JP2013200862A - Method and device for diversifying query results - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a method and device for diversifying query results.SOLUTION: In the method for diversifying the query result, a combination set of related keywords is constructed by extracting the related keywords from domain ontology for respective query keywords of a query, a query result set is acquired by performing search by the combination of the respective related keywords in the combination set of the related keywords, the corresponding number of query results are extracted from the respective query result sets to construct a final query result set, and the constructed final query result set is ranked to acquire diversified query results.

Description

  The present invention relates to an information search technique, and more particularly, to a method and apparatus for diversifying query results.

  Traditional information retrieval techniques typically diversify by post-processing or re-ranking retrieved documents, such as clustering or classifying search results, re-ranking results by means of mean variance analysis, etc. Done.

  With the development of information retrieval technology, user demands for diversification of search results and disambiguation of information retrieval queries are increasing. In particular, diversification of search results means that there may be several interpretations of query keyword input by the user, and the retrieved query results must cover these different interpretations. To do. The purpose of search result diversification is to reduce as much as possible the risk that users are not satisfied by balancing the relevance and novelty of search results. Query disambiguation focuses on identifying all possible query intents based on user input keywords and expressing the intent in a more accurate manner.

  Query disambiguation serves as a new way to support search diversification, which can effectively save computational costs and make the results more understandable, especially when the results are large in scale. Diversified searches are typically performed in the prior art by using statistical analysis (or machine learning, etc.) in the query log.

  Specifically, an existing method for diversifying query results adopts a form of conversion between queries as shown in FIG. 1 and includes the following.

  Step S101 is to generate k related queries R (Q) for a given query Q by analyzing a large sample of query logs.

  Step S102 is to obtain an initial DOC list by extracting n / (k + 1) (n represents the number of user documents) results from each set of query results. And

  Step S103 is to re-rank the first DOC list with the relevance feedback technique.

  A corresponding device for diversifying query results as shown in FIG.

  A query unit 201 configured to store user query keywords;

  A query log memory unit 202 configured to store user query logs;

  A query disambiguation unit 203 configured to determine query keywords related to the target query according to the user's query keywords and query log;

  A subquery memory unit 204 configured to store query keywords associated with the target query;

  A document memory unit 205 configured to store the searched documents;

  A keyword search unit 206 configured to search for documents in the document memory unit 205 using the subquery keywords;

  A subquery result memory unit 207 configured to store query results of a search by each subquery;

  A query result combining unit 208 configured to combine the respective query results;

  A query result memory unit 209 configured to store the combined query results;

  A query result ranking unit 210 configured to rank the combined query results;

  A diversified rank list memory unit 211 configured to store a final diversified query result of the target query.

Specifically, for example, given a query keyword “window” and a target query q = (window), sub-query keywords are acquired as “window XP”, “house window”,. A set of sub-queries for q is R (q) = {(q ₁ , q, window XP), (q ₂ , q, house window),...}, a search by the target query q, and a set of sub-queries A search by each subquery in R (q) is performed by one set of document list S (q) = {(q, document list 1), (q ₁ , document list 2), (q ₂ , document list 3),. ...} are obtained respectively, and n / (k + 1) to form a new set of query results RF (q) for q. Document is selected from each of the document list. Where n is a predetermined value representing the scale of the results, k represents the number of subqueries, the documents in RF (q) are ranked based on the degree of match between the document and the user's interest, Thereby, a diversified query result of the user query is acquired.

  As can be seen from the above method of diversifying query results, a set of sub-queries is determined on the basis of the query log in the prior art, but the inventor of the present invention has now identified that user's Query logs are generated based on user input query keywords that may not accurately represent the intent of the true query, and cannot be used in company search or other search situations, or the size of the query log The query log may be an unreliable data source, and therefore produces diversified query results that are not accurate, because it may not be sufficient to support query disambiguation is there.

  Embodiments of the present invention provide a method and apparatus for diversifying query results to obtain accurate diversified query results.

A method for diversifying query results according to a preferred aspect of the present invention comprises:
Building a set of related keyword combinations by extracting related keywords from the domain ontology for each query keyword;
Searching for each related keyword combination in the related keyword combination set to obtain a query result set;
Constructing a final query result set by extracting a corresponding number of query results from each query result set;
Ranking the constructed final query result set to obtain diversified query results.

An apparatus for diversifying query results according to a preferred aspect of the present invention comprises:
A related keyword combination set building unit configured to build a set of related keyword combinations by extracting related keywords from the domain ontology for each query keyword;
A query unit configured to search for each related keyword combination in the related keyword combination set to obtain a query result set;
A final query result set construction unit configured to construct a final query result set by extracting a corresponding number of query results from each query result set;
A ranking unit configured to rank the constructed final query result set to obtain diversified query results.

  According to the method and apparatus for diversifying query results of the present invention, a set of related keywords is determined for a given keyword query in a domain ontology, and the query set expanded by the combination of these related keywords. Is constructed, thereby avoiding an extended query set being determined from untrusted query logs, and thus diversified query results are more accurate.

