JP7127351B2 - KANSEI SEARCH PROGRAM, INFORMATION PROCESSING DEVICE, AND KANSEI SEARCH METHOD - Google Patents

Description

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

In recent years, various web services have been provided. Among them, search services are frequently used services such as hotel searches, restaurant searches, and product searches on EC (Electronic Commerce) sites.

However, when searching for products such as food and clothes, where emotional values such as atmosphere and image are more important than values defined by specs, what are the product descriptions and emotional keywords used by the person searching? not necessarily match. Therefore, there is a problem that a search system based on keyword matching does not function effectively.

Therefore, an emotional search system is known that can intuitively search for products whose emotional value is emphasized.

In the Kansei search system, side information (external features, personality, etc.) indicating the type of image to be searched is defined. For words, Kansei information is defined in which an image of a search target is represented by a quantitative value (kansei vector) for each aspect information.

In conventional emotion search systems, multiple words (each expression) included in a query input by a user for searching are unconditionally converted into emotion vectors, and search results are displayed based on the degree of similarity of the emotion vectors. do.

For example, an example in which a user enters a query of "broad dress shirt that can be used even when off" to perform a search will be described. In this query "broad shirt that can be used even when off", the words "broad shirt" that indicates the material and "shirt" that specifies the type of shirt can be said to be indispensable conditions desired by the user. On the other hand, the term "can be used even when turned off" is an expression containing ambiguity.

In a conventional sensitivity search system, "broadcloth", "shirt", and "can be used even off" included in a user query are replaced with sensitivity vectors, and the values of these sensitivity vectors are combined to calculate a comprehensive sensitivity vector. Then, it outputs a search result based on the degree of similarity with the value of this comprehensive sensitivity vector.

JP 2011-48527 A

However, as described above, in the conventional affective search system, a plurality of words (each expression) contained in a query input by a user for searching are unconditionally converted into affective vectors, and the similarity of the affective vectors is mainly used. to display the search results.

As a result, the search result may show "a casual shirt that is more dress-like" instead of "a dress shirt". result is not reflected. That is, a search result not intended by the user is output.

In one aspect, an object of the present invention is to enable a user's intention to be reflected in a search result to be obtained in an emotional search.

For this reason, this sensitivity search program designates, in a plurality of words included in a query, terms that belong to an attribute item vocabulary set that is preliminarily extracted to be used as attribute items of a database to be searched, and that the user cannot remove the terms. and among the plurality of words belonging to other than the attribute item vocabulary set, a plurality of first words that can be developed into sensibility information as expressions containing ambiguity if the occurrence frequency exceeds a predetermined threshold on the other hand, presumes words whose frequency of appearance is equal to or less than the predetermined threshold as vocabulary specifying the essential condition, sets priorities to the plurality of first words, and according to the priorities, A computer is caused to execute a process of carrying out a sensitivity search on search target information for a first word selected from among a plurality of first words.

According to one embodiment, it is possible to obtain a search result that reflects the user's intention in the sensitivity search.

FIG. 10 is a diagram showing the relationship of appearance frequencies of vocabularies in the explanatory sentence vocabulary set; 1 is a diagram schematically showing the configuration of an information processing system as an example of an embodiment; FIG. FIG. 4 is a diagram illustrating a display screen of a client terminal in the information processing system as one example of an embodiment; 1 is a diagram illustrating a hardware configuration of a management server in an information processing system as one example of an embodiment; FIG. 1 is a diagram showing a functional configuration of a management server in an information processing system as one example of an embodiment; FIG. FIG. 10 is a diagram illustrating a Kansei vector definition dictionary in the information processing system as one example of an embodiment; 9 is a flowchart for explaining re-search processing for the same user query in the information processing system as an example of the embodiment; 4 is a flowchart illustrating processing in an information processing system as an example of an embodiment;

Hereinafter, embodiments of the present sensitivity search program, information processing apparatus, and sensitivity search method will be described with reference to the drawings. However, the embodiments shown below are merely examples, and are not intended to exclude the application of various modifications and techniques not explicitly described in the embodiments. That is, the present embodiment can be modified in various ways (such as by combining the embodiment and each modified example) without departing from the spirit of the embodiment. Also, each drawing does not mean that it has only the constituent elements shown in the drawing, but can include other functions and the like.

(A) Overview First, an overview of processing by the information processing system 1 (see FIG. 2) as an example of the embodiment will be described.
In this information processing system 1, in the search query text input by the user, (1) words (vocabulary) that can be developed into sensitivity vectors as expressions containing ambiguity, and (2) conditions that the user cannot remove are specified. Each word is estimated.

For example, the search query "broadcloth fashionable pants" for searching clothing includes three words (vocabulary): "broadcloth", "fashionable" and "pants".

Among these three vocabularies, "fashionable" is an expression (vocabulary) that contains ambiguity and has a high degree of abstraction. Vocabulary containing such ambiguity tends to appear statistically with high frequency because it is also used to modify various other words in descriptions of products. Hereinafter, the vocabulary of expressions that include ambiguity and have a high degree of abstraction may be referred to as "ambiguous abstract expressions".

On the other hand, "broad fabric" and "pants" are terms that specify conditions (essential conditions) that the user cannot remove. or "skirt", the user will not be satisfied with the search results.

The term “broad cloth”, which is a term that specifies conditions that users cannot exclude, is used to modify the characteristics of individual products in a specific and limited manner. It tends to be infrequent. Hereinafter, a term that specifically and restrictively modifies a feature may be referred to as a "specific restrictive modifying expression".

However, even if it is a vocabulary that specifies conditions that the user cannot exclude, general terms (for example, "skirt") that are set as attribute items in the search DB (Data Base) have a high frequency of use (appearance frequency). .

Further, hereinafter, a query (search query) input by a user to perform a search may be referred to as a user query.

FIG. 1 is a diagram showing the relationship of the frequency of occurrence of vocabularies in the explanatory sentence vocabulary set.

The search target product DB 101 includes a set of description texts that constitute description texts of products and attribute items of the DB (DB attribute items). A vocabulary that constitutes a DB attribute item may be referred to as a DB attribute item vocabulary set.

In the example shown in FIG. 1, the vocabulary included in the product description stored in the search target product DB 101 (explanatory text set) and the vocabulary used as the attribute item of the DB (DB attribute item, DB attribute item vocabulary set) are collectively represented as an explanatory sentence vocabulary set X.

This descriptive text vocabulary set X includes a set A of ambiguous abstract expressions and a set B of specific restrictive modifying expressions. Further, the set B of specific restrictive modifying expressions includes a DB attribute item vocabulary set C (the set C also includes DB attribute items).

