

  Note that the related word extraction unit 20 uses a related word dictionary that indicates a related relationship between words, and a relationship that exists between any word included in the word set and another word included in the same word set. Count the number of words. And the related word extraction part 20 may extract the word group containing the concept which is easy to raise | generate a serendipity by extracting the word group which the number of the related words counted falls in the predetermined range.

  Returning to FIG. 7, the description will be continued. The word space limiting unit 21 further limits the distributed expression space formed by each word included in the word set extracted by the related word extracting unit 20. Specifically, the word space limiting unit 21 counts the number of documents including a sentence including the word set extracted by the related word extraction unit 20 among all the documents stored in the document database 13. Then, the word space limiting unit 21 calculates a value obtained by dividing the counted value by the number of documents, that is, a DF of the word set, and extracts a word set whose calculated DF value is equal to or less than a predetermined threshold. As a result, the information providing apparatus 10 can extract a word set including a concept that is more likely to cause serendipity based on all fields.

  The word space expansion unit 22 generates a plurality of word sets in which verbs included in the word set are replaced with predetermined operational verbs. As a result, the word space expanding unit 22 expands the distributed expression space limited by the word space limiting unit 21 without greatly changing the concept included in the word set.

  For example, FIG. 10 is a diagram for explaining an example of a process for expanding the distributed representation space executed by the information providing apparatus according to the embodiment. For example, the word space expanding unit 22 acquires a word set such as “icon click display” as the word set extracted by the word space limiting unit 21. In such a case, the word space expansion unit 22 registers the verb “display” included in the word group in the expanded word database 15 as shown in FIG. 10B, as shown in FIG. A word set is generated by replacing each word in the operational verb list.

  As a result, for example, as shown in FIG. 10C, the word space expansion unit 22 newly generates word sets such as “increase icon click”, “divide icon click”, and “remove icon click”. The word space expansion unit 22 also outputs the original word set “icon click display” to the suggested word extraction unit 23 as a newly generated word set.

  Thus, the word space limiting unit 21 and the word space expanding unit 22 calculate a value obtained by dividing the number of sentences including the word set extracted by the related word extracting unit 20 by the number of sentences in all fields. Then, a new word set is generated by combining a word set whose calculated value is smaller than a predetermined threshold and an operational verb. Since the newly generated word set is a word set obtained by converting the verb of the original word set, when mapped on the distributed expression space, the vector will indicate a direction similar to the original word set. Although predicted, such vectors are predicted to represent similar concepts. As a result, even when the number of extracted word pairs is small, the information providing apparatus 10 includes a word set including a concept similar to the word set extracted by the word space limiting unit 21, that is, a concept that easily causes serendipity. Can be generated.

  In addition, when the information providing apparatus 10 processes information other than words, the word space expansion unit 22 indicates a predetermined operation in the same manner as an operational verb for a set of information having a predetermined relevance. A plurality of new information sets may be generated by combining information. In addition, the word space expansion unit 22 may generate a set in which information of a predetermined type is changed to information indicating a predetermined operation among information included in a set of information having a predetermined relevance.

  Returning to FIG. 7, the description will be continued. The suggested word extraction unit 23 extracts words having the same concept as the concept generated by the extracted word set from words belonging to a field different from the identified field. For example, the suggested word extraction unit 23 inputs each word included in the word set generated by the word space expansion unit 22 to W2V, and acquires a distributed representation of each word. Subsequently, the suggested word extraction unit 23 calculates the sum of the acquired distributed expressions, that is, the distributed expression of the word set. Then, the proposed word extraction unit 23 is a distributed expression in a direction similar to the distributed expression of the word set among the distributed expressions stored in the distributed expression space database 14, and is separated from the distributed expression of the word set. Identify distributed representations. In other words, the proposed word extraction unit 23 uses a distributed representation of words included in a word set to determine a word having a concept that satisfies a predetermined condition for similarity to the concept generated by the word set as a field to which the input information belongs. Extract words from different fields. More specifically, the proposed word extraction unit 23 extracts a word corresponding to a distributed expression whose similarity to the distributed expression of the word set is equal to or less than a predetermined threshold.