3 is a flowchart of a method for diversifying query results in the prior art. FIG. 6 is a schematic structural diagram of an apparatus for diversifying query results in the prior art. 4 is a flowchart of a method for diversifying query results according to an embodiment of the present invention. 4 is a flowchart of a method for obtaining a minimum subgraph according to an embodiment of the present invention. 4 is a flowchart of a method for determining a set of query results according to one embodiment of the present invention. 4 is a flowchart of a method for obtaining a query result according to an embodiment of the present invention; 4 is a flowchart of a ranking method according to an embodiment of the present invention. 3 is a flowchart of a method of ranking by similarity according to an embodiment of the present invention; FIG. 3 is a schematic structural diagram of an apparatus for diversifying query results according to an embodiment of the present invention.

  Embodiments of the present invention provide a method and apparatus for diversifying query results, wherein a set of related keywords is determined for a given keyword query in a domain ontology, and the combination of these related keywords. Is used to build an expanded query set, thereby avoiding determining an expanded query set from untrusted query logs, thus making diversified query results more accurate.

  As shown in FIG. 3, a method for diversifying query results according to one embodiment of the present invention includes:

  Step S301 is to construct a combination set of related keywords by extracting the related keywords from the region ontology for each query keyword.

  Step S302 is to perform a search with each combination of related keywords in the combination set of related keywords to obtain a query result set.

  Step S303 is to construct a final query result set by extracting a corresponding number of query results from each query result set.

  Step S304 is to rank the constructed final query result set to obtain diversified query results.

  Each related keyword is extracted from the region ontology, thus making the selection of related keywords more accurate. In addition, by interpreting a given query by building a query graph based on a combination of related keywords, the interpretation is closer to the user's real intentions, thereby diversifying query results more accurately . Domain ontology represents domain knowledge as a set of concepts along with relationships between concepts.

Specifically, in step S301, for each keyword of a given query, a keyword related keyword is extracted from the region ontology, and then a combination set of related keywords is constructed from the related keywords. The combination set of related keywords constructed is S (Q) = {(c ₁ , c ₂ ,..., C _m ) | c ₁ ∈C ₁ && c ₂ ∈C ₂ && ... c _m ∈C _m } Here, C _i is a related keyword set of the i-th keyword among the m keywords in the given query.

  When extracting a related keyword of a given query keyword, word matching can be performed between the query keyword and the ontology concept, and then the acquired matching ontology concept is set as the related keyword. Of course, those skilled in the art can extract related keywords from the region ontology by other methods.

  Related keyword combinations can be further filtered to make the query results more accurate, thereby getting closer to the user's intention.

  Specifically, in step S301, after building a related keyword combination set for a given query according to the region ontology,

  For each combination of related keywords in the combination set of related keywords, a minimum subgraph connecting all the keywords in the combination is extracted from the region ontology. Here, the minimum subgraph is a subgraph that has the minimum number of edges in the subgraph connecting each keyword in the region ontology and connects all related keywords in the closest way.

  As shown in FIG. 4, given five keywords in a related keyword combination, the extracted subgraph connects all five keywords and has a minimum number of edges.

  At this time, as shown in FIG. 5, step S302 for obtaining a query result set by searching for each combination of related keywords in the combination set of related keywords includes the following.

  Step S501 is to determine, for each minimum subgraph, a subquery including a keyword and other nodes in the minimum subgraph.

  In step S502, a search is performed with keywords and other nodes in each subquery, and the same number of subquery result sets as the number of minimum subgraphs are obtained. And

  Step S503 is to determine the query result set as a set composed of the respective subquery result sets.

であり、ユーザの入力クエリについて関連キーワードの組み合わせ、Ｓ（Ｑ）＝｛（ｃ１，ｃ２，…，ｃｍ）｜ｃ１∈Ｃ１＆＆ｃ２∈Ｃ２＆＆…ｃｍ∈Ｃｍ｝を決定することができる。 For example, a user inputs a query including _m keywords, Q = {k ₁ ,..., K _m }, and a set of related keywords for an arbitrary keyword k _i , C _i = {c _i1 , c _{i2,. ...,} it is possible to extract the _{c ini}} from the region ontology, ni represents the number of relevant keywords for _{k i,} also relevance values for each related keyword for _{k i, R i = {r} i1, r _{i 2} ,..., r _ini } can be calculated based on the region ontology. In total

And a combination of related keywords for the user's input query, S (Q) = {(c1, c2,..., Cm) | c1εC1 && c2εC2 && ... cmεCm} can be determined.

  For each combination of related keywords, the smallest query semantic graph can be determined from the region ontology, where the smallest query semantic graph contains each keyword in the combination, each keyword is a node of the query semantic graph, and the smallest The query semantic graph may also include other nodes for each keyword that is combined. A minimal query semantic graph, which can also be expressed as a subquery of a given user query, is a subgraph having the smallest number of edges in the subgraph connecting the respective keywords.

