JP7037778B2 - Search device and method - Google Patents

Description

The present invention relates to a search technique, and more particularly to a search technique suitable for performing a search using words with a high degree of abstraction.

In information retrieval, for the frequency of appearance of features related to the search target (for example, words contained in explanatory texts and related articles, features obtained from the results of physical analysis of music and video). In many cases, characteristic words for the search target category (for example, travel, fashion, cosmetics, music, etc.) are obtained from statistical information and added as information tags for search. As a method for extracting a characteristic word, TF-IDF (Trm Frequency-Inverse Division Frequency), LDA (Latent Dirichlet Allocation), or the like is used.

However, words with a high degree of abstraction and words composed of multiple words (hereinafter referred to as phrases) rarely appear in the text and may be leaked from the tags, so search using words with a high degree of abstraction. It is often difficult to do.

In a general information retrieval site, an information tag indicating the content of the content is added to each content provided by the service operator, and the content content that completely or partially matches the search condition word specified by the user is added. Present. When displaying the above information on the information retrieval site by performing statistical analysis of the tags of the contents selected by the user against the history of the contents selected by the user, and considering the bias of the conditions selected by the user as the user's preference information. By setting the filter condition of, optimization (personalization) for each user may be performed.

FIG. 1 is an example of a search system using an information tag. In this example, an information tag as shown in FIG. 2 is added to each content related to travel information. A search is performed using this tag and user preference information.

Content provided as a service includes products, articles, songs, images, videos, etc. By adding tags that represent the characteristics of each content, display, search, user preference analysis, etc. for each content category, etc. It is used in. As the tag, for example, when the content type is "cosmetics", there is a tag as shown in FIG.

By the way, the contents of these differ depending on the content type. That is, since the content of the tag type is different for each content type such as "travel", "cosmetics", "fashion", "music", and "drinking water", there has been a problem when dealing with a plurality of types of content in a cross-cutting manner. Therefore, it is conceivable to treat a person's "sensitivity" (the user's "impression", the user's own "emotion"), which is less dependent on the content type, as common information and treat it as an information tag.

However, for example, for the same impression word such as "clean", the meanings of "cosmetics", "fashion", and "drinking water" are not exactly the same, so even if you simply search for the word "clean", you can search for it. There is a problem that the content presented as a search result does not match the user's image.

As a method of adding a tag to the content, the following method is generally used.
(1) Using a method such as morphological analysis for text information such as description of target content (for example, product), introductory article, blog, writing in posting column, writing on SNS (Social Networking Service), etc. , Words and phrases are extracted, and those that represent the characteristics of the product are managed as tags.
(2) In the case of music content, in addition to the tags obtained from text information related to the music content (for example, the content of lyrics, introductory articles, etc.), for example, the tempo of the music signal, the time-series change in the volume of the sound, and the frequency. Information related to time-series changes in spectrum components, genre information, singer's voice quality and singing characteristics are managed as tags.
(3) In the case of video content, in addition to tags obtained from text information such as articles and scenarios related to video content, color changes in the video scene, direction, impression of BGM, character settings and lines of performers, voice actors. , Information judged from multiple elements such as supervision is managed as a tag.

FIG. 4 is an example of an automatic tag attachment system. As the input text data of X1, it is assumed that the content is described, the explanation, the introductory article, the evaluation comment on the bulletin board, the comment on the content on the SNS, and the like. At the same time, it is assumed that attributes such as content type, identification ID, and content name are added. This text information can be cut out at the word level by performing preprocessing such as removing unnecessary characters such as symbols, adjusting uppercase and lowercase letters, and aligning half-width and full-width characters in X2, and then performing morphological analysis in X3. become. Here, as an example of morphological analysis, an output example by the MeCab morphological analysis engine will be described. For example, if the text "This cushion is fluffy and cute" (FIG. 5 (A)) is morphologically analyzed as the input text, the output of FIG. 5 (B) is obtained. When only "adjectives" are extracted, the words "cushion", "fluffy", and "cute" are extracted as the output of the keyword extraction X3 (FIG. 5 (C)). The user dictionary X4 may be referred to for keyword extraction.