The relationship of the frequency of occurrence of vocabulary in the explanatory sentence vocabulary set X can be expressed as follows.

{Frequency of vocabulary set A}>{Frequency of (vocabulary set BC)}
In this information processing system 1, according to the following procedures 1 to 4, (1) words (vocabularies) that can be developed into sensitivity vectors as expressions including ambiguity, and (2) vocabulary that specifies conditions that the user cannot remove. are estimated respectively.

For estimating the vocabulary, the frequency (appearance frequency) of the vocabulary used in the entire set of descriptions of the product to be searched is used.

Procedure 1: In the information processing system 1, the vocabulary in the product introduction description in the search target product DB 101 (see FIG. 1) is extracted in advance and made into a set X. Also, the frequency information of these vocabularies is obtained.

Procedure 2: Vocabularies used as DB attribute items in the search target product DB 101 are extracted in advance, and these vocabularies are defined as a set C. Vocabulary preset as item information in such a DB to be searched is unconditionally presumed to be (2) the user's unavoidable condition.

Procedure 3: Presume that the vocabulary corresponding to set C in the user query is (2) a vocabulary that specifies a condition that the user cannot omit.

Step 4: Among the remaining vocabularies in the user query, the vocabularies whose frequency of appearance in the set X exceeds a predetermined threshold are estimated as (1) vocabularies that can be developed into sensitivity vectors as expressions containing ambiguity. In addition, words below the threshold value are assumed to be (2) vocabulary specifying conditions that cannot be removed by the user.

In this way, with respect to vocabulary other than the set C, words whose frequency is less than a predetermined threshold are (1) words that can be developed into a sentiment vector as expressions containing ambiguity, and words that exceed the threshold are (2) It is presumed to be a word that includes an unavoidable condition.

For example, it is assumed that "shirt", "T-shirt", "polo shirt", etc. are included as DB attribute items of the search target product DB 101, and an example in which the user query is "a broad shirt that can be used even when off" is shown.

This user query includes three terms: "can be used even when off", "broadcloth", and "shirt".

Here, since "shirt" is included as a DB attribute item in the search target product DB 101, it is identified as (2) a vocabulary specifying a condition that the user cannot remove.

Also, in the set X, if the appearance frequency of "can be used even when turned off" is greater than the threshold and the appearance frequency of "broad" is less than or equal to the threshold, "can be used even when turned off" is (1) an expression containing ambiguity , and "broad" is presumed to be a vocabulary specifying (2) a condition that the user cannot remove.

In addition, in the information processing system 1, (1) words that can be developed into sensitivity vectors as expressions including ambiguity are developed into sensitivity vectors, with priority given to vocabulary that appears frequently in the search target product DB 101. Then, a full-text search is performed in consideration of the sensitivity vector developed each time this development is performed.

(B) Configuration FIG. 2 is a diagram schematically showing the hardware configuration of the information processing system 1 as an example of the embodiment.

As shown in FIG. 2, the information processing system 1 includes a management server 10, a network 30, and one or more (two in the example shown in FIG. 1) client terminals 20-1 and 20-2.

The management server 10 and the client terminals 20-1 and 20-2 are connected via a network 30 so as to be communicable.

The client terminals 20-1 and 20-2 are information processing devices (computers) such as PCs (Personal Computers), smart phones, tablets, etc., and are used by users to use services provided by the management server 10. FIG. In this embodiment, the user utilizes the search system provided by the management server 10 using the client terminals 20-1 and 20-2.

Note that, hereinafter, as a code indicating a client terminal, codes 20-1 and 20-2 are used when it is necessary to specify one of a plurality of client terminals, but code 20 is used when an arbitrary client terminal is indicated. .

The client terminal 20 includes, for example, a CPU (Central Processing Unit), a memory, an input device, and an output device (not shown), and implements various functions by the CPU executing programs.

For example, in the client terminal 20, a user inputs a query (user query; search condition) for performing a search using an input device such as a keyboard or a touch panel, and a search transmitted from the management server 10 in response to this search query. Get results. The acquired search results are output to an output device (display device) such as a display or touch panel.

Note that the user query may be input by voice input using a microphone (not shown). Also, the search results may be output by voice using a speaker (not shown).

Further, in the information processing system 1, the user uses an input device to input a text related to a search target as a search condition (user query) in order to use the search service. A user query may be a keyword, or may be an arbitrary text sentence (free sentence, free text).

FIG. 3 is a diagram illustrating a display screen of the client terminal 20 in the information processing system 1 as one example of the embodiment.

The display screen illustrated in FIG. 3 displays a search service provided by the management server 10 as a web page by a browser.

This display screen includes an interface screen 501 and a search result display screen 503 .

Interface screen 501 is used for interaction between the user and the search service. For example, on the interface screen 501 illustrated in FIG. 3, a query from the search service to the user, "Where are you looking for clothes to wear?" is displayed.

A user inputs a user query (search query) from a user query input box 502 in the form of answering a question from the search service displayed on this interface screen 501 . In the example shown in FIG. 3, the user has entered a user query "broad dress shirt that can be used even when off".

The search service searches the search target product DB 101 based on the user query input from this user query input box 502 .

A search result display screen 503 outputs the search result (search result) performed by the search system. In the search result display screen 503 illustrated in FIG. 3, images and descriptions of a plurality of shirts are displayed as a list as a result of the search performed by the management server 10 based on the input user query.

The network 30 is, for example, a computer network such as Ethernet (registered trademark). The client terminal 20 and the management server 10 may be connected to the network 30 by wire or wirelessly.

The management server 10 is, for example, an information processing device (computer) having a server function. The management server 10 includes a search target product DB 101 (see FIG. 1), searches the search target product DB 101 based on a user query input from the client terminal 20 by the user, and sends the search result to the client terminal 20. respond.

In this embodiment, as a search service provided by the information processing system 1, an example of a product search system that handles clothing as search target products will be described.

FIG. 4 is a diagram illustrating the hardware configuration of the management server 10 in the information processing system 1 as an example of the embodiment.

The management server 10 is an information processing device (computer) having a server function, and as shown in FIG.

As shown in FIG. 4, the management server 10 is provided with a search target product DB 101, a product DB attribute item vocabulary DB 102, an affective vector definition dictionary 103, and an explanation sentence vocabulary frequency information DB 104. FIG. The search target product DB 101, product DB attribute item vocabulary DB 102, sensitivity vector definition dictionary 103, and explanation sentence vocabulary frequency information DB 104 will be described later.

The storage device 13 is a storage device such as a HDD (Hard Disk Drive), SSD (Solid State Drive), SCM (Storage Class Memory), etc., and stores various data and programs.