  In a combination of related keywords, select a keyword randomly, set the selected keyword as the first node where traversal begins, traverse each path that connects the keyword to another node in the region ontology, A minimum subgraph can be obtained by selecting the shortest path to the target node as the path in the minimum subgraph until the minimum subgraph to be connected is determined. Here, when there are two paths having the same path length, one of them can be selected at random. The target node here indicates a keyword in a combination of related keywords.

  In step S303, there are two methods for constructing a final query result set by extracting a corresponding number of query results from each query result set. The first method is to extract a preset number of query results from each subquery result set. The second method is to extract a corresponding number of query results from each subquery result set according to the relevance of the subquery for a given query, so that the more the subquery is related to a given query, More results will be selected from the subquery result set and added to the final query result set.

  Specifically, as shown in FIG. 6, according to the relevance of each subquery to a given query, a corresponding number of query results can be obtained from each subquery result set, particularly as follows.

として決定することであり、式中ｍはクエリキーワードの数を表し、ｒｉはクエリキーワードに対するクエリキーワードについて領域オントロジから抽出された関連キーワードのマッチ値を表し、｜Ｅ｜はサブグラフにおけるエッジの数を表す。 Step S601 calculates the subgraph weight of each minimum subgraph.

Where m represents the number of query keywords, ri represents the match value of the related keyword extracted from the region ontology for the query keyword for the query keyword, and | E | represents the number of edges in the subgraph. Represent.

  Step S602 is to extract a corresponding number of query results from the subquery result set corresponding to each minimum subgraph according to the subgraph weight of the minimum subgraph.

  In step S602, a corresponding number of query results can be obtained from the subquery result set corresponding to each minimum subgraph, particularly as follows, according to the subgraph weights of the minimum subgraph.

  The query results extracted from the subquery result set corresponding to the smallest subgraph are the top k * a results, where a is the ratio of the subgraph weight of the current smallest subgraph to the sum of the subgraph weights of all smallest subgraphs. Where k is a predetermined value representing the scale of the final query result set, and k * a represents the largest integer less than the product of k and a.

  Furthermore, one embodiment of the present invention provides a method for ranking query results so that the user can more conveniently view query results that are closer to the intent of the query, as shown in FIG. Step S304 of ranking the constructed final query result set to obtain diversified query results includes:

  Step S701 is to determine the value of the relevance of the query result for the corresponding minimum subgraph for each query result in the final query result set.

  Step S702 is to determine the weight of the query result according to the relevance value of the query result for the corresponding minimum subgraph and the subgraph weight of the minimum subgraph for each query result in the final query result set. And

  Step S703 is to obtain the diversified query results by ranking the query results in the final query result set according to the query result weights.

  Specifically, in step S702, the weight of the query result can be determined according to the relevance value of the query result with respect to the corresponding minimum subgraph and the subgraph weight of the minimum subgraph as follows.

  The query result weight is determined as the product of the relevance value of the query result for the corresponding minimum subgraph and the subgraph weight of the minimum subgraph.

  Further, in step S703, by directly ranking the query results obtained according to the query result weights, or further taking into account the similarity between the query results, the obtained query results are made according to the query result weights. At this time, as shown in FIG. 8, step S703 specifically includes the following:

  Step S801 is to determine the query result having the highest weight as the query result ranked first, and to determine the similarity between each two query results.

として決定することであり、式中ｓはクエリ結果の重みを表し、ｄは現在のクエリ結果を表し、Ｄはすでにランク付けされている１組のクエリ結果を表し、類似性（ｄ，ｄ´）はｄとｄ´との間の類似性を表す。そして Step S802 calculates the similarity weight of each query result to be ranked.

Where s represents the weight of the query result, d represents the current query result, D represents the set of query results that are already ranked, and the similarity (d, d ′ ) Represents the similarity between d and d '. And

  Step S803 is to repeatedly rank the query results excluding the query result ranked first by the similarity weight.

  A specific example illustrating a method for diversifying query results according to an embodiment of the present invention is described below.

  When the user's query keyword is “peony” and “Beijing”, C (“peony”) = {(“peony flower”, 0.5), (“peony TV”, 0.2), (“peony river” , 0.2), ...} and C ("Beijing") = {("Beijing City", 0.8) ("Beijing Watch", 0.07), ("Beijing Events", 0.05), ...} can be determined from the region ontology, where ("peony flower", 0.5) represents the match value of the related keyword "peony flower" of "peony" with respect to "peony".

  After the combination of each related keyword is determined, a subgraph connecting each keyword is acquired. For example, one set of minimum subgraphs is S (graph) = {(g1, peony flower, Beijing, 0.65), (G2, peony TV, Beijing, 0.5), (g3, peony flower, li commuter, beijing affair, 0.138), ...}, and the subgraph g1 is 0.65 so that it can be easily derived. G2 has a subgraph weight of 0.5 and g3 has a subgraph weight of 0.138.