  For example, FIG. 11 is a diagram for explaining the limitation and expansion of the distributed representation space executed by the information providing apparatus according to the embodiment. For example, as shown in FIG. 11A, the related word extraction unit 20 extracts word groups # 1 to #N as word groups including a concept that easily causes serendipity. In this case, as shown in FIG. 11B, the word space limiting unit 21 calculates the DF values of the word groups # 1 to #N for all the documents stored in the document database 13, and calculates them. The word set is filtered by the DF value. As a result, the word space limiting unit 21 extracts, for example, word groups # 1, # 3, #N, etc. as word groups whose DF value is equal to or less than a predetermined threshold, as shown in FIG. To do.

  Further, as shown in FIG. 11D, the word space expansion unit 22 expands the word set using the Kobak & bacnar checklist. For example, the word space expansion unit 22 generates word groups # 1-1 to # 1-m in which the verbs included in the word group # 1 are replaced with the operational verbs included in the Kovark & Bakunal checklist. Similarly, the word space expansion unit 22 generates word sets # 3-1 to # 3-m in which the verbs included in the word set # 3 are replaced with operational verbs, and operates the verbs included in the word set #N. The word groups # N-1 to #Nm replaced with the target verbs are generated.

  Subsequently, as shown in FIG. 11E, the proposed word extraction unit 23 sets each word group # 1-1 to # 1-m, # 3-1 to # 3-m, # N-1 to ##. The following processing is executed for each N−m. That is, the suggested word extraction unit 23 inputs a word included in the word set to W2V, acquires a vector, is a word similar to the sum of the acquired vector, and belongs to a field to which the user's remark belongs, that is, a specific word The word of the field different from the selected field is acquired from the distributed expression space.

  Hereinafter, an example of processing in which the proposed word extraction unit 23 acquires words from a field different from the identified field that are similar to the sum of the vectors of the word sets from the distributed expression space will be described with reference to the drawings. For example, FIG. 12 is a diagram illustrating an example of a process of extracting similar words by the information providing apparatus according to the embodiment. For example, as shown in FIG. 12A, the suggested word extraction unit 23 acquires the word set “view wearing glasses”. In such a case, as shown in FIG. 12B, the proposed word extraction unit 23 inputs each word included in the word set, that is, “glasses”, “wear”, and “view” to W2V. Compute a distributed representation of each word, ie a vector.

  Subsequently, as shown in FIG. 12C, the proposed word extraction unit 23 extracts a vector similar to the sum of the vectors of each word in the word set from the distributed representation space database 14, and the extracted vector indicates Get a word. For example, as shown in FIG. 12D, the suggested word extraction unit 23 acquires a word indicated by a vector having a large cosine distance value from the word set vector.

  For example, in the example shown in FIG. 12D, “age”, “spouse”, “appearance”, “acting”, “optimization”, “moderator”, “real estate_owned”, etc., in descending order of the cosine distance value. Is acquired. Here, “real estate_owned” is a word extracted as one word by morphological analysis when the word “real estate” and the word “owned” are used as one idiom by being consecutive.

  Here, it can be said that a word having a large cosine distance value is a word having a concept similar to the concept indicated by the word set. However, a word simply having a large cosine distance value is not intended to provide a concept that the user has not predicted, so it is predicted that the serendipity is low. On the other hand, a word with a small cosine distance value includes a word having a concept different from the concept indicated by the word set, but a word that provides a concept that the user cannot think of at first glance, that is, has a high serendipity. Words are also expected to be included.

  For example, FIG. 13 is a diagram for explaining an example of the relationship between word vectors. In the example illustrated in FIG. 13, an example of a distributed expression space in which the words “MAN”, “WOMAN”, “UNCLE”, “AUNT”, “KING”, and “QUEEN” are mapped is described. Here, as shown in FIG. 13A, a vector from the word “MAN” to the word “WOMAN”, a vector from the word “UNCLE” to the word “AUNT”, and from the word “KING” to the word “QUEEN” These vectors are considered to indicate the same concept, such as “add the concept of a woman excluding the concept of a man from the word”. For this reason, as shown in FIG. 13B, the directions of the arrows of such vectors coincide.