The network interface 14 is an interface communicably connected to the network 30, and is, for example, a LAN (Local Area Network) interface.

The memory 12 is a storage memory including ROM (Read Only Memory) and RAM (Random Access Memory). The ROM of the memory 12 is written with an OS (Operating System), a software program for search processing, and data for this program. A software program on the memory 12 is appropriately read and executed by the CPU 11 . Also, the RAM of the memory 12 is used as a primary storage memory or a working memory.

The CPU 11 is a processing unit (processor) containing a control unit (control circuit), an arithmetic unit (arithmetic circuit), a cache memory (register group), etc., and performs various controls and operations. The CPU 11 implements various functions by executing the OS and programs stored in the memory 12 . In the information processing system 1, a search system (product search system) is realized by the CPU 11 executing a sensitivity search program.

FIG. 5 is a diagram showing the functional configuration of the management server 10 of the information processing system 1 as an example of the embodiment.

By executing the OS and programs stored in the memory 12, the CPU 11 creates a search target product DB 101, a product DB attribute item vocabulary DB 102, an affective vector definition dictionary 103, and a description sentence vocabulary frequency information DB 104, which are illustrated in FIG. Realize each function.

Further, by executing the OS and programs stored in the memory 12, the CPU 11 executes the product DB attribute item vocabulary extraction unit 111, the explanation sentence vocabulary extraction unit 112, the explanation sentence vocabulary frequency calculation unit 113, The functions of the sensitivity vector calculation unit 114, the user query input unit 115, the input text analysis unit 116, the sensitivity vector expansion vocabulary identification unit 110, the hybrid search unit 120, and the search result display unit 126 are realized.

In addition, the sensitivity vector expansion vocabulary identification unit 110 functions as a DB attribute item vocabulary identification unit 117, a vocabulary frequency setting unit 118, a sensitivity vector expansion vocabulary identification unit 119, a product DB attribute item vocabulary DB 102, and a description sentence vocabulary frequency information DB 104. Prepare.

Further, the hybrid search unit 120 has functions as a search control unit 121 , a sensitivity vector expansion unit 122 , a sensitivity search execution unit 123 , a keyword search execution unit 124 and a search result integration unit 125 .

The search target product DB 101 is a database that stores information about products to be searched.

The search target product DB 101 stores, as product information, product IDs, category attributes and their items, product descriptions, and product sensitivity vectors, for example.

The product ID is identification information that uniquely identifies the product. For example, information composed of a combination of alphanumeric characters and symbols such as “ID=0237” is set in advance for each product as the product ID.

A category attribute is a type of information about a product. For example, information such as "type of shirt (shirt)", "color", and "pattern" is used as category attributes. An item is a concrete value of a category attribute.

Specific combinations of category attributes and items are exemplified below.

(Category attribute: item) = (shirt: dress shirt), (color: white), (pattern: stripe), ...
The description is text that describes the product. For example, the sentence "a loose shirt made of broadcloth" is used.

A sensitivity vector is information that quantitatively expresses an image (impression) of a search target (product in this embodiment) using a vector (sensitivity vector) for each aspect information.

That is, the kansei vector expresses the impression that a person receives from the vocabulary by numerical values (kansei score, vector value) on one or more (three types in this embodiment) kansei axes. It is information indicating the type of image represented by the emotional expression, such as "characteristics", "appearance", "personality", etc., and at least one side information for one emotional expression according to a predetermined classification system such as an existing ontology. shall be given.

As side information, the categories of the general noun semantic attribute system of the Japanese lexical overview (high grade, soft: "quality", bright: "personality", etc.) may be used.

It is assumed that the sensitivity axis corresponds to side information. Also, the sensitivity vector may be referred to as sensitivity information.

The sensitivity vector has a predetermined value for each dimensional axis (sensitivity axis) of a vector space having a predetermined number of dimensions. For example, it is assumed that the number of dimensions of the sensitivity vector is 3, the first dimension axis=traditional, the second dimension axis=casual, and the third dimension axis=familiar.

That is, in this example, "traditional," "casual," and "comfortable" are used as side information about shirts. The kansei vector expresses each of these side information as a kansei axis using a kansei score, which is a numerical value.

As the sensitivity score, for example, a numerical value of 1 or less may be used. For example, it is assumed that the more traditional, the higher the value of the sensitivity score. Similarly, it is assumed that the more casual the user is, the higher the value of the sensitivity score, and the more sophisticated the user is, the higher the value of the sensitivity score.

In this embodiment, the sensitivity vector is represented by a combination of numerical values representing sensitivity scores as follows.
(traditional, casual, mature) = (0.6, 0.7, 0.2)
The search target product DB 101 has such a sensitivity vector for each product.

In addition, in the search target product DB 101, the sensitivity vector given from the outside (for example, manually defined) may be stored. Also, a value calculated by the sensitivity vector calculation unit 114 using the sensitivity vector definition dictionary 103, which will be described later, may be stored.

Note that the sensitivity vector can be realized using various known techniques, and detailed description thereof will be omitted.

The kansei vector definition dictionary 103 stores numerical values (kansei scores, vector values) corresponding to one or more kansei axes (side information) (three types of traditional, casual, and sophisticated in this embodiment) for a vocabulary. It is configured by associating combinations.

As will be described later, the sensitivity vector definition dictionary 103 is used when the sensitivity vector calculation unit 114 calculates the sensitivity vector for each product description in the search target product DB 101, and also includes vocabulary included in user queries (sensitivity vector expansion vocabulary). ) is used to calculate the sensitivity vector.

FIG. 6 is a diagram exemplifying the emotion vector definition dictionary 103 in the information processing system 1 as an example of the embodiment.

The sensitivity vector definition dictionary 103 illustrated in FIG. 6 is configured by defining sensitivity scores on three sensitivity axes of traditional, casual, and familiar with respect to vocabulary.

For example, for the vocabulary "dress shirt", the sensitivity score for the sensitivity axis "traditional" is 0.9, the sensitivity score for the sensitivity axis "casual" is 0.1, and the sensitivity score for the sensitivity axis "feeling mature" is 0.1.

The sensitivity vector calculation unit 114 refers to the sensitivity vector definition dictionary 103 for each product description in the search target product DB 101 to calculate a sensitivity vector corresponding to the product description of each product ID. For example, each time a vocabulary registered in the Kansei vector definition dictionary 103 appears in a description, the Kansei vector calculation unit 114 adds a Kansei score for each Kansei axis (traditional, casual, familiar). A value obtained by arranging the average values of the sensitivity scores is calculated as a sensitivity vector.