  For each set of subquery results, eg, here for each document in the set of subquery results, by searching with keywords and other nodes in each subgraph, the result (g1) = {(doc1, ωg = 0.65, ωr = 0.9), (doc2, ωg = 0.65, ωr = 0.7),...}, Result (g2) = {(doc3, ωg = 0.5, ωr = 0.8), (doc4, ωg = 0.5, ωr = 0.6)..., Wg represents the subgraph weight of the smallest subgraph, wr represents the relevance of the document to the smallest subgraph, and the document in each set of subquery results is ranked by wr.

個の文書が結果（ｇ１）から選択され、最終的なクエリ結果セットＲＦ（ｑ）に追加され、上位

個の文書が結果（ｇ２）から選択され、最終的なクエリ結果セットＲＦ（ｑ）に追加される。ここで３は、ＲＦ（ｑ）の予想されるスケールを意味する。 The query results obtained from the subquery result set corresponding to the smallest subgraph are the top 3 * a query results most relevant to the subgraph, eg, top

Documents are selected from the results (g1) and added to the final query result set RF (q)

Documents are selected from the results (g2) and added to the final query result set RF (q). Here, 3 means the expected scale of RF (q).

  When RF (q) is {(doc1, 0.65, 0.9), (doc2, 0.65, 0.7), (doc3, 0.5, 0.8)},

  The obtained query results can be directly ranked by the weight of the query results, and the weights of the three documents are s1 = 0.65 × 0.9, s2 = 0.65 × 0.7, and s3 = 0, respectively. Since it is .5 × 0.8, the ranked query result is RF (q) = {doc1, doc2, doc3}.

  Alternatively, the retrieved query results can be further ranked using similarity, where similarity (doc1, doc2) = 0.5, similarity (doc1, doc3) = 0.1, and If similarity (doc2, doc3) = 0.2, the ranked query result is RF (q) = {doc1, doc3, doc2}.

  One embodiment of the present invention further provides an apparatus for correspondingly diversifying query results, as shown in FIG. 9, the apparatus includes:

  A related keyword combination set construction unit 901 configured to construct a related keyword combination set by extracting related keywords from the region ontology for each query keyword of a given query;

  A query unit 902 configured to search with each related keyword combination in the related keyword combination set to obtain a query result set;

  A final query result set construction unit 903 configured to construct a final query result set by extracting a corresponding number of query results from each query result set;

  A ranking unit 904 configured to rank the constructed final query result set to obtain diversified query results.

  In particular, the related keyword combination set construction unit 901

  For each query keyword of a given query, extract keyword related keywords from the domain ontology,

  A related keyword combination set is constructed based on each related keyword.

  The related keyword combination set construction unit 901 that constructs a combination set of related keywords based on each related keyword includes in particular:

The related keyword combination set is constructed as S (Q) = {(c ₁ , c ₂ ,..., C _m ) | c ₁ ∈ C ₁ && c ₂ ∈ C ₂ &&... C _m ∈ C _m }, where C _i is a related keyword set of the i-th keyword among m keywords in a given query.

  In particular, the related keyword combination set construction unit 901 further includes

  For each query keyword in a given query, after extracting the keyword related keywords from the domain ontology,

  After building a related keyword combination set for a given query keyword set in a domain ontology,

  For each combination of related keywords in the combination set of related keywords, a minimum subgraph connecting the keywords in the combination of related keywords is extracted from the region ontology. Here, the minimum subgraph is defined in the region ontology. It is a subgraph having the minimum number of edges in the subgraph connecting each keyword.

  Query unit 902 specifically includes

  For each minimal subgraph, determine a subquery consisting of keywords and other nodes in the minimal subgraph,

  Search by keyword and other nodes in each subquery to get as many subquery result sets as the number of minimum subgraphs,

  A query result set is configured to be determined as a set of respective subquery result sets.

  The final query result set construction unit 903 specifically

  Extract a corresponding number of query results from each subquery result set according to the relevance of each subquery to a given query,

  Configured to combine query results obtained from respective subquery result sets.

  Furthermore, the final query result set construction unit 903 specifically

として決定し、式中ｍはクエリキーワードの数を表し、ｒｉはクエリキーワードに対するクエリキーワードについて領域オントロジから抽出された関連キーワードのマッチ値を表し、｜Ｅ｜はサブグラフにおけるエッジの数を表し、 Subgraph weight for each minimum subgraph

Where m represents the number of query keywords, ri represents the match value of the related keyword extracted from the region ontology for the query keyword for the query keyword, | E | represents the number of edges in the subgraph,

  Extract a corresponding number of query results from the subquery result set corresponding to each minimum subgraph according to the subgraph weight of the minimum subgraph,

  Configured to combine query results obtained from respective subquery result sets.

  Specifically, the final query result set construction unit 903 that extracts a corresponding number of query results from the subquery result set corresponding to each minimum subgraph according to the subgraph weights of the minimum subgraph includes in particular:

  The query results extracted from the subquery result set corresponding to the smallest subgraph are the top k * a query results most relevant to the smallest subgraph, where a is the current minimum for the sum of the subgraph weights of all the smallest subgraphs. It represents the ratio of the subgraph weight of the subgraph, k is a predetermined value representing the scale of the final query result set, and k * a represents the largest integer less than the product of k and a.

