JP2011257891A - Search apparatus - Google Patents

Abstract

PROBLEM TO BE SOLVED: To construct a search system capable of outputting a search result on a medium suited to a query by effectively using an existing search system.SOLUTION: A query input unit 20 inputs a query into various types of search engines E. A similarity calculation unit 36 calculates multimedia similarity between contents of the search result from a Web search engine E1, or a reference engine, and contents of the search result from other search engines E. A search result replacement unit 38 generates a search result for an output by replacing the search result of the reference engine with the search result from the other search engine E based on the multimedia similarity.

Description

  The present invention relates to a technique for retrieving data related to a query.

  In a search technique using a query, content existing on the web is collected in advance by a technique called crawling and stored in a database of a search system. At the time of accumulation, each data is indexed (index) and weighting indicating the degree of association between the keyword and the content is performed.

  The types of content to be searched include a wide range of media such as text (character strings), still images, moving images (hereinafter simply referred to as “moving images”), audio, etc., and data is retrieved using a search system corresponding to each content. Processing such as collection is performed. Then, using the degree of association (weight value) with the query (keyword) input by the user, a ranking is assigned to each content to generate a search result.

  As a technique for searching for a plurality of media, a technique for performing a ranking search with respect to a search condition from a search target database holding a plurality of documents and images to be searched is known (see Patent Document 1).

JP 2004-94813 A

  The degree of relevance between the keyword and the content described above differs in calculation algorithm depending on the type of media, and is generally managed and stored by a search system corresponding to each media.

  However, as in the patent literature, when a search system is realized using a plurality of media as search targets, it is necessary to integrate by a process such as normalizing the relevance calculated by different algorithms. Becomes complicated. In addition, while existing search systems such as text search, image search, and video search calculate the relevance with an algorithm suitable for each media, it is costly to develop a new algorithm that integrates the relevance. It was.

  In addition, in the case of a search engine for each existing medium, it is impossible to know which media the information desired by the user is expressed unless the search results from a plurality of search engines are compared. On the other hand, it was difficult to select which media was appropriate.

  The present invention has been made in view of the above-described problems, and the object of the present invention is to build a search system that can output a search result in a medium suitable for a query by utilizing an existing search system. It is to be.

  In order to solve the above-described problem, the first aspect of the present invention is to obtain a search result by outputting a query to a plurality of search engines that search for content including various media and acquiring the search result from each search engine. And a search result from a search engine for which the content including a plurality of various media as a search target is used as a reference, and the similarity between each content of the search result serving as the reference and the content of the search result from another search engine A similarity calculation means for calculating the search, and a search for generating a search result for output by replacing the content of the reference search result with a search result by another search engine based on the calculated similarity And a result generating means.

  According to the first aspect, a search result for output is generated by replacing a search result from a reference search engine with a search result from another search engine based on the similarity between contents. For this reason, the search results of a plurality of search engines are integrated based on the media characteristics of the reference content. Therefore, it is possible to construct a search system that can output a search result in a medium suitable for a query by utilizing an existing search system.

  Also, in the second aspect of the present invention, the content including a plurality of the various media is web content, and the similarity calculation means uses the search result from a search engine for searching the web content as the reference. The search result generation means is configured to replace the web content as the search result with the content of the search result by another search engine.

  According to the second aspect, search results from a plurality of search engines are replaced with reference to web content including a plurality of media such as text, still images, moving images, and sounds. For this reason, it is possible to integrate search results of search engines that search for media included in web content.

  In the third aspect of the present invention, the various media include at least one of text, a still image, a moving image, and audio, and the similarity calculation unit is included in the content acquired from each search engine. The similarity is calculated by comparing feature quantities of text, still image, moving image, and audio between the contents.

  In addition, according to the third aspect, since the similarity is calculated based on the text, still image, moving image, and audio feature amounts included in the content, it is possible to perform replacement with search results having similar characteristics of various media. .