In the following, there is a case where calculation of sensitivity vectors for a vocabulary is expanded into sensitivity vectors.

The sensitivity vector calculation unit 114 also calculates sensitivity vectors for the vocabulary included in the user query by the same method.

The product DB attribute item vocabulary DB 102 is a database that stores a list of vocabulary used as category attribute names and vocabulary used as item names in the search target product DB 101 .

Hereinafter, vocabulary used as category attribute names and item names in the search target product DB 101 may be referred to as product DB attribute item vocabulary. That is, the product DB attribute item vocabulary DB 102 stores product DB attribute item vocabulary.

The product DB attribute item vocabulary extraction unit 111 creates a product DB attribute item vocabulary DB 102 by extracting a list of vocabularies used as category attribute names and item names from the search target product DB 101 and listing them. The creation of the product DB attribute item vocabulary DB 102 by the product DB attribute item vocabulary extraction unit 111 may be performed, for example, when the search system is initialized, or may be performed each time the search target product DB 101 is updated. .

The explanation sentence vocabulary frequency information DB 104 is a database that stores appearance frequency information of each vocabulary appearing in each explanation sentence of the product stored in the search target product DB 101 .

The explanatory text vocabulary extraction unit 112 performs, for example, morphological analysis processing on each explanatory text of the product stored in the search target product DB 101, thereby extracting the vocabulary used in the description and creating a list (vocabulary list). ). The descriptive text vocabulary extraction unit 112 may also perform post-processing such as compound noun conversion processing that combines consecutive noun morphemes in the descriptive text, and shaping processing such as aligning conjugations of verbs and adjectives into final forms. good. This post-processing can improve the accuracy of the frequency information.

Creation of the explanatory sentence vocabulary frequency information DB 104 by the explanatory sentence vocabulary extraction unit 112 may be performed, for example, at the initial setting of the search system, or may be performed each time the search target product DB 101 is updated.

The explanatory sentence vocabulary frequency calculation unit 113 calculates the appearance frequency of each vocabulary extracted by the explanatory sentence vocabulary extraction unit 112 . The calculation of the appearance frequency of this vocabulary can be realized by using a known technique, and the explanation thereof will be omitted. After calculating the frequency of occurrence, the explanatory sentence vocabulary frequency calculator 113 normalizes by dividing each frequency by the maximum frequency. The explanatory sentence vocabulary frequency calculation unit 113 registers the normalized frequency information in the explanatory sentence vocabulary frequency information DB 104 .

User query input unit 115 receives a user query input by a user. For example, in the web page illustrated in FIG. 3 , information input by the user from the user query input box 502 of the interface screen 501 is input to the user query input unit 115 . The user query input unit 115 is an example of an input unit that receives queries.

The input text analysis unit 116 morphologically analyzes the user query received by the user query input unit 115 . The input text analysis unit 116 extracts the vocabulary included in the user query using the same method as the processing by the explanatory text vocabulary extraction unit 112 . The input text analysis unit 116 post-processes the vocabulary extracted from the user query, such as compound noun processing that combines continuous noun morphemes into one, and shaping processing such as aligning conjugations such as verbs and adjectives into final forms. You can do it as

The sensitivity vector expansion vocabulary specifying unit 119 specifies whether or not to expand each vocabulary included in the user query into sensitivity vectors. That is, the sensitivity vector expansion vocabulary specifying unit 119 specifies vocabulary to be matched by sensitivity vectors and vocabulary to be matched by keywords in the hybrid search unit 120, which will be described later, and ranks the expansion into sensitivity vectors.

The DB attribute item vocabulary identification unit 117 refers to the product DB attribute item vocabulary DB 102 to determine whether or not the vocabulary included in the user query is used as the category attribute name or item name in the search target product DB 101 . When the vocabulary extracted from the user query matches the vocabulary registered in the product DB attribute item vocabulary DB 102, the DB attribute item vocabulary identification unit 117 determines that the vocabulary is the product DB attribute item vocabulary.

That is, by referring to the product DB attribute item vocabulary DB 102, the DB attribute item vocabulary identification unit 117 identifies one or more terms included in the user query that correspond to the product DB attribute item vocabulary (second term ).

The vocabulary frequency setting unit 118 refers to the explanatory text vocabulary frequency information DB 104 for vocabulary that is included in the user query and does not correspond to the product DB attribute item vocabulary, and sets the appearance frequency (appearance frequency information) of each vocabulary. Identify. Note that the appearance frequency can be obtained by a known technique, and the explanation thereof is omitted.

A value obtained by normalizing frequency information may be given to each word as a weight score for Kansei vector expansion.

The sensitivity vector expansion vocabulary identification unit 119 estimates (separates) whether the vocabulary included in the user query is to be searched by matching by sensitivity vectors or by keyword matching. Kansei vector expansion vocabulary identifying unit 119 is an example of a first estimating unit.

For each vocabulary extracted from the user query by the input text analysis unit 116, the emotion vector expansion vocabulary identification unit 119 refers to the explanatory sentence vocabulary frequency information DB 104 (set X), and stores the vocabulary in the search target product DB 101. Acquire the appearance frequency information in each description of the product.

The sensitivity vector expansion vocabulary identification unit 119 compares the frequency of appearance of each vocabulary extracted from the user query with a predetermined threshold value, and determines that the vocabulary is (1) a vocabulary that can be expanded to sensitivity vectors as an expression containing ambiguity. or (2) it is a vocabulary that specifies a condition that the user cannot omit.

Hereinafter, (1) a vocabulary that can be expanded into an affective vector as an expression containing ambiguity is sometimes referred to as an affective vector expanded vocabulary. This sensitivity vector expansion vocabulary corresponds to a plurality of first words that can be developed into sensitivity information (sensitivity vector) as an expression including ambiguity (sensitivity expression) among the plurality of words (vocabulary) included in the query.

In addition, hereinafter, (2) a vocabulary specifying a condition that a user cannot remove may be referred to as an essential condition vocabulary. This essential condition vocabulary corresponds to the second word other than the first word among the vocabulary included in the query.

If the appearance frequency of the vocabulary extracted from the user query is higher than a predetermined threshold, the sensitivity vector expanded vocabulary identification unit 119 estimates that the vocabulary is the sensitivity vector expanded vocabulary.

Further, when the appearance frequency of a vocabulary extracted from a user query is equal to or less than a predetermined threshold value, the sensitivity vector expansion vocabulary identifying unit 119 estimates that the vocabulary is a prerequisite vocabulary.