  In particular, the ranking unit 904

  For each query result in the final query result set, determine the relevance value of the query result for the corresponding smallest subgraph,

  For each query result in the final query result set, determine the query result weight according to the relevance value of the query result for the corresponding minimum subgraph and the subgraph weight of the minimum subgraph,

  A query result weight is configured to rank the query results in the final query result set to obtain a diversified query result.

  The ranking unit 904 that determines the weight of the query result according to the value of the relevance of the query result for the corresponding minimum subgraph and the subgraph weight of the minimum subgraph includes in particular:

  The query result weight is determined as the product of the relevance value of the query result for the corresponding minimum subgraph and the subgraph weight of the minimum subgraph.

  The ranking unit 904 for ranking the query results obtained by the query result weights includes in particular:

  The query results are directly ranked according to the query result weights. Or

として決定される。式中ｓはクエリ結果の重みを表し、ｄは現在のクエリ結果を表し、Ｄはすでにランク付けされている１組のクエリ結果を表し、類似性（ｄ，ｄ´）はｄとｄ´との間の類似性を表し、第１位にランク付けされたクエリ結果を除くクエリ結果が類似性の重みによって繰り返してランク付けされる。 The query result with the highest weight is determined as the query result ranked first, the similarity between each two query results is determined, and the similarity weight of each ranked query result is ,

As determined. Where s represents the weight of the query result, d represents the current query result, D represents a set of already ranked query results, and the similarity (d, d ′) is d and d ′. The query results excluding the query result ranked first are repeatedly ranked according to the similarity weight.

  Embodiments of the present invention provide a method and apparatus for diversifying query results, wherein a set of related keywords for a set of keywords for a given query is determined in a region ontology, and these related keywords An expanded query set is constructed by the combination, thereby avoiding determining an expanded query set from untrusted query logs, thus making diversified query results more accurate.

  Those skilled in the art will appreciate that the embodiments of the invention may be embodied as a method, system, or computer program product. Accordingly, the present invention may be embodied in the form of an entirely hardware embodiment, an all software embodiment, or an embodiment of a combination of software and hardware. Furthermore, the present invention is embodied in one or more computer-usable storage media (including but not limited to disk memory, CD-ROM, optical memory, etc.) that contain computer-usable program code. Can be embodied in the form of a computer program product.

  The present invention has been described in flowcharts / block diagrams of methods, devices (systems) and computer program products according to embodiments of the invention. It should be understood that each flow / block of the flowchart / block diagram, and combinations of flowcharts / block diagrams, may be embodied in computer program instructions. These computer program instructions can build a means for instructions executed on the processor of a computer or other programmable data processing device to perform the functions specified in the flowchart / block diagram blocks of the flowchart. To produce the machine, it can be loaded onto a general purpose computer, a dedicated computer, an embedded processor, or another programmable data processing device processor.

  These computer program instructions may also be stored in a computer readable memory that may instruct a computer or other programmable data processing device to operate in a particular manner. This allows the instructions stored in the computer readable memory to create a product that includes instruction means for performing the functions specified in the flowchart / block diagram blocks.

  These computer program instructions are executed in a sequence of operational steps on a computer or other programmable data processing device, and the instructions executed on the computer or other programmable device are represented in the flowchart / block diagram of the flowchart. It can also be loaded onto a computer or other programmable data processing device to create a computer-implemented process that provides steps for performing the functions specified in the block.

  Although preferred embodiments of the present invention have been described, those skilled in the art who benefit from the underlying inventive concepts can make additional changes and modifications to these embodiments. Accordingly, the appended claims are to be construed to include the preferred embodiments and all modifications and variations that fall within the scope of the invention.

  It will be apparent to those skilled in the art that various modifications and variations can be made to the present invention without departing from the spirit and scope of the invention. Accordingly, it is intended that the present invention include those modifications and variations as long as the modifications and variations are included in the scope of the claims appended hereto and their equivalents.

  Further, a part or all of the above-described embodiment can be described as in the following supplementary notes, but is not limited thereto.

(Appendix 1)
A method for diversifying query results,
Building a set of related keywords by extracting the related keywords from the domain ontology for each query keyword of a given query;
Searching for each related keyword combination in the related keyword combination set to obtain a query result set;
Constructing a final query result set by extracting a corresponding number of query results from each query result set;
Ranking the constructed final query result set to obtain a diversified query result; and a method for diversifying the query results.

(Appendix 2)
The step of constructing a set of related keywords by extracting related keywords from the domain ontology for each query keyword of a given query comprises:
For each query keyword of a given query, extracting a keyword related keyword from the region ontology;
The method for diversifying query results according to claim 1, further comprising the step of constructing a set of related keyword combinations based on the related keywords.

(Appendix 3)
The step of constructing a combination set of related keywords based on the related keywords comprises:
Build a set of related keywords as S (Q) = {(c ₁ , c ₂ ,..., C _m ) | c ₁ ∈ C ₁ && c ₂ ∈ C ₂ &&... C _m ∈ C _m } where C The method for diversifying the query result according to appendix 2, wherein _i is a related keyword set of an i-th keyword among m keywords in a given query.