  ADVANTAGE OF THE INVENTION According to this invention, the search system which can output the search result in the medium suitable for a query can be constructed | assembled using the existing search system.

The block diagram which shows an example of a function structure of a search server. The figure which shows the data structural example of an index. The flowchart which shows the specific processing content of search result integration processing. The 1st conceptual diagram for the illustration of calculation of a multimedia similarity. The figure which shows an example of a multimedia similarity. Display example of search results. The 2nd conceptual diagram for the illustration of calculation of a multimedia similarity.

[Apparatus configuration of this embodiment]
An embodiment when the search device of the present invention is applied to the search server shown in FIG. 1 will be described with reference to the drawings. The apparatus of this embodiment performs a web search using a query.

  A block diagram showing an example of the functional configuration of the search server 1 of the present embodiment is shown in FIG. The search server 1 is connected to the user terminal T via a communication network such as the Internet so as to be able to communicate with each other.

  The user terminal T is a terminal having an input function for a user to input a query for search and an output function for displaying and outputting a search result corresponding to the query, and includes a CPU, an input device, a display device, and the like. This is realized by a personal computer or a portable terminal.

  The search server 1 performs a search based on the query input at the user terminal T, and returns the search result to the user terminal T.

  The search server 1 is configured to include a search engine E for searching various contents, generates search results for each content in response to a query input at the user terminal T, and integrates them to the user terminal. Return to T.

  As shown in FIG. 1, the search server 1 includes a query receiving unit 10, a query input unit 20, a search result integration unit 30, a web search engine E1, an image search engine E3, and a video search engine as various search engines E. E5, a product search engine E7, and a search result output unit 40.

  The query receiving unit 10 receives and receives a search query from the user terminal T. The query is composed of a single keyword or a combination of a plurality of keywords.

  The query input unit 20 outputs the query received by the query receiving unit 10 to each search engine E, thereby causing the search processing in each search engine E to be performed.

  The search result integration unit 30 integrates the search results from the search engines E obtained by the query output of the query input unit 20 and generates a search result that is finally output to the user terminal T. As shown in FIG. 1, the search result integration unit 30 includes a reference engine setting unit 32, a search result acquisition unit 34, a similarity calculation unit 36, and a search result replacement unit 38. Will be described later.

  The search result output unit 40 is for outputting the search result generated by the search result integration unit 30 to the user terminal T. Specifically, the search result output unit 40 is display data that can access the image data of the search result (for example, HTML). Web page data) is generated and transmitted to the user terminal T.

  The various search engines E compare the keyword included in the query with the keyword stored in the index D to obtain a search result for the query. Here, the web search engine E1 searches for web content including media such as text, images, and moving images. The image search engine E3 searches for still images, and the moving image search engine E5 searches for moving images. The product search engine E7 searches product information posted on the sales site.

  Each search engine E collects content on the Internet by crawling in so-called robot search, and indexes the content. For example, the web search engine E1 collects web pages by circulating URLs posted on the web pages, and indexes the URLs using keywords included in the web pages.

  The image search engine E3 collects image data from the image posting site, and indexes the URL of the image posting site based on tag information (keyword) given to the image data. At this time, the tag information may be set by dividing text information described in the vicinity of the image into words by morphological analysis or the like on the web page on which the image is posted. Similarly, the video search engine E5 and the product search engine E7 perform crawling and indexing to create the index D.

  As shown in FIG. 1, each search engine E has an index D (D1, D3, D5, D7).

  The index D is a so-called transposed index, and as shown in FIG. 2, shows an index keyword, the URL of a web page (content data) on which content is posted, and the relationship between the keyword and content data. It is a database that stores weight values and content data in association with each other. At the time of the above crawling, each search engine E indexes and stores a keyword (including tag information) extracted from the content data in the URL and content data collected by crawling.

  Further, a weight value indicating the degree of association between the keyword and the content data is calculated by TF / IDF or the like and stored in association with it.