Furthermore, the sensitivity vector expansion vocabulary identification unit 119 presumes that the vocabulary included in the user query and identified by the DB attribute item vocabulary identification unit 117 as corresponding to the product DB attribute item vocabulary is the essential condition vocabulary. .

The sensitivity vector expanded vocabulary is subjected to sensitivity vector search by sensitivity search execution section 123 in hybrid search section 120, which will be described later. On the other hand, in the hybrid search unit 120, keyword search by the keyword search execution unit 124 is performed for the essential condition vocabulary.

In addition, the sensitivity vector expansion vocabulary identification unit 119 sorts the vocabulary in descending order of the set frequency information as the vocabulary with the highest priority to be expanded to the sensitivity vector, and finally, the vocabulary corresponding to the product DB attribute item vocabulary. Arrange them so that they line up. The sensitivity vector expansion vocabulary specifying unit 119 is an example of a setting unit.

For example, if there are a plurality of emotion vector-expanded vocabularies in the vocabulary included in the user query, the order of priority for expanding the vocabulary with a high frequency of occurrence in the search target product DB 101 into emotion vectors among the plurality of emotion vector-expanded vocabularies. set higher.

As a result, the hybrid search unit 120, which will be described later, preferentially searches for vocabularies with a high frequency of appearance in the descriptions of the search target product DB 101 for sensitivity vector search.

Note that setting the priority order may be expressed as weighting. A large (heavy) weight corresponds to a high priority, and a small (light) weight corresponds to a low priority.

The hybrid search unit 120 expands the vocabularies into sensitivity vectors in descending order of priority, and searches user query vocabularies that have not been expanded by keyword matching. Also, the expanded vocabulary is searched by matching using sensitivity vectors. After that, the logical product (AND) of these two search results is taken as the final search result.

The hybrid search unit 120 continues this search until there are no more words to expand while expanding the vocabularies into sensitivity vectors in descending order of priority.

The search control unit 121 expands the vocabulary up to a specific rank into sensitivity vectors, executes a search using the sensitivity vectors, executes a full-text search using keywords, and integrates the results. The processing is repeated while developing the vocabularies into sensitivity vectors in order from the vocabulary having the highest priority of sensitivity vector development.

The search control unit 121 develops the sensitivity vector in descending order of the weight for the sensitivity vector development, and performs a full-text search every time it is developed.

The sensitivity vector expansion unit 122 calculates the sensitivity vectors of the vocabularies up to the rank specified by the search control unit 121 by the sensitivity vector calculation unit 114 while referring to the sensitivity vector definition dictionary 103, and converts the sensitivity vectors of the obtained vocabularies. Find the average value.

The emotion search execution unit 123 extracts a product description (=product ID) having an emotion vector similar to the average value of the emotion vectors calculated by the emotion vector development unit 122 with a degree of similarity greater than or equal to a certain threshold. Similarity determination can be performed using various known techniques such as cosine distance. The sensitivity search execution unit 123 is an example of a sensitivity search unit.

The keyword search execution unit 124 performs a full-text search (keyword search) on the descriptions of the search target product DB 101 using, as keywords, the vocabulary included in the user query that has not been expanded into the sensitivity vector. The keyword search execution unit 124 is an example of a keyword search unit.

The search result integration unit 125 extracts logical product results of the search results of the emotional search execution unit 123 and the search results of the keyword search execution unit 124, and arranges the results in descending order of similarity of the sensitivity search execution unit 123.

In the case of the second or later search by the search control unit 121, the search result integration unit 125 arranges the newly searched search result (the current search result).

The search result display unit 126 displays the search result list integrated by the search result integration unit 125 to the user. The search result integration unit 125 and the search result display unit 126 are examples of an output processing unit.

By displaying the search results in descending order, it is as if the search results were sorted according to the essential condition vocabulary.

In addition, the search result display unit 126 presents the search results to the user in the order of development and search, so that the part of the expression containing ambiguity is expressed as if it were another expression that matches the sensibility of the user. The search results are presented to the user in a sorted manner so that the results matching the desired specific conditions are obtained as they are. The display of the search results to the user may be interactively additionally displayed in chronological order, or may be collectively displayed.

For example, the search result display unit 126 creates image data to be displayed on the search result display screen 503 of the display screen illustrated in FIG.

Further, in the information processing system 1, the sensitivity vector search may be performed for all vocabulary included in the user query.

That is, in the hybrid search unit 120, the search control unit 121 searches again for the vocabulary determined as the essential condition vocabulary by the emotion vector expansion vocabulary identification unit 119 among all the vocabulary included in the user query as the emotion vector expansion vocabulary. may be performed.

Hereinafter, among all the vocabularies included in the user query, the vocabularies that have not undergone the emotion vector search may be referred to as unexpanded vocabularies.

If there is an unexpanded vocabulary, the search control unit 121 sets the unexpanded vocabulary to 1 until the sensitivity vector search is performed for all the vocabularies included in the user query (until there are no more unexpanded vocabularies). Re-searching for the same user query may be repeated while switching to the Kansei vector expansion vocabulary one by one.

FIG. 7 is a flowchart (steps A1 to A5) for explaining re-search processing for the same user query in the information processing system 1 as an example of the embodiment.

At step A1, a user enters a user query. In this example, the user query is a text sentence "a broad shirt that can be used even when off".

In step A2, the emotion vector-expanded vocabulary identification unit 119 estimates the emotion vector-expanded vocabulary (A), essential condition vocabulary (B), and product DB attribute item vocabulary for all the vocabularies included in the user query.

In the user query input in step A1, the frequencies of appearance of the words "can be used even when turned off" and "broad" in the description of the search target product DB 101 are 0.2 and 0.01, respectively. The affective vector expansion vocabulary identification unit 119 compares these appearance frequencies with a preset threshold value to estimate the vocabulary "can be used even when turned off" as the affective vector expansion vocabulary, and the vocabulary "broad" as the essential condition vocabulary. presume. It is also assumed that the DB attribute item vocabulary specifying unit 117 has determined that the vocabulary "shirt" is the product DB attribute item vocabulary.

At step A3, a first search based on the user query is performed.
The sensitivity vector development unit 122 searches for the vocabulary "can be used even when off" for sensitivity vector search, and searches for the vocabulary "broad" and "shirt" as keywords.

The sensitivity vector expansion unit 122 expands the vocabulary "can be used even when turned off" estimated as the sensitivity vector expansion vocabulary into sensitivity vectors. That is, the sensitivity vector developing unit 122 causes the sensitivity vector calculation unit 114 to develop the vocabulary "can be used even when off" into sensitivity vectors.