(Appendix 4)
After building a set of related keywords by extracting related keywords from the domain ontology for each query keyword,
For each related keyword combination in the related keyword combination set, the step of extracting from the region ontology a minimum subgraph connecting all the keywords in the combination,
Searching for each related keyword combination in the related keyword combination set to obtain a query result set comprises:
For each minimum subgraph, determining a subquery consisting of keywords and other nodes in the minimum subgraph;
Searching on keywords and other nodes in each subquery to obtain the same number of subquery result sets as the number of minimum subgraphs;
Determining a query result set as a set of respective subquery result sets;
The minimum subgraph is a subgraph that has a minimum number of edges in a subgraph connecting each keyword in a region ontology, and connects all related keywords in the closest way. A method for diversifying query results.

(Appendix 5)
Said step of constructing a final query result set by extracting a corresponding number of query results from each query result set;
Extracting a corresponding number of query results from each subquery result set according to the relevance of each subquery to a given query;
The method for diversifying the query results according to claim 4, further comprising the step of combining query results obtained from each subquery result set.

ｍはクエリキーワードの数を表し、ｒｉはクエリキーワードに対するクエリキーワードについて領域オントロジから抽出された関連キーワードのマッチ値を表し、｜Ｅ｜はサブグラフにおけるエッジの数を表す
として決定するステップと、
最小サブグラフのサブグラフ重みに従って、各最小サブグラフに対応するサブクエリ結果セットから対応する数のクエリ結果を抽出するステップとを含む
ことを特徴とする付記５に記載のクエリ結果を多様化するための方法。 (Appendix 6)
Said step of extracting a corresponding number of query results from each subquery result set according to the relevance of each subquery to a given query;
Subgraph weight for each minimum subgraph

m represents the number of query keywords, ri represents the match value of the related keyword extracted from the region ontology for the query keyword for the query keyword, and | E | is determined to represent the number of edges in the subgraph;
The method for diversifying query results according to appendix 5, comprising: extracting a corresponding number of query results from a subquery result set corresponding to each minimum subgraph according to the subgraph weight of the minimum subgraph.

(Appendix 7)
Extracting the corresponding number of query results from the subquery result set corresponding to each minimum subgraph according to the subgraph weight of the minimum subgraph,
The query results extracted from the subquery result set corresponding to the smallest subgraph are the top k * a query results most relevant to the smallest subgraph, where a is the current relative to the sum of the subgraph weights of all the smallest subgraphs. Represents the ratio of subgraph weights of the smallest subgraph, k is a predetermined value representing the scale of the final query result set, and k * a represents the largest integer less than the product of k and a. A method for diversifying the query result according to appendix 5.

(Appendix 8)
The step of ranking the constructed final query result set to obtain diversified query results comprises:
For each query result in the final query result set, determining a relevance value of the query result for the corresponding minimum subgraph;
Determining, for each query result in the final query result set, a query result weight according to the relevance value of the query result for the corresponding minimum subgraph and the subgraph weight of the minimum subgraph;
Ranking the query results in the final query result set according to the weights of the query results, and obtaining the diversified query results, to diversify the query results as described in appendix 4 the method of.

(Appendix 9)
For each query result in the final query result set, said step of determining the query result weight according to the relevance value of the query result for the corresponding minimum subgraph and the subgraph weight of the minimum subgraph,
9. The method for diversifying a query result according to appendix 8, wherein the weight of the query result is determined as a product of a relevance value of the query result for the corresponding minimum subgraph and a subgraph weight of the minimum subgraph.

ここで、ｓはクエリ結果の重みを表し、ｄは現在のクエリ結果を表し、Ｄはすでにランク付けされている１組のクエリ結果を表し、類似性（ｄ，ｄ´）はｄとｄ´との間の類似性を表す
として決定し、第１位にランク付けされたクエリ結果を除くクエリ結果を類似性の重みによって繰り返してランク付けする
ことを特徴とする付記８に記載のクエリ結果を多様化するための方法。 (Appendix 10)
The step of ranking the constructed final query result set to obtain diversified query results comprises:
Either rank the query results directly according to the weight of the query results, or determine the query result with the highest weight as the query result ranked first, and determine the similarity between each two query results Determine and rank the similarity weight of each query result,

Where s represents the weight of the query result, d represents the current query result, D represents a set of already ranked query results, and similarity (d, d ′) is d and d ′. The query result according to appendix 8, characterized in that the query result excluding the query result ranked first is ranked repeatedly by the similarity weight. A way to diversify.

(Appendix 11)
A device for diversifying query results,
A related keyword combination set building unit configured to build a set of related keyword combinations by extracting related keywords from the domain ontology for each query keyword;
A query unit configured to search for each related keyword combination in the related keyword combination set to obtain a query result set;
A final query result set construction unit configured to construct a final query result set by extracting a corresponding number of query results from each query result set;
An apparatus for diversifying query results, comprising: a ranking unit configured to rank the constructed final query result set to obtain diversified query results.