There are various methods for determining the importance of a word in a document, such as using an LDA (Latent Dirichlet Allocation) or a vector that matches the meaning of the word. Here, an example using a method called TF-IDF will be described (https://ja.m.wikipedia.org/wiki/Tf-idf). The TF of TF-IDF represents the frequency of occurrence of words (the number of times a specific word appears in a document), and the IDF represents the frequency of reverse documents (the natural logarithm of the number of documents including documents in the entire corpus). It has the following significance.
TF: Words that appear more frequently in a document dd are more important to characterize the document IDF: Words such as "a" and "the" that generally appear in any document are not important TF -IDF represents "importance of word ww in document dd". Since this TF-IDF is an amount determined for each word w and the document dd, if TF-IDF (w, d) is used, TF-IDF (w, d) = TF (w, d) × defined by the following equation. IDF (w)
Here, in order to obtain the importance of words appearing for each content type, sentences with the same content attribute (pg) are aggregated as the same sentence and the value of TF-IDF is used.
TF-IDF (w, pg) = TF (w, pg) x IDF (w)
The importance obtained in this way indicates that the importance of the word in the content type is higher as the numerical value is larger. In FIG. 4, the importance score value of a word is calculated by generating statistical information X5 and calculating importance score value X6.

The keyword extracted from each text sentence is stored together with the relationship with the value n_count (w) of the number of times the keyword w appears in the sentence (X7). For example, it is assumed that the keywords "cushion", "fluffy", and "cute" are extracted once from the text related to "content name A". In that case, the output example of the keyword extraction X3 in FIG. 4 is as follows as shown in FIG. 6 (A).
Content name A: Cushion: 1, Fluffy: 1, Cute: 1
In the tag processing X7 of FIG. 4, the number of times of the extracted words is totaled for each content name added to the input text X1. For example, as shown in FIG. 6B, the keyword extraction result of the text input for the "content name A" different from the above extraction result is the content name A: pouch: 1, soft: 1, cute: 1, fluffy: 2.
age,
sum_count (w) = Σn_count (w)
(FIG. 6 (C)), the aggregated result in the tag processing X7 is as follows, as shown in FIG. 6 (D).
Content name A: Fluffy: 3, Cute: 2, Light: 1, Cushion: 1, Pouch: 1
Further, in the aggregation tagging process X8 of FIG. 4, words unnecessary as tags and those whose appearance count is less than a certain value are deleted from the aggregation result of the keyword related to the content name A in this way (filter processing X9), and remain. For the word w, the score value using the value of the score value weight_score (w) of the importance of the word calculated by the importance score value calculation process X6 of FIG. 4 (word_score (w) = func (weight_score (w)). , Sum_count (w)) (FIG. 6 (E)) to determine the keywords. For example, the importance score of each keyword is calculated as word_score (w) = func (weight_score (w) × sum_count (w)). (FIG. 6 (F)), a tag whose value is equal to or more than a certain value and whose number of applicable keywords is within the upper limit is added as an information tag. When adding, the tags are arranged or added in descending order of score value. The score value of each keyword may be normalized so that the total value becomes 1. The tag is output as a tag as shown in FIG. 6 (G), for example (X10).
Content name A: Fluffy: 0.53, Cute: 0.32, Light: 0.15

For example, when a word expressing an impression of the content such as "fluffy" or "cute" is added as a tag of the content, the corresponding content can be searched by combining these tag words or a plurality of tag words.

Here, consider a case where a word such as "small devil" that has an abstract and associated impression is added as a tag and the tag word is used for a search. In the tag addition method as described above, the word "small devil" exists in the text data from which the keyword is extracted, and it represents the characteristics of the product (for example, the word is detected at a certain ratio or more). ) It is added as an information tag only in the case. However, in reality, the probability that the word "small devil" will be added as an information tag for content is low. Therefore, in order to add such an abstract word as an information tag, it is necessary for the system operator to manually add a tag for each content, which is a time-consuming work.