  When a query is input from the query input unit 20, each search engine E searches the index D based on the query and generates a list of URLs as a search result. When generating this list of search results, ranking is given to the URLs of the search results in descending order of the weight value, which is the degree of association of the web page with the query.

  As a ranking algorithm of the search engine E, in addition to the method of using the importance of the keyword in the web page as in the above TF / IDF, the importance of the web page by analyzing the link relation between the web pages. You may combine the method of using. Further, as the weight value of the web page, the position of the keyword (title, text, etc.) in the web page, the size of the font, and the like may be taken into account, and a known technique can be adopted as appropriate.

  As the index D, only the document ID indicating the keyword and the content data and the weight value are stored in association with each other, and the URL and the content data are stored separately in association with the document ID in another database. Also good.

[Detailed description of the search result integration unit]
Next, a specific operation of the search result integration unit 30 will be described with reference to FIGS.

  First, the reference engine setting unit 32 of the search result integration unit 30 selects and sets one search engine E as a reference engine from among a plurality of search engines E (step S11). The reference engine is a search engine E that is used as a reference for ranking when integrating the search results output from the search engines E, and one of the plurality of search engines E is selected. 4 to 6, the web search engine E1 is set as the reference engine. The reference engine may be set by the search server 1 or may be set by a user who inputs a query.

  Then, the search result acquisition unit 34 acquires a predetermined number of search results searched based on the query output by the query input unit 20 (for example, the top N search results, N is an arbitrary natural number) from each search engine E. To do. At this time, a list of URLs as search results and content data indicated by the URLs are also acquired (step S12).

  For example, in the example of FIG. 4, a search result group A as a result of searching the web search engine E1 with the query “Tokyo” is acquired. Further, a search result group B as a result of searching for “Tokyo” is acquired from the image search engine E3, a search result group C is acquired from the moving image search engine E5, and a search result group F is acquired from the product search engine E7.

  Next, the similarity calculation unit 36 calculates the similarity between the search results from the reference engine and the search results from the other search engines E (steps S13 to S14). The similarity calculated here is referred to as “multimedia similarity”.

  The multimedia similarity is an index that comprehensively considers the similarity between media such as text, images, and audio included in content data.

  In this embodiment, each item (one of the search results) in the search result list of the basic engine is compared with an item group (N search results) of the search result list of another search engine, Calculate multimedia similarity. Specifically, it is calculated based on the following equation.

[Formula 1]
Multimedia similarity = (text similarity + image similarity + video similarity) / (number of types of media included in content)

  The number of types of media included in the content is set according to the search target of each search engine E. For example, if the content includes text, images, and moving images as search target media as in the case of web search, the number is “3”. If it is an image search, “1” is set. Further, even in the case of image search or moving image search, text information (title, tag information, etc.) may be included in the search target. In this case, the number of types of media may be set to ‘2’. In the combination of search engines E that calculate multimedia similarity, the value with the smaller number of media types may be used, and the following text similarity, image similarity, and video similarity can be calculated (value is 0). You may set the number of media.

  The text similarity is a similarity between texts included in the content. For example, the similarity is obtained from the cosine distance between the vectors by expressing the feature amount of the text by a multidimensional vector based on the number of appearances of the keyword included in the text.

  As an example, assume that one search result in the search result group A of the reference engine is A1, and a search result group B of the image search engine E3 that is another engine. It is assumed that N contents are included in the search result group B, and each content in the search result group is represented by B1, B2, B3,.

  The text similarity between the search result A1 and the search result group B is obtained as follows.

[Formula 2]
Text similarity = MAX {similarity (text A1, text B1), similarity (text A1, text B2), ..., similarity (text A1, text BN)} / N

  MAX {} means a function for selecting the maximum value, and similarity () means a function for calculating the similarity. That is, the text similarity is obtained by averaging the maximum similarities between the most similar texts in the search result group.