(Appendix 12)
The related keyword combination set construction unit is
For each query keyword of a given query, extract keyword related keywords from the region ontology,
The apparatus for diversifying query results according to appendix 11, wherein the apparatus is configured to construct a combination set of related keywords based on the related keywords.

(Appendix 13)
The related keyword combination set construction unit that constructs a combination set of related keywords based on the related keywords,
A related keyword combination set is constructed such that S (Q) = {(c ₁ , c ₂ ,..., C _m ) | c ₁ ∈C ₁ && c ₂ ∈C ₂ &&... C _m ∈C _m } And
Here C _i, the apparatus for diversifying the query results according to Note 12, which is a related keyword set of the i-th keyword in the m keywords in a given query.

(Appendix 14)
The related keyword combination set construction unit is
After building a set of related keywords by extracting related keywords from the domain ontology for each query keyword,
For each related keyword combination in the related keyword combination set, it is configured to extract from the region ontology the smallest subgraph connecting all the keywords in the combination,
The query unit is
For each minimal subgraph, determine a subquery consisting of keywords and other nodes in the minimal subgraph,
Search by keyword and other nodes in each subquery to get as many subquery result sets as the number of minimum subgraphs,
Configured to determine the query result set as a set of each subquery result set,
The supplementary note 11, wherein the minimum subgraph is a subgraph having a minimum number of edges in a subgraph connecting each keyword in a region ontology, and connecting all related keywords in a closest manner. A device for diversifying query results.

(Appendix 15)
The final query result set construction unit is
Extract a corresponding number of query results from each subquery result set according to the relevance of each subquery to a given query,
The apparatus for diversifying the query results according to appendix 14, wherein the query results obtained from the respective subquery result sets are combined.

ｍはクエリキーワードの数を表し、ｒｉはクエリキーワードに対するクエリキーワードについて領域オントロジから抽出された関連キーワードのマッチ値を表し、｜Ｅ｜はサブグラフにおけるエッジの数を表す
として決定し、
最小サブグラフのサブグラフ重みに従って、各最小サブグラフに対応するサブクエリ結果セットから対応する数のクエリ結果を抽出し、
各サブクエリ結果セットから取得したクエリ結果を結合するように構成される
ことを特徴とする付記１５に記載のクエリ結果を多様化するための装置。 (Appendix 16)
The final query result set construction unit is
Subgraph weight for each minimum subgraph

m represents the number of query keywords, ri represents the match value of the related keyword extracted from the region ontology for the query keyword for the query keyword, and | E | represents the number of edges in the subgraph,
Extract a corresponding number of query results from the subquery result set corresponding to each minimum subgraph according to the subgraph weight of the minimum subgraph,
The apparatus for diversifying query results according to appendix 15, characterized in that it is configured to combine query results obtained from each subquery result set.

(Appendix 17)
In the final query result set construction unit for extracting a corresponding number of query results from the subquery result set corresponding to each minimum subgraph according to the subgraph weight of the minimum subgraph,
The query results extracted from the subquery result set corresponding to the smallest subgraph are the top k * a query results most relevant to the smallest subgraph, where a is the current relative to the sum of the subgraph weights of all the smallest subgraphs. Represents the ratio of subgraph weights of the smallest subgraph, k is a predetermined value representing the scale of the final query result set, and k * a represents the largest integer less than the product of k and a. An apparatus for diversifying the query results according to attachment 16.

(Appendix 18)
The ranking unit is
For each query result in the final query result set, determine the relevance value of the query result for the corresponding smallest subgraph,
For each query result in the final query result set, determine the query result weight according to the relevance value of the query result for the corresponding minimum subgraph and the subgraph weight of the minimum subgraph,
The query result according to appendix 14, wherein the query result is configured to rank the query result in the final query result set according to the weight of the query result to obtain a diversified query result. Equipment for.

(Appendix 19)
For each query result in the final query result set, the ranking unit that determines the query result weight according to the relevance value of the query result to the corresponding minimum subgraph and the subgraph weight of the minimum subgraph,
The query result according to appendix 18, wherein the query result weight is determined to be determined as a product of the value of the relevance of the query result for the corresponding minimum subgraph and the subgraph weight of the minimum subgraph. Device to do.

ここで、ｓはクエリ結果の重みを表し、ｄは現在のクエリ結果を表し、Ｄはすでにランク付けされている１組のクエリ結果を表し、類似性（ｄ，ｄ´）はｄとｄ´との間の類似性を表す
として決定し、第１位にランク付けされたクエリ結果を除くクエリ結果を類似性の重みによって繰り返してランク付けするように構成される
ことを特徴とする付記１８に記載のクエリ結果を多様化するための装置。 (Appendix 20)
The ranking unit that ranks query results in the final query result set by query result weights,
Either rank the query results directly according to the weight of the query results, or determine the query result with the highest weight as the query result ranked first, and determine the similarity between each two query results Determine and rank the similarity weight of each query result,

Where s represents the weight of the query result, d represents the current query result, D represents a set of already ranked query results, and similarity (d, d ′) is d and d ′. The supplementary note 18 is characterized in that it is configured so that the query results excluding the query result ranked first are repeatedly ranked according to the similarity weight. Device for diversifying the described query results.