  On the other hand, when the distance between the word “KING” and the word “UNCLE” is long, the word “UNCLE” cannot always be easily recalled from the word “KING”. The co-occurrence of the concept indicated by the word “KING” and the concept indicated by the word “UNCLE” is considered to be small. However, even in such a concept with low co-occurrence, a similar concept can be shown when the vector direction is facing, so that serendipity can occur as shown in FIG. It is considered that the nature is high.

  For this reason, the information providing apparatus 10 is a word having a concept similar to a concept of an idea having a high co-occurrence with a word included in the user's utterance as a word having a high possibility of causing serendipity to the user. Thus, it is only necessary to provide a word having low co-occurrence from words included in the user's utterance. Here, a word with low co-occurrence is a word whose vectors are separated from each other. For example, a word belonging to a different field is predicted to be applicable.

  Therefore, the proposed word extraction unit 23 calculates the distance between the vector of the word set and a vector indicating another word, and identifies a vector whose calculated distance is more than a predetermined threshold. The word identified in this way is a word including a concept similar to the concept of an idea having high co-occurrence with the user's utterance, and is predicted to be a word having low co-occurrence with the user's utterance. Is done.

  For example, FIG. 14 is a diagram for explaining an example of a concept that a word proposed by the information providing apparatus has. As shown in FIG. 14, the words # 1 to # 3 constituting the word set # 1 belonging to the field # 1 have concepts # 1 to # 3, respectively. Here, concept # 1 is formed by element group # 1 which is a plurality of elements forming the concept, and concept # 2 is formed by element group # 2 which is a plurality of elements forming the concept, and concept # 3. Is formed by an element group # 3 which is a plurality of elements forming the concept. For this reason, the concept of the word set # 1 is considered to be the sum of the concepts # 1 to # 3, that is, the sum of the element groups # 1 to # 3.

  Here, as shown in FIG. 14A, the possibility of causing serendipity increases as the vector distance increases. However, if the distance is simply separated, it is too far from the concept of the word set # 1, so that it is difficult for serendipity to occur. Therefore, as shown in FIG. 14B, the proposed word extraction unit 23 has a word that has a part in common with the word set # 1 to such an extent that it can be guaranteed that it is meaningful enough to generate serendipity. To extract. More specifically, the proposed word extraction unit 23 is a word having a concept # 4 formed by an element group # 4 similar to the concept of the word group # 1, that is, the sum of the element groups # 1 to # 3. Thus, the word # 4 belonging to the field # 2 that is sufficiently similar to the field # 1 is extracted.

  For example, the suggested word extraction unit 23 extracts a vector whose cosine distance from a word set vector is greater than a predetermined threshold (or falls within a predetermined range), and the extracted vector and each word included in the word set The cosine distance of each word vector is calculated. Then, the suggested word extraction unit 23 specifies a vector in which the sum of the calculated cosine distances is smaller than a predetermined threshold, and sets the word indicated by the specified vector as a suggested word. That is, the proposed word extraction unit 23 has a distributed representation in which the cosine distance with the sum of the distributed representations of the word set is larger than a predetermined threshold and the sum of the cosine distances with the distributed representation of each word is smaller than the predetermined threshold. The word corresponding to is extracted.

  In this way, the suggested word extraction unit 23 calculates the sum of the distributed representations of the words included in the word set, and calculates the word corresponding to the distributed representation whose similarity with the calculated sum of the distributed representations satisfies the predetermined condition. Extract. For example, the suggested word extraction unit 23 is a distributed representation having the same direction as the sum of the distributed representations of words included in the word set (that is, the vector of the word sets), and is a distributed representation of the words included in the word set. A word corresponding to the distributed expression whose distance from the sum is equal to or greater than a predetermined threshold is extracted.