In addition, on bulletin boards where evaluations of content can be written, the operator may prepare a list of information tag candidates, select the words to be tagged from the list, and tag them. However, in this case, since tagging is performed by the subjective judgment of the user who writes the evaluation, there is a problem that the tagging standard lacks accuracy.

It is preferable that the search can be performed using words with a high degree of abstraction, and that the recommendation information of the presented content is provided to the user for convenience when the user uses the search result.

Prior art documents related to this invention include Japanese Patent Application Laid-Open No. 2017-062717. In this prior art document, in addition to attributes such as the performers and genres of the content, metawords such as words that linguistically express the mood and memorable scenes that the customer feels are associated with the content so that the operator can respond to the customer. When searching for content using the metaword determined based on the search result, the corresponding metaword is presented together with the content as a search result to support communication between the operator and the customer.

It should be noted that the present invention should not be understood in a limited manner due to the above-mentioned problems, and the contents thereof are defined in the scope of claims and will be described in detail below with reference to examples.

JP-A-2017-062717

It is preferable to provide a search technique suitable for searching using a word with a high degree of abstraction that is unlikely to appear in the text group that is the base of the tag.

According to the present invention, in order to achieve the above-mentioned object, the configuration as described in the claims is adopted. Here, prior to explaining the invention in detail, the description of the scope of claims will be supplementarily explained.

According to one aspect of the invention, in order to achieve the above object, a search device is used: for searchable content that associates one or more primary metawords contained in a pre-registered primary metaword set with the searchable content. Primary metaword association with the primary metaword association data holding means for the content to be searched that holds the primary metaword association data; Primary metaword association data holding means for secondary metaword that holds primary metaword association data for secondary metaword; Displayable content that holds display target content in association with at least one secondary metaword included in the above secondary metaword set. Retention means; Search input means for inputting search input text; Secondary metaword extraction means for extracting one or more secondary metawords from the search input text input from the search input means; Secondary metaword extraction Related primary metaword determining means for determining the primary metaword related to the one or more secondary metawords by referring to the primary metaword association data retention means for the secondary metaword for the one or more secondary metawords extracted by the means. And; a search means for searching the search target content using the primary metaword determined by the related primary metaword determining means; and; when the search target content is searched in relation to the at least one secondary metaword, the said A display target content display means for displaying the display target content stored in the display target content holding means in relation to at least one secondary metaword; and a search target for displaying the search target content retrieved by the search means. It is configured to include content display means.

The primary metaword is a component that defines the secondary metaword. The primary metaword is a word or phrase that can be easily extracted from a corpus or the like, and often corresponds to, but is not limited to, the above-mentioned basic metaword and Kansei metaword. Secondary metawords (also called idiom metawords and idiom tags) are words and phrases with a high degree of abstraction that are difficult to extract from corpora and the like, but are not limited to these.

The primary metaword association data for the search target content associates the primary metaword with the search target content as a tag. The primary metaword association data for the secondary metaword associates the primary metaword with the secondary metaword as a tag.

In this configuration, since the display target content is associated with the secondary metaword, it is recommended to the user to view the search result by displaying the display target content in relation to the display of the search result of the secondary metaword. Guidance information such as information and hint information can be provided.

In this configuration, the display target content may be text described in relation to the secondary metaword. The display target content may be adapted and optimized according to environmental conditions such as the type of target user content, user profiling (age, gender, preference tendency), season, time zone, and weather.

Further, in this configuration, the primary metaword association data for the search target content has one or a plurality of primary metawords for each of the search target contents, and a score value related to the search target content of the primary metaword. The primary metaword association data for the secondary metaword has a plurality of primary metawords for each of the secondary metawords and a score value associated with the secondary metaword of the primary metaword; the search means is extracted. For the secondary metawords that have been created, the score values related to the secondary metawords of multiple primary metawords are retrieved by referring to the primary metaword association data for the secondary metawords, and the associated primary metawords are the primary content for the search target content. With reference to the metaword association data, the score values related to the search target content of the plurality of primary metawords are extracted, and the score values related to the secondary metawords of the plurality of primary metawords and the search target contents of the plurality of primary metawords are related. It may be configured to search the search target content based on the score value to be searched.