It is assumed that (traditional, casual, mature) = (0.6, 0.7, 0.2) is obtained as a sensitivity vector by expanding the vocabulary "can be used even when off" to a sensitivity vector.

The sensitivity search execution unit 123 uses the obtained sensitivity vector (0.6, 0.7, 0.2) to search the search target product DB 101 (sensitivity vector search) (first search). As a result, the item description (=item ID) having an affective vector similar to the average value of the affective vector with a degree of similarity equal to or higher than a certain threshold is extracted.

In addition, the keyword search execution unit 124 performs a keyword search on the search target product DB 101 using the words “broad” and “shirt” as keywords.

The search result integration unit 125 integrates the search results from the emotional search execution unit 123 and the search results from the keyword search execution unit 124 . The search result display unit 126 presents the integrated search results to the user.

In the example shown in FIG. 7, in the first search in step A3, search results are output in the following order.

Dress shirt X (a loose-fitting broadcloth dress shirt)
Dress shirt Y (comfortable broadcloth dress shirt)
Dress shirt Z (dress shirt with a sense of familiarity. Features broad fabric.)

By performing a keyword search on the essential condition vocabulary and the product DB attribute item vocabulary in the first search result, it is possible to obtain a search result that satisfies the conditions that the user considers indispensable.

The search result display unit 126 presents the search results in the above-described order on the search result display screen 503 of the web page illustrated in FIG. 3, for example.

In addition, the search result display unit 126 may display the search results on the search result display screen 503 and present a message such as "such search results have been obtained", for example.

If the user is not satisfied with the search results displayed on the search result display screen 503, the user makes an input requesting display of other search results, such as "What else?", on the interface screen 501. may

As a result, in step A4, a second search (research) based on the same user query is performed.

In the second search based on the same user query, the search control unit 121 selects the unexpanded vocabulary included in the user query that is determined to be the essential condition vocabulary and is not the product DB attribute item vocabulary ( In this embodiment, "broad") is preferentially targeted for the sensitivity vector search.

The affective vector expansion unit 122 searches for the words “broad” and “can be used even when off” for affective vector search, and searches for the word “shirt” as a keyword.

The sensitivity vector developing unit 122 develops the vocabulary “broad” and the vocabulary “can be used even when off” into sensitivity vectors by the sensitivity vector calculation unit 114 .

It is assumed that (traditional, casual, mature) = (0.7, 0.5, 0.2) is obtained as a sensitivity vector by developing the vocabulary "broad" and the vocabulary "can be used even off" into sensitivity vectors.

The sensitivity search execution unit 123 uses the obtained sensitivity vector (0.7, 0.5, 0.2) to search the search target product DB 101 (sensitivity vector search) (second search). As a result, the item description (=item ID) having an affective vector similar to the average value of the affective vector with a degree of similarity equal to or higher than a certain threshold is extracted.

In addition, the keyword search execution unit 124 performs a keyword search on the search target product DB 101 using the vocabulary "shirt" as a keyword.

The search result integration unit 125 integrates the search results from the emotional search execution unit 123 and the search results from the keyword search execution unit 124, and outputs the search results.

In the example shown in FIG. 7, in the second search in step A4, search results are output in the following order.

Dress shirt X (a loose-fitting broadcloth dress shirt)
Dress shirt Y (comfortable broadcloth dress shirt)
Dress shirt Z (dress shirt with a sense of familiarity. Features broad fabric.)
・・・
Dress shirt α (bright casual dress shirt)
Dress shirt β (a comfortable shape-memory dress shirt)

In the result of the second search, dress shirts α and β whose material is not broad are included in the search result by carrying out the sensitivity vector search for the word “broad”. In other words, the search results that are unintended by the user are included in the search results by performing the sensitivity vector search for the word "broad", which is a condition that the user cannot remove.

However, since there is a possibility that the search result of the sensitivity vector search does not match the user's intention, the search result of the word "broad" is displayed at the lower level, and the search result that matches the user's intention is displayed at the higher level.

As a result, it is possible to obtain search results that satisfy the conditions that the user considers indispensable from the top of the search results, and to obtain unexpected search results that the user does not intend from the bottom of the search results. .

The search result display unit 126 presents the search results in the above-described order on the search result display screen 503 of the web page illustrated in FIG. 3, for example.

In addition, the search result display unit 126 may display the search results on the search result display screen 503 and present a message such as "such search results have been obtained", for example.

If the user is not satisfied with the search results displayed on the search result display screen 503, the user can re-enter an input requesting display of other search results, such as "What else?", on the interface screen 501. You can do it.

As a result, in step A5, a third search (research) based on the same user query is performed.

In this example, in the third search based on the same user query, there are no unexpanded vocabularies that are determined to be essential condition vocabularies and are not product DB attribute item vocabularies. Therefore, the search control unit 121 selects unexpanded vocabulary included in the user query and determined to be product DB attribute item vocabulary ("shirt" in this embodiment) as a subject of the emotion vector search. do.

The sensitivity vector development unit 122 searches the vocabulary "shirt", the vocabulary "broad", and the vocabulary "can be used even when off" for sensitivity vector search.

The sensitivity vector developing unit 122 causes the sensitivity vector calculation unit 114 to develop the vocabulary “shirt”, the vocabulary “broad”, and the vocabulary “can be used even when off” into sensitivity vectors.

By expanding all of the vocabulary “shirt”, the vocabulary “broad”, and the vocabulary “can be used off-duty” into the Kansei vector, (traditional, casual, mature) = (0.8, 0.4, 0.2) is obtained as the Kansei vector. shall be assumed.

The sensitivity search execution unit 123 uses the obtained sensitivity vector (0.8, 0.4, 0.2) to search the search target product DB 101 (sensory vector search) (third search). As a result, the item description (=item ID) having an affective vector similar to the average value of the affective vector with a degree of similarity equal to or higher than a certain threshold is extracted.

The search result integration unit 125 outputs the search result by the sensitivity search execution unit 123 as a search result.

In the example shown in FIG. 7, in the third search in step A5, search results are output in the following order.

Dress shirt X (a loose-fitting broadcloth dress shirt)
Dress shirt Y (comfortable broadcloth dress shirt)
Dress shirt Z (dress shirt with a sense of familiarity. Features broad fabric.)
・・・
Dress shirt α (bright casual dress shirt)
Dress shirt β (a comfortable shape-memory dress shirt)
・・・
Casual shirt γ (somewhat casual...)
Casual shirt ε (silk...)

In this third search result, by performing an affective vector search on all the words "shirt", "broad shirt", and "can be used even when you are off" included in the user query, casual shirts γ, ε that are not white shirts is included in the search results. In other words, the search results that are not intended by the user will be included in the search results by performing the sensitivity vector search even for the word "shirt", which is a condition that the user cannot remove.