  For example, when the proposed word extracting unit 23 receives a word set such as “graph collective addition”, it calculates the sum of the word “graph” vector, the word “collective” vector, and the word “add” vector. . Next, the suggested word extraction unit 23 specifies a vector of the word “rotation” as a vector whose cosine distance with the calculated vector sum is larger than a predetermined threshold. In this case, the proposed word extraction unit 23 performs cosine distance between the vector of the word “rotation” and the vector of the word “graph”, the cosine distance between the vector of the word “rotation” and the vector of the word “collective”, and the word “rotation”. The cosine distance between each vector and the vector “add” is calculated, and it is determined whether or not the sum of the calculated cosine distances is smaller than a predetermined threshold.

  That is, the suggested word extraction unit 23 determines whether or not the distance from each word vector constituting the word set vector is smaller than a predetermined threshold. Then, the proposed word extraction unit 23 proposes the word “rotation” when it is determined that the distance between the word “rotation” vector and each word vector constituting the word set vector is smaller than a predetermined threshold. A word.

  Returning to FIG. 7, the description will be continued. The output unit 24 outputs the word extracted by the suggested word extraction unit 23 as a proposal. For example, when the proposed word extraction unit 23 extracts the word “rotation”, the output unit 24 generates a sentence that makes a proposal such as “How is it rotated?” And transmits the generated sentence to the output device 31. To do. As a result, the output device 31 can output a word extracted by the information providing device 10, that is, a word that causes a serendipity to the user.

  The learning unit 25 learns the distributed expression space stored in the distributed expression space database 14. For example, the learning unit 25 performs morphological analysis of each document included in the document database 13 and learns a distributed expression based on the relationship between words included in each document using the W2V technology. Then, the learning unit 25 registers the learning result in the distributed representation space database 14.

[4. Flow of processing executed by information providing apparatus 10]
Next, the flow of processing executed by the information providing apparatus 10 will be described with reference to FIGS. First, the flow of extraction processing executed by the information providing apparatus 10 will be described with reference to FIG. FIG. 15 is a flowchart illustrating a flow of extraction processing executed by the information providing apparatus according to the embodiment.

  As illustrated in FIG. 15, when the information providing apparatus 10 acquires input information (step S101), the information providing apparatus 10 specifies a field to which the acquired input information belongs (step S102). Subsequently, the information providing apparatus 10 extracts a word set satisfying a predetermined relationship from the document data belonging to the specified field (step S103). That is, the information providing apparatus 10 extracts a word set from a specific syntax having a cε dictionary structure. Then, the information providing apparatus 10 selects a word group that can cause serendipity from the extracted word group (step S104), and the concept and similarity generated by the word group satisfy a predetermined condition using the distributed representation of the word group. A word is extracted (step S105). That is, the information providing apparatus 10 extracts words that belong to a field other than the field to which the word group belongs and have a concept similar to the concept of the word group. And the information provision apparatus 10 outputs the extracted word as a hint (step S106), and complete | finishes a process.

  Next, referring to FIG. 16, the flow of various processes for extracting a word that belongs to a field other than the field to which the word group belongs and has a concept similar to the concept of the word group will be described more specifically. explain. FIG. 16 is a flowchart for explaining a specific processing flow of the extraction processing executed by the information providing apparatus according to the embodiment. Note that steps S201 to S206 shown in FIG. 16 are more specific processes of steps S103 to S105 shown in FIG.

  For example, the information providing apparatus 10 performs morphological analysis of document data belonging to the field to which the input information belongs (step S201), and specifies a character string that matches a predetermined pattern (step S202). And the information provision apparatus 10 produces | generates a word set from the word contained in the specified character string (step S203).

  Further, the information providing apparatus 10 selects a word group that is likely to cause serendipity from the generated word group (step S204), and generates a variation of the word group from the selected word group using an operational verb (step S205). Then, the information providing apparatus 10 inputs the selected word set to W2V, and the degree of similarity including the direction, that is, the degree of similarity not including the direction when the cosine distance with the vector of the word set is within a predetermined range. That is, a word whose sum of cosine distances with a vector of words included in the word set is equal to or less than a predetermined threshold is extracted (step S206).

  Next, the flow of processing for selecting a word group that is likely to cause serendipity from the word group and generating a variation of the word group from the selected word group will be described more specifically with reference to FIG. FIG. 17 is a flowchart for explaining a specific processing flow of processing for selecting a word group that is likely to cause serendipity by the information providing apparatus according to the embodiment. Note that steps S301 to S303 shown in FIG. 17 are more specific processes of steps S204 and S205 shown in FIG.