Further, in this configuration, the search device associates one or a plurality of secondary metawords included in the above secondary metaword set with the search target content, and holds the secondary metaword association data for the search target content for the search target content. With the following metaword association data retention means; the search is performed for the search target content using the search means for one or more secondary metaword sets included in the secondary metaword set, and the search result is used as the basis for the search result. Alternatively, the score value of the search target content is calculated for the plurality of secondary metawords, and the secondary metaword association data holding means for the search target content is used as the score value related to the one or more secondary metawords and the corresponding search target content. The search means further has a learning means to be registered in the above; the search means is the secondary metaword association data for the search target content when the learning means has already been learned for the secondary metaword extracted by the secondary metaword extraction means. It may be configured to search by referring to.

Further, in this configuration, the secondary metaword extraction means extracts one or a plurality of primary metawords in addition to the one or a plurality of secondary metawords from the search input text input from the search input means; the search. The means may be configured to search the search target content using the extracted one or a plurality of primary metawords in addition to the primary metaword determined by the related primary metaword determining means.

Further, in this configuration, at least one primary metaword of the association data for the secondary metaword held in the association data holding means for the secondary metaword has a content type, a season, a situation, and a user's age / gender. The score value may be changed depending on the parameters of the parameter group.

Further, in this configuration, the above parameters may be content types, seasons, situations, and / or user attributes.

It should be noted that the present invention can be realized not only as an apparatus or a system but also as a method. Of course, a part of such an invention can be configured as software. It is also natural that software products (computer programs) used to execute such software on a computer are also included in the technical scope of the present invention.

The above-mentioned aspects and other aspects of the present invention are described in the claims and will be described in detail below with reference to Examples and the like.

According to the present invention, it is possible to provide a search technique suitable for searching using words and phrases with a high degree of abstraction that are unlikely to appear in the text group that is the base of the tag.

It is a figure explaining the example of the search system using the conventional information tag. It is a figure explaining the example of the conventional information tag. It is a figure explaining the example of the conventional information tag by taking cosmetics as an example. It is a figure explaining an example of adding an information tag to a content. It is a figure explaining the preprocessing of the input text using the morphological analysis. It is a figure explaining the example which adds the information tag with the importance score value to the content. It is a figure explaining the schematic example of the search of an Example. It is a figure explaining the idiom metaword of an embodiment. It is a figure explaining the metaword and response text example related to a content. It is a figure explaining an example of an idiom metaword. It is a block diagram which shows the specific configuration example of an Example. It is a figure explaining the premise of the search example of an Example. It is a figure explaining the search example of an Example. It is a figure explaining the example of a metaword. It is a figure explaining the format example of a metaword. It is a figure explaining the format example of a metaword. It is a block diagram which shows the other concrete configuration example of an Example. It is a block diagram which shows the other concrete configuration example of an Example. It is a figure which shows the example of a theme word and a concrete explanation. It is a figure explaining the tag of a theme word.

Hereinafter, the search device according to the embodiment of the present invention will be described.

The search device of this embodiment adds a sensibility tag (hereinafter, sensibility metaword) for each content to a search method using a conventional tag (hereinafter, basic metaword), and uses a word with a high degree of abstraction of the user. You can also search for requests. Furthermore, for a word with a higher degree of abstraction (hereinafter referred to as idiom metaword), a response message related to it should be defined, and when presenting the content information of the search result by the idiom metaword, it should be presented as a reason for recommendation. It has become. Specifically, an idiom metaword is defined as a metaword whose semantic interpretation is represented by a set of a plurality of Kansei metawords and basic metawords (hereinafter, also referred to as metaword factors). By adopting the search method of this embodiment, as in the embodiment of FIG. 7, the idiom metaword, the sensitivity metaword, and the basic metaword are extracted even for the search request from the user with a high degree of abstraction, and the idiom metaword, the sensitivity metaword, and the basic metaword are extracted. Based on this, it will be possible to match content that crosses multiple content types.