  The image similarity is a similarity between images included in the content. This similarity is obtained, for example, from the Euclidean distance between the vectors by expressing the feature quantities such as colors, shapes, and patterns extracted from the images by multidimensional vectors. If the image similarity is obtained for the search result A1 of the reference engine E1 and the search result group B of the image search engine E3, the following equation is obtained.

[Formula 3]
Image similarity = MAX {similarity (image A1, image B1), similarity (image A1, image B2), ..., similarity (image A1, image BN) / N

  The moving image similarity is a similarity between moving images included in the content. This similarity is represented by, for example, image features such as colors, shapes, and patterns extracted from a moving image, and feature amounts such as object motion and audio signals expressed by multidimensional vectors. Calculated by distance. If the moving image similarity is obtained for the search result A1 of the reference engine E1 and the search result group B of the image search engine E3, the following equation is obtained.

[Formula 4]
Video similarity = MAX {similarity (video A1, video B1), similarity (video A1, video B2), ..., similarity (video A1, video BN) / N

  As described above, the text similarity, image similarity, and video similarity between the search results of the reference engine and the search results of other engines are obtained and divided by the number of media types included in the content. The multimedia similarity that is the average value of is calculated. Based on the multimedia similarity, it is possible to determine which media characteristics have a lot of search results of the reference engine. Further, by selecting the maximum value using MAX {} for calculation of each similarity, the multimedia similarity can be calculated from the most similar content data by comparing the search result groups.

  The calculated multimedia similarity is temporarily stored in a memory or the like with a data structure as shown in FIG. 5 in association with a combination of search results.

  In the case where a plurality of media of the same type are included in the content as in the search results A1 and A2 in FIG. 4, the maximum similarity is extracted for each of them as described above, and the average is extracted. A value may be used. Moreover, it is preferable that the value of each similarity is normalized with respect to the maximum similarity in each medium (for example, according to the type of text or image).

  The similarity calculation unit 36 selects one of the search results from the reference engine (step S13), and multimedia between the selected search result content data and a search result group from another search engine E. The similarity is calculated (step S14).

  For example, when the search result A1 from the web search engine E1 that is the reference engine is selected as shown in FIG. 4, the search result A1 and the search result groups B, C, and F from other search engines E Multimedia similarity between them is calculated.

  Based on the multimedia similarity calculated by the similarity calculation unit 36, the search result replacement unit 38 selects a search result of another search engine E that is a replacement candidate for the search result of the reference engine (step S15).

  Specifically, the highest search result in the search result group is selected from the search result group of the other search engine having the maximum multimedia similarity calculated for one search result of the reference engine. Select. For example, in FIG. 5, since the one with the highest multimedia similarity calculated with respect to the search result A1 of the web search is the image search, the highest search result B1 of the search result group B of the image search is selected.

  Then, the search result replacing unit 38 determines whether or not the selected search result has already been replaced as an output search result (step S17). If the result of this determination is that it has already been replaced (step S17; Yes), the process proceeds to step S15, and the next higher-order search result is replaced from the search result group having the maximum multimedia similarity. Select as a candidate.

  If it is determined that the replacement has not been completed (step S17; No), the search result of the replacement candidate other engine is stored in the memory as a search result for output (step S18).

  For example, when the image characteristic is high in the content of the search result of the reference engine like the search result A1, the upper search result B1 is selected and replaced as the search result for output among the search results of the image search. When the moving image feature is high as in the search result A2, the upper search result C1 is selected from the search results of the moving image search and replaced.

  The search result replacement unit 38 determines whether or not all search results from the reference engine have been selected for selection of search results in step S13, and if there is any unselected one (step S19; No), step S13. The process is transferred to. Thereby, the multimedia similarity is comprehensively calculated for each search result of the web search in combination with each of the search results of the image search, the video search, and the product search.

  In the example of FIG. 4, multimedia is obtained for each combination between N search results A1 to AN from the web search engine E1 and search result groups B, C, and F from other search engines as shown in FIG. Similarity is calculated.