  For example, the information providing apparatus 10 calculates each DF of the generated word set (step S301). Further, the information providing apparatus 10 selects a word set in which the DF value of the generated word set is equal to or less than a predetermined threshold (step S302). And the information provision apparatus 10 produces | generates the variation of the selected word group using an operational verb list | wrist (step S303).

[5. (Modification)
In the above, an example of the extraction process performed by the information providing apparatus 10 has been described using the aspect illustrated in FIG. However, the embodiment is not limited to this. Hereinafter, the variation of the extraction process which the information provision apparatus 10 performs is demonstrated.

[5-1. (Various parameters)
In order to extract words that can cause serendipity, for example, the information providing apparatus 10 described above extracts word pairs having a DF value of a predetermined threshold value (hereinafter referred to as a first threshold value) or less from the extracted word sets. Extracted. Further, the information providing apparatus 10 is a vector whose cosine distance from a word set vector is within a predetermined range (hereinafter referred to as a first range), A vector in which the sum of the cosine distances is equal to or less than a predetermined threshold (hereinafter referred to as a second threshold) was extracted. Here, as the various threshold values employed by the information providing apparatus 10, any threshold value can be employed.

  For example, when the information providing apparatus 10 outputs a word that is difficult to understand at first glance but is prone to serendipity when considered, the value of the first threshold may be set lower. May be made smaller. Further, the information providing apparatus 10 may set the first range to be narrower.

[5-2. (About word sets that are prone to serendipity)
In the example described above, the information providing apparatus 10 selects a word that is highly likely to cause serendipity from the extracted word set in order to increase the possibility of causing serendipity. For example, the related word extraction unit 20 selects a word set based on the appearance frequency of each word set, the relationship between words, and the like. However, the embodiment is not limited to this.

  For example, the information providing apparatus 10 may select a word group that can cause serendipity in advance for each field. Such selection processing may be performed in advance by a human hand. For example, the information providing apparatus 10 extracts a word set including a plurality of words having a predetermined relationship from the document data, and presents the extracted word set to the operator. And the information provision apparatus 10 may select beforehand the word set which the operator selected as a word set which can raise serendipity.

[5-3. Processing performed by information providing device]
In the above description, the information providing apparatus 10 specifies a plurality of pieces of information having a predetermined relationship from the information of the field to which the input information belongs, and specifies information having the same concept as that generated by the specified pieces of information. Extracted from information belonging to a different field. However, the embodiment is not limited to this. That is, the information providing apparatus 10 can output information having an explicitly discontinuous relationship while maintaining an implicit connection with the input information from information in a field different from the field to which the input information belongs. If so, the output information may be extracted by different processing.

  For example, FIG. 18 is a diagram illustrating an example of a process for outputting information related to information in an input field and information in a different field. In the example shown in FIG. 18, as a variation of the process executed by the information providing apparatus 10, both the information on the field to which the input information belongs and the information on the field different from the field to which the input information belongs are analyzed. An example of processing for outputting information relevant to information was described.

  For example, as shown in FIG. 18, when a user thinks of a new idea, fields to which information that may be helpful belongs include education, power supply, banking, and the like. For this reason, as shown in FIG. 18A, it is difficult for the user to consider a new idea in consideration of information in all these fields.

  Therefore, the information providing apparatus 10 executes the following process. First, the information providing apparatus 10 receives input information from a user. In such a case, the information providing apparatus 10 specifies the field to which the input information belongs. Further, the information providing apparatus 10 extracts a field (different field) different from the identified field. Then, the information providing apparatus 10 analyzes both information belonging to the field to which the input information belongs and information belonging to a field different from the field, and specifies information relevant to both information. Then, the information providing apparatus 10 outputs the specified information.

  For example, as shown in FIG. 18B, the user inputs arbitrary information such as a word that has come up to the information providing apparatus 10 as input information. In such a case, the information providing apparatus 10 identifies the field to which the input information belongs, as shown in (C) of FIG. For example, if the input information is information such as “propane gas”, the information providing apparatus 10 identifies “gas supply industry” as the field to which the input information belongs.