However, since there is a possibility that the search result of the sensitivity vector search for the vocabulary "shirt" and the vocabulary "broad" may not match the user's intention, they are displayed at the lower level, and the search result that matches the user's intention is displayed at the higher level.

As a result, it is possible to obtain search results that satisfy the conditions that the user considers indispensable from the top of the search results, and to obtain unexpected search results that the user does not intend from the bottom of the search results. .

(C) Operation Processing in the information processing system 1 as an example of the embodiment configured as described above will be described according to the flowchart (steps S1 to S17) shown in FIG.

In step S<b>1 , the product DB attribute item vocabulary extraction unit 111 extracts a list of vocabulary used as category attribute names and item names from the search target product DB 101 .

In step S2, the product DB attribute item vocabulary extraction unit 111 creates a DB attribute item vocabulary DB 102 by listing the vocabulary extracted in step S1.

In step S3, the description vocabulary extraction unit 112 extracts the vocabulary used in the description by performing, for example, morphological analysis processing on each description of the product stored in the search target product DB 101. .

In step S<b>4 , the explanatory sentence vocabulary frequency calculation unit 113 calculates the appearance frequency of each vocabulary extracted by the explanatory sentence vocabulary extraction unit 112 . After calculating the frequency of occurrence, the explanatory sentence vocabulary frequency calculator 113 normalizes by dividing each frequency by the maximum frequency.

In step S5, the explanatory sentence vocabulary frequency calculation unit 113 creates the explanatory sentence vocabulary frequency information DB 104 by registering the normalized frequency information in the explanatory sentence vocabulary frequency information DB 104 in step S4.

In step S6, the sensitivity vector calculation unit 114 refers to the sensitivity vector definition dictionary 103 for each product description in the search target product DB 101, and calculates the value of the sensitivity vector corresponding to the product description of each product ID. calculate.

For example, each time a vocabulary registered in the Kansei vector definition dictionary 103 appears in a description, the Kansei vector calculation unit 114 adds a Kansei score for each Kansei axis (traditional, casual, familiar). A value obtained by arranging the average values of the sensitivity scores is calculated as a sensitivity vector.

In step S7, a user query is input from the user, and the user query input unit 115 accepts this input user query.

In step S8, the input text analysis unit 116 analyzes the user query input by the user.

In step S9, the DB attribute item vocabulary identification unit 117 refers to the product DB attribute item vocabulary DB 102 to determine whether or not the vocabulary included in the user query is used as a category attribute name or item name in the search target product DB 101. to decide. That is, the DB attribute item vocabulary identification unit 117 determines whether or not the vocabulary included in the user query is the product DB attribute item vocabulary.

In step S10, the vocabulary frequency setting unit 118 refers to the description sentence vocabulary frequency information DB 104 for vocabulary that is included in the user query and does not correspond to the product DB attribute item vocabulary, and sets the appearance frequency (appearance frequency information) of each vocabulary. set.

In step S11, the sensitivity vector-expanded vocabulary identification unit 119 compares the appearance frequency of each vocabulary extracted from the user query with a predetermined threshold value to determine whether the vocabulary is an emotion vector-expanded vocabulary or an essential condition vocabulary. identify.

In addition, the sensitivity vector expansion vocabulary identification unit 119 sorts the vocabulary in descending order of the set frequency information as the vocabulary with the highest priority to be expanded to the sensitivity vector, and finally, the vocabulary corresponding to the product DB attribute item vocabulary. Arrange them so that they line up.

In step S12, the search control unit 121 confirms whether or not there is a vocabulary included in the user query that has not yet been developed in the sensitivity vector.

As a result of the confirmation, if unexpanded vocabulary remains (see Yes route of step S12), the process proceeds to step S13.

In step S13, the sensitivity vector expansion unit 122 includes the unexpanded vocabulary among the vocabulary included in the user query in the sensitivity vector search target, and causes the sensitivity vector calculation unit 114 to expand the sensitivity vector.

In step S14, the sensitivity search execution unit 123 executes sensitivity vector search for the product description in the search target product DB 101 using the sensitivity vector obtained in step S13.

In step S15, the keyword search execution unit 124 performs a keyword search (full-text search) on the search target product DB 101 using essential condition vocabulary and product DB attribute item vocabulary. Note that if there is no vocabulary to be searched for by keyword, the process of step S15 is skipped.

In step S16, the search result integration unit 125 integrates the result of the sensitivity vector search performed in step S14 and the result of the keyword search performed in step S15. After that, the process returns to step S12.

On the other hand, as a result of the confirmation in step S12, if there is no unexpanded vocabulary left (see No route in step S12), the process proceeds to step S17.

In step S17, the search result display unit 126 outputs the search results integrated by the search result integration unit 125 in step S16.

(D) Effect As described above, according to the information processing system 1 as an example of the embodiment, the sensitivity vector expansion vocabulary identification unit 119 determines the appearance frequency in the description of the search target product DB 101 for each vocabulary included in the user query. Then, the Kansei vector expansion vocabulary and the essential condition vocabulary are estimated based on the above.

In addition, the DB attribute item vocabulary identification unit 117 identifies product DB attribute item vocabulary whose vocabulary is used as category attribute names and item names in the search target product DB 101 in the vocabulary included in the user query.

In the hybrid search unit 120, a sensitivity vector search is performed on the sensitivity vector expanded vocabulary, and a keyword search is performed on the essential condition vocabulary and the product DB attribute item vocabulary. output as

That is, by performing a keyword search on essential condition vocabulary and product DB attribute item vocabulary assumed to be conditions intended by the user, the user's search intention can be reflected in the search results. In addition, by performing an affective vector search on the affective vector expanded vocabulary, it is possible to perform an intuitive affective search, and to obtain search results in which the affective vector expanded vocabulary is ambiguous.

The kansei vector expansion vocabulary identification unit 119 sorts the words in descending order of the set frequency information as the vocabularies with the highest priority to be expanded to the kansei vector, and finally, the vocabularies corresponding to the product DB attribute item vocabulary are arranged. Sort by Then, in the hybrid search unit 120, vocabularies with a high appearance frequency in the descriptions of the search target product DB 101 are preferentially searched for by sensitivity vectors.

(E) Others The technology disclosed herein is not limited to the above-described embodiments, and various modifications can be made without departing from the spirit of the embodiments. Each configuration and each process of the present embodiment can be selected as required, or may be combined as appropriate.