  If the search result replacement unit 38 determines that all search results have been selected (step S19; Yes), the search result for output transmitted to the user terminal T based on the search result obtained by performing the replacement process in step S18. List data is generated (step S20). FIG. 6 is an example of the generated search result list for output. As shown in the figure, the search results R1 to RN for output are generated by replacing the search result A1 with the search result B1, the search result A2 with the search result C1, and the search result A3 with the search result B2. .

  As described above, according to this embodiment, the multimedia similarity between the search result in the search engine E set as the reference engine and the search result group in the other search engine E is calculated, A search result for output is generated by replacing the search result of the reference engine with the search result of another search engine E based on the media similarity. Accordingly, the search results from the other search engines E are integrated in consideration of the media characteristics of the search results from the reference engine. Therefore, a search system that integrates a plurality of media can be constructed by utilizing the existing search engine E.

  In addition, the multimedia similarity calculated between the individual search results of the reference engine and the search result group of other engines is used as an index of what kind of media each search result of the reference engine is expressed. be able to. Therefore, a search result for output is generated by a medium that well expresses the search result for the query. Therefore, the user can obtain a search result by an appropriate medium without performing a search / comparison with the search system for each medium.

[Modification]
The embodiment described above is an example to which the present invention is applied, and the applicable range is not limited to the above example.

  For example, in the above-described embodiment, the search result of the replacement candidate having the highest multimedia similarity is described as selecting the search result of the replacement candidate from the higher ones among the search results of the other engines having the higher multimedia similarity. Among them, a search result having a high degree of similarity with the search result of the reference engine may be selected as a replacement candidate.

  In this case, the multimedia similarity between each search result of the reference engine and the search result group of the other engine is calculated as shown in FIG. 5, and each search result of the reference engine is calculated as shown in FIG. The multimedia similarity is calculated between each search result of the engine. Then, among the search result groups of other engines with high multimedia similarity, the search results with high individual multimedia similarity are selected as replacement candidates.

  The calculation of the multimedia similarity between the search results is performed based on the above formula 1, and the text similarity, the image similarity, and the video similarity are calculated between the contents of the search results.

  In the example of FIG. 7, among the search result group B of the image search having a high multimedia similarity with the search result A1 in FIG. 5, the search result B2 having a high calculated multimedia similarity is selected as a replacement candidate. The Rukoto. For this reason, it is possible to select a search result having a higher content similarity as a replacement candidate from search engines having similar media characteristics of the search results of the reference engine.

  Alternatively, the multimedia similarity between the search result of the reference engine and the search result of the other engine may be calculated as shown in FIG. 7, and the one with the high multimedia similarity may be selected as a replacement candidate. Thereby, a search result for output can be generated based on a search result having high similarity among search results of other engines.

  Further, in the above embodiment, the description has been given that the search result is replaced with the one having the maximum multimedia similarity. However, when all the multimedia similarities are less than a predetermined threshold (for example, 0.5), The reference engine search result may be used as an output search result. That is, when the media characteristics are not similar to the search results of other search engines E, the search results of the reference engine can be used for output without replacement.

  In addition, a search engine E for searching a composite medium including a plurality of media such as a web search and a product search as a search target rather than a search engine E as a search target for a content by a single medium such as an image search or a video search. It is desirable to set as a reference engine. By using the search engine E that searches for composite media as a reference engine, multimedia similarity can be calculated from the viewpoint of multiple media such as text, images, and videos included in the content. It is possible to realize a search that integrates.

  Further, when the search result from the reference engine is replaced with the search result of another search engine E, the search result to be replaced may be selected so that the search result to be replaced does not overlap. Specifically, the above search results are replaced from the top of the search results of the reference engine, and the search results of the other search engines E that have performed the replacement are stored in a memory or the like. Then, when replacing a search result lower than the reference engine, if the maximum search result has already been replaced, replacement is performed using the maximum search result. As a result, it is possible to prevent duplicate contents from being included as search results for output.