  Subsequently, the information providing apparatus 10 extracts a field different from the identified field. For example, the information providing apparatus 10 extracts “advertisement industry” as a field different from the identified field “gas supply industry”. The information providing apparatus 10 may select a plurality of fields different from the field to which the input information belongs. Further, the information providing apparatus 10 may select a field that has a meaning that is different from the field to which the input information belongs (for example, a field that is far in the distributed expression space) by using the various processes described above. .

  Then, the information providing apparatus 10 analyzes information on both the field to which the input information belongs and the extracted field, and specifies information related to both information. For example, the information providing apparatus 10 compares the attributes, concepts, meanings, and the like of words belonging to each field using the distance and direction between words in the distributed expression space, and identifies information related to both fields. Further, for example, the information providing apparatus 10 extracts words that are close in direction and long in the distributed expression space. And the information provision apparatus 10 outputs the extracted word, as shown to (F) in FIG.

  As a result of executing such a process, the information providing apparatus 10 can output a combination of words that does not appear in normal thinking, so that serendipity can be generated for the user. In addition to the processing described above, the information providing apparatus 10 can employ any analysis method as long as it outputs information related to both the field to which the input information belongs and the different field.

  Note that the information providing apparatus 10 described above can employ arbitrary information as input information. For example, the information providing apparatus 10 employs text, business features, contents such as sound and images, biometric information, and the like as input information, and is related to information in both the field to which the input information belongs and the different field. As information, text, business characteristics, contents, biometric information, and the like may be output. In addition, the information providing apparatus 10 can be used within a range in which arbitrary processes are not contradicted among the writing processes described in the above-described embodiments.

[5-4. Others]
In addition, among the processes described in the above embodiment, all or part of the processes described as being automatically performed can be performed manually, or the processes described as being performed manually can be performed. All or a part can be automatically performed by a known method. In addition, the processing procedures, specific names, and information including various data and parameters shown in the document and drawings can be arbitrarily changed unless otherwise specified. For example, the various types of information illustrated in each drawing is not limited to the illustrated information.

  Further, each component of each illustrated apparatus is functionally conceptual, and does not necessarily need to be physically configured as illustrated. In other words, the specific form of distribution / integration of each device is not limited to that shown in the figure, and all or a part thereof may be functionally or physically distributed or arbitrarily distributed in arbitrary units according to various loads or usage conditions. Can be integrated and configured. For example, the word space limiting unit 21 and the word space extending unit 22 illustrated in FIG. 7 may be integrated.

  In addition, the above-described embodiments can be appropriately combined within a range in which processing contents do not contradict each other.

[5-5. program〕
Further, the information providing apparatus 10 according to the above-described embodiment is realized by, for example, a computer 1000 configured as shown in FIG. FIG. 19 is a hardware configuration diagram illustrating an example of a computer that implements extraction processing. The computer 1000 includes a CPU 1100, RAM 1200, ROM 1300, HDD 1400, communication interface (I / F) 1500, input / output interface (I / F) 1600, and media interface (I / F) 1700.

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

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

  The CPU 1100 controls an output device such as a display and a printer and an input device such as a keyboard and a mouse via the input / output interface 1600. The CPU 1100 acquires data from the input device via the input / output interface 1600. In addition, the CPU 1100 outputs the generated data to the output device via the input / output interface 1600.

  The media interface 1700 reads a program or data such as an information providing program stored in a recording medium 1800 that is an example of a non-transitory computer-readable storage medium, and provides the read program or data to the CPU 1100 via the RAM 1200. The CPU 1100 loads such a program from the recording medium 1800 onto the RAM 1200 via the media interface 1700, and executes the loaded program. The recording medium 1800 is, for example, an optical recording medium such as a DVD (Digital Versatile Disc) or PD (Phase change rewritable disk), a magneto-optical recording medium such as an MO (Magneto-Optical disk), a tape medium, a magnetic recording medium, or a semiconductor memory. Etc.