The metaword factors (basic metaword, sensitivity metaword) included in this idiom metaword are fixedly defined as shown in Fig. 8 (A), and vary depending on the content type, season, situation, age / gender of the user, etc. Consists of what to do. Only the maximum number of factors is defined for the number of factors included in the Kansei metaword group and the basic metaword group, and the number of factors may vary. Also, there may be no variable factor. However, the minimum and maximum number of factors included in the idiom metaword is defined. The Kansei metaword group and the basic metaword group include individual metawords and their respective score values as shown in FIG. 8 (B).

The metaword factors in the idiom metaword may be arranged in a mixed order, but in this explanation, they are arranged in descending order of score value from front to back for each fixed metaword and variable metaword. By searching based on this idiom metaword for requests from users, it is possible to obtain content information corresponding to the meaning of the idiom metaword without considering variable factors such as content type and season. Become.

As shown in FIG. 9A, the tag attached to the content consists of a basic metaword, a sensitivity metaword, and an idiom metaword. The idiom metaword retains the information as shown in FIG. 9B. Metaword factors and reasons for recommendation associated with idiom metawords may change or change depending on the domain of the target content, user profiling (age, gender, preference tendency), season, time zone, environmental conditions such as weather, etc. May be optimized. The "idiom name" is a name for displaying the idiom metaword. An example of the response text (recommended text) is as shown in FIG. 9 (C). Examples of idiom metawords and corresponding metaward factors are shown in FIG.

FIG. 11 shows a specific configuration example of the embodiment. In this figure, the search device 100 includes a text input unit 10, an idiom metaword extraction unit 20, a primary metaword determination unit 30, a search unit 40, and a display unit. 50, an idiom metaword dictionary 601, an idiom metaword tag storage unit 602, a search target content tag storage unit 603, a search target content information database 604, and a recommended text storage unit 605. The search device 100 is typically configured using one or more computer resources (computer systems). The search device 100 is realized, for example, by installing a computer program recorded on the recording medium 201 or a computer program transmitted via a communication network (not shown) in the computer system 200. The computer system includes a CPU, a main memory, a bus, an external memory, various input / output interfaces, and the like, and may be a personal computer, a smartphone, an information home appliance, or the like.

The text input unit 10 inputs free input text, catch phrase, etc., performs morphological analysis, semantic analysis, and the like, and cuts out words and phrases. If there is a reserved word that precedes or follows the idiom metaword, the idiom metaword may be cut out based on this. The idiom metaword extraction unit 20 extracts one or a plurality of idiom metawords with reference to the idiom metaword dictionary 601. The idiom metaword dictionary 601 holds candidate words or phrases in addition to the same word or phrase for each entry of the idiom metaword, and when these words or phrases appear, the corresponding idiom metaword is displayed. You can extract it. The primary metaword determination unit 30 refers to the tag storage unit 602 for the idiom metaword, and extracts the primary metaword (basic metaword or sensitivity metaword) associated with the extracted idim metaword and the importance score value of each. The search unit 40 refers to the tag storage unit 603 for the search target content for the primary metaword associated with the extracted idiom metaword, calculates the matching degree with the search target content, and searches one or more with a large matching degree. Identify the target content. If the degree of matching is smaller than a predetermined threshold value, the search fails. The matching degree may be calculated by multiplying the importance score value of the primary metaword associated with the idiom metaword and the importance score of the primary metaword associated with the search target content, and accumulating the values. This will be described later with reference to FIGS. 12 and 13. The display unit 50 extracts and displays information about the search target content determined based on the matching degree from the internal or external address with reference to the search target content information database 604, and also refers to the recommended text storage unit 605. Then, the recommendation text of the idiom metaword is extracted and displayed. Images and sounds may be used instead of the recommended text.