  In addition, when determining the replacement based on the multimedia similarity, the replacement may be performed in consideration of the ranking by weighting the ranking given to each search result. For example, when the rankings 1 to N of the search results are ranked, the multimedia similarity calculated for each search result of the other search engine E is set to 1/1, 1/2, Weighting is performed by multiplying 1/3,..., 1 / N weight values.

  In this way, by weighting the ranking with respect to the multimedia similarity, it is possible to replace the search results with the search results highly relevant to the query among the search results of each search engine E.

  In addition, as an expression for calculating the multimedia similarity, the average value of the text similarity, the image similarity, and the moving image similarity has been described as in Expression 1, but the maximum value or the minimum value among them is used. Also good.

  Specifically, the maximum value among the calculated text similarity, image similarity, and moving image similarity may be set as the multimedia similarity. According to this, the search result can be replaced by the multimedia similarity that places importance on the high similarity media.

  A value obtained by multiplying the text similarity, the image similarity, and the moving image similarity may be set as the multimedia similarity. According to this, since the multimedia similarity value does not increase unless all the similarity values are increased, the replacement can be performed by the search result in which the respective media are totally similar.

  In addition, the voice search engine may be included as a search target medium by adding the voice similarity to the multimedia similarity using a voice search engine that searches voice data such as music and recorded voice. It is also possible to use search engines for other media such as news, blogs and maps.

  The operation of the embodiment can be performed by incorporating appropriate computer software into the computer. The contents of the present invention are not limited to the above embodiment. In the present invention, various modifications can be made to the specific configuration within the scope of the claims.

  For example, each component only needs to exist as a functional block, and does not need to exist as independent hardware. As a mounting method, hardware or computer software may be used. Furthermore, one functional element in the present invention may be realized by a set of a plurality of functional elements, and a plurality of functional elements in the present invention may be realized by one functional element.

  Moreover, the functional element may be arrange | positioned in the position physically separated. In this case, the functional elements may be connected by a network. It is also possible to realize functions or configure functional elements by grid computing.

D Index E Search engine E1 Web search engine E3 Image search engine E5 Video search engine E7 Product search engine T User terminal 1 Search server 10 Query reception unit 20 Query input unit 30 Search result integration unit 32 Reference engine setting unit 34 Search result acquisition unit 36 similarity calculation unit 38 search result replacement unit 40 search result output unit

Claims (5)

A search result acquisition means for outputting a query to a plurality of search engines that search for contents including various media and acquiring search results from each search engine;
Based on a search result from a search engine that searches for content including a plurality of the various media, the similarity between each content of the search result serving as the reference and the content of the search result from another search engine is calculated. Similarity calculation means;
Search result generating means for generating a search result for output by replacing the content of the reference search result with a search result by another search engine based on the calculated similarity;
A search device comprising: The content including a plurality of the various media is web content,
The similarity calculation means includes:
A search result from a search engine that searches the web content is set as the reference,
The search result generating means includes
The search apparatus according to claim 1, wherein the web content that is the search result is replaced with content that is a search result from another search engine. The various media include at least one of text, still images, moving images, and sounds,
The similarity calculation means includes:
3. The similarity is calculated by comparing text, still image, video, and audio feature amounts included in the content acquired from each search engine between the content. Search device. Computer
A search result acquisition step of outputting a query to a plurality of search engines that search content including various media and acquiring search results from each search engine,
Based on a search result from a search engine that searches for content including a plurality of the various media, the similarity between each content of the search result serving as the reference and the content of the search result from another search engine is calculated. Similarity calculation step;
A search result generation step of generating a search result for output by replacing the content of the reference search result with a search result by another search engine based on the calculated similarity,
The search method characterized by performing.   A program for causing a computer to execute the search method according to claim 4.

Images