For example, in the above-described embodiment, an example in which the search target product DB 101, the product DB attribute item vocabulary DB 102, the emotion vector definition dictionary 103, and the description sentence vocabulary frequency information DB 104 are provided in the management server 10 was shown (see FIG. 4). ), but not limited to: At least a part of the search target product DB 101, product DB attribute item vocabulary DB 102, emotion vector definition dictionary 103, and explanation sentence vocabulary frequency information DB 104 may be provided in another server computer, and various modifications may be made. can.

Also, in the above-described embodiment, an example of a product search system that handles clothes as products has been shown as a search system, but the present invention is not limited to this, and various modifications can be made.

Furthermore, in the above-described embodiment, an example of a three-dimensional product sensitivity vector having "traditional", "casual", and "familiar" as sensitivity axes is shown, but it is not limited to this. Instead, it can be implemented with various modifications. For example, the number of dimensions may be two or less or four or more, and sensitivity axes other than "traditional," "casual," and "familiar" may be provided.

Moreover, the present embodiment can be implemented and manufactured by those skilled in the art based on the above disclosure.

(F) Supplementary Note The following Supplementary Note will be disclosed with respect to the above embodiment.

(Appendix 1)
estimating a plurality of first words that can be developed into emotional information as expressions containing ambiguity in the plurality of words included in the query;
Prioritizing the plurality of first words;
A feeling search program for causing a computer to execute a process of performing a feeling search on search target information for a first word selected from the plurality of first words according to the priority.

(Appendix 2)
estimating a second word other than the first word in a plurality of words included in the query;
The Kansei search program according to Supplementary Note 1, causing the computer to execute a process of performing a keyword search on the search target information for the second word, and integrating and outputting the result of the Kansei search and the result of the keyword search.

(Appendix 3)
3. The sensibility search program according to appendix 1 or 2, causing the computer to execute a process of estimating the first word based on the appearance frequency in the search target information.

(Appendix 4)
an input unit that accepts queries;
a first estimation unit for estimating a plurality of first words that can be developed into sensitivity information as expressions containing ambiguity in the plurality of words included in the query;
a setting unit that sets a priority order for the plurality of first words;
An information processing apparatus, comprising: a sensitivity search unit that performs a sensitivity search on search target information for a first word selected from the plurality of first words according to the priority order.

(Appendix 5)
a second estimation unit for estimating a second word other than the first word in a plurality of words included in the query;
a keyword search unit that performs a keyword search on the search target information for the second word;
5. The information processing apparatus according to appendix 4, further comprising an output processing unit that integrates and outputs the result of the sensitivity search and the result of the keyword search.

(Appendix 6)
The first estimation unit
6. The information processing apparatus according to appendix 4 or 5, wherein the first word is estimated based on the appearance frequency in the search target information.

(Appendix 7)
estimating a plurality of first words that can be developed into emotional information as expressions containing ambiguity in the plurality of words included in the query;
a process of prioritizing the plurality of first words;
and performing a sensitivity search on search target information for a first word selected from the plurality of first words according to the priority.

(Appendix 8)
A process of estimating a second word other than the first word in a plurality of words included in the query;
a process of performing a keyword search for the search target information for the second word;
7. The kansei search method according to appendix 7, further comprising a process of integrating and outputting the result of the kansei search and the result of the keyword search.

(Appendix 9)
The sensibility search method according to appendix 7 or 8, comprising a process of estimating the first word based on the appearance frequency in the search target information.

1 Information Processing System 10 Management Server 20-1, 20-2, 20 Client Terminal 30 Network 11 CPU
12 memory 13 storage device 14 network interface 101 search target product DB
102 Product DB attribute item vocabulary 103 Kansei vector definition dictionary 104 Description sentence vocabulary frequency information DB
111 Product DB attribute item vocabulary extraction unit 112 Description vocabulary extraction unit 113 Description sentence vocabulary frequency calculation unit 114 Kansei vector calculation unit 115 User query input unit 116 Input text analysis unit 117 DB attribute item vocabulary identification unit 118 Vocabulary frequency setting unit 119 Kansei Vector expansion vocabulary identification unit 120 Hybrid search unit 121 Search control unit 122 Kansei vector expansion unit 123 Kansei search execution unit 124 Keyword search execution unit 125 Search result integration unit 126 Search result display unit 501 Interface screen 502 User query input box 503 Search result display screen

Claims (4)

Vocabularies belonging to an attribute item vocabulary set, which are pre-extracted vocabularies used as attribute items of a database to be searched in a plurality of words included in a query, are presumed to be vocabularies specifying essential conditions that a user cannot remove, and With respect to words belonging to the attribute item vocabulary set other than the attribute item vocabulary set, those whose frequency of appearance exceeds a predetermined threshold are estimated as a plurality of first words that can be developed into sensibility information as expressions containing ambiguity; estimating words below the predetermined threshold as vocabulary specifying the essential condition ;
Prioritizing the plurality of first words;
A feeling search program for causing a computer to execute a process of performing a feeling search on search target information for a first word selected from the plurality of first words according to the priority. performing a keyword search on the search target information for the vocabulary specifying the essential condition ;
2. The Kansei search program according to claim 1, causing said computer to execute a process of integrating and outputting the result of the Kansei search and the result of the keyword search. an input unit that accepts queries;
in the plurality of words included in the query, vocabularies belonging to an attribute item vocabulary set obtained by extracting in advance vocabularies used as attribute items of a database to be searched are presumed to be vocabularies specifying essential conditions that a user cannot remove; Among the words belonging to other than the attribute item vocabulary set, those whose appearance frequency exceeds a predetermined threshold are estimated as a plurality of first words that can be developed into sensitivity information as expressions containing ambiguity, while the appearance frequency is equal to or less than the predetermined threshold, an estimation unit for estimating vocabulary specifying the essential condition ;
a setting unit that sets a priority order for the plurality of first words;
An information processing apparatus, comprising: a sensitivity search unit that performs a sensitivity search on search target information for a first word selected from the plurality of first words according to the priority order. the computer
Vocabularies belonging to an attribute item vocabulary set, which are pre-extracted vocabularies used as attribute items of a database to be searched in a plurality of words included in a query, are presumed to be vocabularies specifying essential conditions that a user cannot remove, and With respect to words belonging to the attribute item vocabulary set other than the attribute item vocabulary set, those whose frequency of appearance exceeds a predetermined threshold are estimated as a plurality of first words that can be developed into sensibility information as expressions containing ambiguity; A process of estimating words below the predetermined threshold as vocabulary specifying the essential condition ;
a process of prioritizing the plurality of first words;
and performing a sensitivity search on search target information for a first word selected from the plurality of first words according to the priority.

Images