The importance score of the primary metaword can be prepared using the TF-IDF described above. The idiom metaword importance score can be prepared using TF-IDF for the text corresponding to the idiom metaword.

12 and 13 show an example of calculating the degree of matching between the idiom metaword and the content to be searched. In this example, the idiom metawords “adult dignity” and “small devil” and products A, B and The degree of matching with C is calculated. The idiom metaword tags for the idiom metawords "small devil" and "adult dignity" are as shown in FIG. 12 (A), and the search target content tags for products A, B, and C are shown in FIG. 12 (B). It is shown. For these products, the degree of matching with each idiom metaword is calculated. In this calculation example, the metaword factor (primary metaword) of each idiom metaword and the same word in the tag of each product are multiplied by the score value on the idiom side of that word and the score value of the tag of the product. , The value obtained by adding it to the metaword factors of all idiom metawords is used as the matching score value.

The calculation results for each idiom metaword are as shown in FIG. As a result, the product B is displayed for the "adult dignity ", and the recommendation text of the idiom metaword is also displayed.

The text input unit 10, the idiom metaword extraction unit 20, the primary metaword determination unit 30, the search unit 40, and the display unit 50 in FIG. 11 execute each step of the corresponding search method.

It is also possible to calculate an idiom metaword that is highly relevant to each content offline and manage it as a tag for each content in advance. For example, as shown in FIG. 17, a learning unit 70 may be provided to add an idiom metaword and a corresponding importance score value to the search target content tag. By doing so, the idiom metaword can be treated in the same way as the primitive metaword (basic metaword and Kansei metaword). At that time, there may be a plurality of idiom metawords related to one content. This idiom metaword is premised on being defined in advance by the service provider, but may be added in response to a request from the user side. In FIG. 17, the parts corresponding to FIG. 11 are designated by the corresponding reference numerals.

FIG. 14 shows examples of idiom metawords, Kansei metawords, and basic metawords. Further, FIG. 15 shows an example of the format of the metaword in JSON (Javascript (registered trademark) Object Notification) format. It may be described in the format shown in FIGS. 16A and 16B.

FIG. 18 shows another modification of the specific configuration example of FIG. In this figure, a metaword extraction unit 202 is provided to extract not only the idiom metaword but also the basic metaword and / or the sensation metaword from the input text, and based on the extracted basic metaword and / or the sensation metaword and the idiom metaword. It is possible to perform a complex search using the determined basic metaword and / sensitivity metaword. In FIG. 18, the parts corresponding to FIG. 11 are designated by the corresponding reference numerals.

In the above explanation, the catch phrase is a promotional phrase such as a campaign theme for products and content information. The idiom metaword is basically fixed, and although additional modifications are made, the frequency of changes is low. A plurality of idiom metawords may be associated with each of the user request input text, the catch phrase, and the emotional motivation phrase as the user's desire information. In the case of input from the user, a plurality of Kansei factors are extracted from text, voice, image, and biometric information, and associated with idiom metadata having many metaword factors. Multiple Kansei metaword factors and basic metaword factors are associated with idiom metawords. Separately from this, a theme word may be set and a plurality of Kansei metaword factors and basic metaword factors may be associated with each other. It is assumed that the metaward factors added to these are related based on information such as seasons, usage scenes, product information, related articles, and reputation on the Internet, and are added when such information is updated. The metaward factor that is used may also be changed. When the theme word has a strong connection between the added Kansei metaword factor and the Kansei metaword factor added to the idiom metaword, the theme word and the idiom metaword may be associated with each other. If there is no corresponding theme phrase, it will not be associated. If the metaword factors attached to the idiom metaword and theme word are changed, these relationships will also be changed. You may use the theme phrase as the recommendation text.