  For example, when the computer 1000 functions as the information providing apparatus 10 according to the embodiment, the CPU 1100 of the computer 1000 implements the function of the control unit 16 by executing a program loaded on the RAM 1200. The HDD 1400 stores data in the storage unit 12, that is, the document database 13, the distributed expression space database 14, and the extended word database 15. The CPU 1100 of the computer 1000 reads these programs from the recording medium 1800 and executes them. However, as another example, these programs may be acquired from other devices.

[6. effect〕
As described above, when receiving the input information, the information providing apparatus 10 specifies a specific syntax from the input information and extracts a word group embedded in the specified specific syntax. In addition, the information providing apparatus 10 calculates the degree of similarity with the distributed expression of other words using the distributed expression of the word group included in the specific syntax. Then, the information providing apparatus 10 outputs information extracted based on the calculated similarity. For this reason, the information providing apparatus 10 can provide information that does not come to the mind of the user, and thus can output information that can cause serendipity to the user.

  In addition, the information providing apparatus 10 specifies a specific syntax using the cε dictionary method. For this reason, the information providing apparatus 10 can output a word having a feature similar to the feature of the word group from the word group included in the sentence having a pattern that has the possibility of causing serendipity in a certain field. Therefore, it is possible to provide information that is hard to come up with than the user while ensuring the possibility of causing serendipity for the user.

  Moreover, the information provision apparatus 10 has the analogy similar to a word group based on the equivalent conversion theory, and outputs the word which belongs to the field | area different from this word group. Since it is possible to provide information that is more difficult for the user to remember while maintaining the concept generated by the word group, it is possible to output information that can generate serendipity for the user.

  Further, the information providing apparatus 10 corresponds to a word corresponding to a distributed expression similar to the sum of the distributed expressions of the word group or a distributed expression in which the similarity to the sum of the distributed expressions of the single tool group falls within a predetermined range. Extract information. Therefore, the information providing apparatus 10 can output information that can cause the user to generate serendipity.

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

  Moreover, the above-mentioned “section (module, unit)” can be read as “means”, “circuit”, and the like. For example, the control unit can be read as control means or a control circuit.

DESCRIPTION OF SYMBOLS 10 Information provision apparatus 11 Communication part 12 Storage part 13 Reference database 14 Distributed expression database 15 Extended word database 16 Control part 17 Reception part 18 Field specific part 19 Pattern extraction part 20 Related word extraction part 21 Word space limitation part 22 Word space expansion part 23 Proposed word extraction unit 24 Output unit 25 Learning unit 30 Input device 31 Output device

Claims (7)

A reception unit for receiving input information;
A pattern identifying unit that extracts a specific syntax and a group of words embedded in the specific syntax from input information;
A similarity calculation unit that calculates a similarity using a distributed representation of the specified word group included in the specified pattern;
And an output unit that outputs information extracted based on the similarity calculated by the similarity calculation unit. The information providing apparatus according to claim 1, wherein the pattern specifying unit uses a cε dictionary method. 3. The information provision according to claim 1, wherein the output unit outputs a word having an analogy similar to that of the word group and belonging to a field different from the word group based on an equivalent conversion theory. apparatus. The information providing apparatus according to claim 1, wherein the output unit outputs a word for a distributed expression similar to the sum of the distributed expressions of a word group. 5. The output unit according to claim 1, wherein the output unit outputs a word corresponding to a distributed expression whose similarity with the sum of the distributed expressions of the word group falls within a predetermined range. Information provision device. A reception process for receiving input information;
A pattern identification step of extracting a specific syntax and a group of words embedded in the specific syntax from input information;
A similarity calculation step of calculating a similarity using a distributed representation of the specified word group included in the specified pattern;
An output step of outputting information extracted based on the similarity calculated by the similarity calculation step. Acceptance procedure to accept input information,
A pattern identification procedure for extracting a specific syntax and a group of words embedded in the specific syntax from input information;
A similarity calculation procedure for calculating a similarity using a distributed representation of the specified word group included in the specified pattern;
An information providing program for causing a computer to execute an output procedure for outputting information extracted based on the similarity calculated by the similarity calculation procedure.

Images