201: Query unit 202: Query log memory unit 203: Query disambiguation unit 204: Subquery memory unit 205: Document memory unit 206: Keyword search unit 207: Subquery result memory unit 208: Query result combination unit 209: Query result memory unit 210 : Query result ranking unit 211: Diversified rank list memory unit 901: Related keyword combination set construction unit 902: Query unit 903: Final query result set construction unit 904: Ranking unit

Claims (10)

A method for diversifying query results,
Building a set of related keywords by extracting the related keywords from the domain ontology for each query keyword of a given query;
Searching for each related keyword combination in the related keyword combination set to obtain a query result set;
Constructing a final query result set by extracting a corresponding number of query results from each query result set;
Ranking the constructed final query result set to obtain a diversified query result; and a method for diversifying the query results. The step of constructing a set of related keywords by extracting related keywords from the domain ontology for each query keyword of a given query comprises:
For each query keyword of a given query, extracting a keyword related keyword from the region ontology;
The method for diversifying query results according to claim 1, comprising building a set of related keyword combinations based on the related keywords. The step of constructing a combination set of related keywords based on the related keywords comprises:
Build a set of related keywords as S (Q) = {(c ₁ , c ₂ ,..., C _m ) | c ₁ ∈ C ₁ && c ₂ ∈ C ₂ &&... C _m ∈ C _m } where C The method for diversifying query results according to claim 2, wherein _i is a related keyword set of an i-th keyword among m keywords in a given query. After building a set of related keywords by extracting related keywords from the domain ontology for each query keyword,
For each related keyword combination in the related keyword combination set, the step of extracting from the region ontology a minimum subgraph connecting all the keywords in the combination,
Searching for each related keyword combination in the related keyword combination set to obtain a query result set comprises:
For each minimum subgraph, determining a subquery consisting of keywords and other nodes in the minimum subgraph;
Searching on keywords and other nodes in each subquery to obtain the same number of subquery result sets as the number of minimum subgraphs;
Determining a query result set as a set of respective subquery result sets;
The minimum subgraph is a subgraph that has a minimum number of edges in a subgraph connecting each keyword in a region ontology and connects all related keywords in the closest way. To diversify query results for. Said step of constructing a final query result set by extracting a corresponding number of query results from each query result set;
Extracting a corresponding number of query results from each subquery result set according to the relevance of each subquery to a given query;
Combining the query results obtained from each subquery result set. The method for diversifying query results according to claim 4. Said step of extracting a corresponding number of query results from each subquery result set according to the relevance of each subquery to a given query;
Subgraph weight for each minimum subgraph

m represents the number of query keywords, ri represents the match value of the related keyword extracted from the region ontology for the query keyword for the query keyword, and | E | is determined to represent the number of edges in the subgraph;
Extracting the corresponding number of query results from the subquery result set corresponding to each minimum subgraph according to the subgraph weights of the minimum subgraphs. 6. A method for diversifying query results according to claim 5 comprising: . Extracting the corresponding number of query results from the subquery result set corresponding to each minimum subgraph according to the subgraph weight of the minimum subgraph,
The query results extracted from the subquery result set corresponding to the smallest subgraph are the top k * a query results most relevant to the smallest subgraph, where a is the current relative to the sum of the subgraph weights of all the smallest subgraphs. Represents the ratio of subgraph weights of the smallest subgraph, k is a predetermined value representing the scale of the final query result set, and k * a represents the largest integer less than the product of k and a. The method for diversifying the query results according to claim 5. The step of ranking the constructed final query result set to obtain diversified query results comprises:
For each query result in the final query result set, determining a relevance value of the query result for the corresponding minimum subgraph;
Determining, for each query result in the final query result set, a query result weight according to the relevance value of the query result for the corresponding minimum subgraph and the subgraph weight of the minimum subgraph;
5. Diversifying the query results according to claim 4, comprising ranking query results in the final query result set according to the weights of the query results to obtain diversified query results. Way for. For each query result in the final query result set, said step of determining the query result weight according to the relevance value of the query result for the corresponding minimum subgraph and the subgraph weight of the minimum subgraph,
9. The method for diversifying a query result according to claim 8, wherein the weight of the query result is determined as a product of the relevance value of the query result for the corresponding minimum subgraph and the subgraph weight of the minimum subgraph. . A device for diversifying query results,
A related keyword combination set building unit configured to build a set of related keyword combinations by extracting related keywords from the domain ontology for each query keyword;
A query unit configured to search for each related keyword combination in the related keyword combination set to obtain a query result set;
A final query result set construction unit configured to construct a final query result set by extracting a corresponding number of query results from each query result set;
An apparatus for diversifying query results, comprising: a ranking unit configured to rank the constructed final query result set to obtain diversified query results.

Images