FIG. 19 is an introductory text of the theme word “Shiretoko Goko”, from which metaword factors as shown in FIG. 20 are extracted. Of these, the idiom metaword (idiom tag) is associated with (great nature, view) → "majestic nature", (fantastic, quiet) → "fantastic silence" from the metaword factor.

This is the end of the description of the embodiment.

The present invention is not limited to the above-described embodiment, and various modifications can be made without departing from the spirit of the present invention.

For example, a metaword factor associated with an idiom metaword can also be defined as a collection of metaword factors with close vector distances when a word is mapped to a semantic vector space as used in Japanese Patent No. 6571231. ..

In one example, the type and number of words used as metawords shall be predefined (referred to as defined primary metawords; also referred to as defined primary metawords), and words similar to defined primary metawords shall be relevant. It is assumed that the name is identified as the primary metaword.
ex. Prescribed primary metaward (warm): warm, warm, warm, fluffy ... (Nayoro word)

In this example, for example, for the idiom metaword "adult hideout", the target is a collection of text sentences such as articles, explanations, synopses, and essays in which the target idiom metaword appears. Of the words that appear in the vicinity of the idiom metaword, the idiom metaword can be expressed as a vector using the specified primary metaword by obtaining the statistical information of the distance between the word and the specified primary metaword after name identification. The vector space expressed by this defined first-order metaword is defined as the semantic vector space. As one of the methods, there is a method such as "Word2Vec" (https://depage.net/bigdata/machine_learning/2016/02/02/word2vec_power_of_word_vector.html). When each idiom metaword is expressed in a semantic vector space, the similarity between a plurality of idiom metawords can be obtained by calculation such as cosine similarity.

For example, when seeking a prescriptive primary metaword related to the target idiom "Adult Hideaway", the reference words are words such as "Adult Hideaway", "Man's Hideaway", and "Adult Hideaway", which are highly similar phrases. , In the text, the relationship between the idiom metaword and the specified primary metaword is obtained from the appearance rate of the specified primary metaword that appears before and after that, or the word that can be identified as the specified primary metaword.

Consider the following example.
Example of text sentence (1) A true adult hideaway that you do not want to teach. A feeling of excitement like a secret base.
(2) A hideaway that you really want to keep secret! A fashionable space with a calm atmosphere.
(3) "Adult hideaway known only to those who are used to playing". We are particular about such a secret place, and the door is opened only to customers who understand its essence.
In the example of these text sentences, when the score values indicating the high degree of relevance of the prescribed primary metaword are normalized and arranged in descending order, the result is as follows.
"Adult hideout"
Adult: 0.241 Calm: 0.146 Secret: 0.117 Exciting: 0.104 Secret: 0.08 Commitment: 0.07 Fashionable: 0.06 Essence: 0.02 Door: 0.01
In this way, the idiom metaword can be associated with the primary metaword and its score value.

10 Text input unit 20 Idiom metaword extraction unit 30 Primary metaword determination unit 40 Search unit 50 Display unit 70 Learning unit 100 Search device 200 Computer system 201 Recording medium 202 Metaword extraction unit 601 Idiom metaword dictionary 602 Idiom metaword tag storage unit 603 Search target content tag storage unit 604 Search target information content database 605 Recommended text storage unit

Claims (6)

A primary metaword association data holding means for search target content that holds primary metaword association data for search target content that associates one or more primary metawords included in a pre-registered primary metaword set with the search target content.
Primary metaword association data for secondary metawords that retains primary metaword association data for secondary metawords that associates multiple primary metawords included in the above primary metaword set with secondary metawords included in a pre-registered secondary metaword set. Holding means and
A display target content holding means for associating and holding display target content with at least one secondary metaword included in the above secondary metaword set, and a display target content holding means.
Search input means for inputting text and search input means
A secondary metaword extraction means that extracts one or more secondary metawords from the search input text input from the search input means, and a secondary metaword extraction means.
For the one or more secondary metawords extracted by the secondary metaword extraction means, the primary metaword related to the one or more secondary metawords is determined by referring to the primary metaword association data holding means for the secondary metawords. Related primary metaword determination means,
A search means for searching the search target content using the above-mentioned primary metaword determined by the above-mentioned related primary metaword determining means, and
When the search target content is searched in relation to the at least one secondary metaword, the display target content stored in the display target content holding means in relation to the at least one secondary metaword is displayed. Display target content display means and
A search device comprising: a search target content display means for displaying search target content retrieved by the search means. The search device according to claim 1, wherein the display target content is a text described in relation to the secondary metaword. The primary metaword association data for the search target content has one or a plurality of primary metawords for each of the search target contents, and a score value related to the search target content of the primary metaword.
The primary metaword association data for the secondary metaword has a plurality of primary metawords for each of the secondary metawords and a score value related to the secondary metaword of the primary metaword.
The search means refers to the primary metaword association data for the secondary metaword for the extracted secondary metaword, extracts the score value related to the secondary metaword of the plurality of primary metawords, and obtains the score values related to the secondary metaword of the plurality of primary metawords. With reference to the primary metaword association data for the search target content, the score values related to the search target content of the plurality of primary metawords are extracted, and the score values related to the secondary metawords of the plurality of primary metawords and the plurality of primary metawords are obtained. The search device according to claim 1 or 2, wherein the search target content is searched based on the score value related to the search target content. The secondary metaword extraction means extracts one or more primary metawords in addition to the one or more secondary metawords from the search input text input from the search input means.
The search means is described in any one of claims 1 to 3 for searching the search target content using the extracted one or a plurality of primary metawords in addition to the primary metawords determined by the related primary metaword determining means. Search device. A step in which the primary metaword association data retention means for search target content retains primary metaword association data for search target content that associates one or more primary metawords included in a pre-registered primary metaword set with the search target content.
Primary metaword association for secondary metawords Primary metaword association for secondary metawords The data retention means associates multiple primary metawords included in the above primary metaword set with secondary metawords contained in a pre-registered secondary metaword set. The steps to hold the data and
A step in which the display target content holding means holds the display target content in association with at least one secondary metaword included in the above secondary metaword set.
The search input method is a step to enter the search input text,
A step in which the secondary metaword extraction means extracts one or more secondary metawords from the search input text input from the search input means.
The related primary metaword determining means refers to the primary metaword association data holding means for the secondary metaword for the one or more secondary metawords extracted by the secondary metaword extracting means to the one or more secondary metawords. Steps to determine the relevant primary metaword,
A step in which the search means searches for the content to be searched using the above-mentioned primary metaword determined by the above-mentioned related primary metaword determining means, and
When the search target content is searched in relation to the at least one secondary metaword, the display target content display means is stored in the display target content holding means in relation to the at least one secondary metaword. The steps to display the above display target content and
A search method, characterized in that the search target content display means includes a step of displaying the search target content retrieved by the search means. Computer,
Primary metaword association for search target content that retains primary metaword association data for search target content that associates one or more primary metawords included in a pre-registered primary metaword set with search target content.
Primary metaword association data for secondary metawords that retains primary metaword association data for secondary metawords that associates multiple primary metawords included in the above primary metaword set with secondary metawords included in a pre-registered secondary metaword set. Holding means,
Display target content holding means for holding display target content in association with at least one secondary meta word included in the above secondary meta word set.
Search input means for inputting text,
A secondary metaword extraction means, which extracts one or more secondary metawords from the search input text input from the search input means.
For the one or more secondary metawords extracted by the secondary metaword extraction means, the primary metaword related to the one or more secondary metawords is determined by referring to the primary metaword association data holding means for the secondary metawords. Related primary metaword determination means,
A search means for searching search target content using the above-mentioned primary metaword determined by the above-mentioned related primary metaword determining means,
When the search target content is searched in relation to the at least one secondary metaword, the display target content stored in the display target content holding means in relation to the at least one secondary metaword is displayed. Display target content display means and
Search target content display means that displays the search target content retrieved by the above search means,
A computer program used to run as.

Images