CN101169780A - A Semantic Ontology-Based Retrieval System and Method - Google Patents

Abstract

The embodiment of the invention discloses a semantic ontology-based retrieval system, which includes a semantic ontology index database and a semantic ontology index processing unit. The semantic ontology search processing unit obtains the text hit file list, and matches the text hit file list with the semantic ontology index in the semantic ontology index database to obtain the document semantic classification table. This enables the retrieval system to identify the semantic information of the files to be retrieved, and makes the search results present semantic classification results. The embodiment of the present invention also discloses a semantic ontology-based retrieval method. The method first establishes a semantic ontology index for a file with an established text index, and performs semantic ontology index matching processing on the text matching result when the user performs a search, so that The final output results present a semantic classification on the traditional text matching results, which is convenient for users to query.

Description

A Semantic Ontology-Based Retrieval System and Method

technical field

The invention relates to information retrieval technology, in particular to a semantic ontology-based retrieval system and method.

Background technique

With the rapid development of retrieval technology, text-based information retrieval technology has gradually matured, forming a complete set of ideas and perfect algorithms, and has been widely used in various search engines, such as Google (Google), AltaVista, etc. , Lycos, Yahoo (Yahoo), etc.

Fig. 1 is a structural block diagram of an existing text search engine. As shown in Figure 1, existing text search engine comprises: spider control module 101, uniform resource location (URL) database 102, web spider 103, URL extraction module 104, webpage database 105, link information extraction module 106, text index module 107 , link database 108 , index database 109 , web page rating module 110 and query server 111 .

The web spider 103 crawls web pages from the Internet, and sends the web pages into the web page database 105 . The URL extracting module 104 extracts URLs from the web pages crawled by the web spider 103 and sends the URLs to the URL database 102 . The spider control module 101 obtains the URL of the webpage from the URL database 102, and controls the web spider 103 to crawl other webpages, and repeats the above steps until all the webpages are crawled.

The system acquires text information from the webpage database 105 and sends it to the text index module 107 , which builds an index and then sends it to the index database 109 . At the same time, the link information extraction module 106 acquires link information from the webpage database 105 and sends it to the link database 108 . The link information in the link database 108 provides the webpage rating module 110 with a basis for webpage rating.

When a user submits a query request through the query server 111, the query server 111 searches the index database 109 for webpages related to the user query request, and the web page rating module 110 combines the user query request with the link information in the link database 108 to evaluate the search results. Evaluation of the degree of relevance is performed, and the search results are sorted according to their degree of relevance through the query server 111, and the final page is organized to be returned to the user.

Although the existing text retrieval technology can search for files containing the user's text query information, it cannot identify the content and meaning of the searched files. This is because the existing text retrieval technology is based on text string matching. The problem with this retrieval technology is that when different words can represent the same meaning or a word has different meanings in different contexts, it will It will limit the precision rate and recall rate of the search, resulting in the search results far from meeting the user's needs. For example, when the user's search keyword is "heaven", it is impossible to judge whether the files that meet the user's search criteria reflect " Paradise Game" or "Paradise Music" content. The proposal of Semantic Web provides an opportunity to solve these problems.

The Semantic Web is a network composed of a group of webpages that can be automatically controlled and identified by computers. It expands the data that computers can recognize for webpages on the basis of the existing Internet, and adds documents for computer use, that is, ontology On the language to label web pages and clarify their semantics, so that web page information can not only be understood by humans, but also automatically controlled and recognized by computers. Semantically annotated webpages generally use Extensible Markup Language (XML) or Hypertext Markup Language (Html) as data for annotation, Resource Description Framework (RDF) as data description model, combined with semantic ontology, so that the annotated data has clear semantics. Ontology is a concept derived from philosophy. Its original meaning refers to the theory of existence, its essence and laws. It was later introduced by the field of artificial intelligence, specifically referring to an explicit specification of conceptualization. Ontology can explicitly and formally express various concepts and interrelationships in the domain, thereby explicitly expressing the semantics of terms, so it plays an important role in semantic query. The semantic ontology referred to here defines the basic terms that make up the concept of the subject domain and the relationship between them, and stipulates the extension rules for combining the basic terms and the relationship between them to define the vocabulary.

The purpose of semantic retrieval is to enhance and improve traditional search results through data obtained from the Semantic Web. Fig. 2 is a structural block diagram of an existing semantic search system. As shown in FIG. 2 , the existing semantic search system includes: query interface 201 , query preprocessing module 202 , semantic ontology reasoning engine 203 , annotation ontology library 204 , traditional search module 205 and result return interface 206 .

The query interface 201 acquires user query information and sends it to the query preprocessing module 202 .

The query preprocessing module 202 analyzes the user's query information, segments it into query keywords through word segmentation technology, and sends them to the semantic ontology reasoning engine 203 .

The semantic ontology inference engine 203 matches and infers the ontology concept vocabulary corresponding to the query keyword according to the ontology concept vocabulary and the relationship between concepts defined in the annotation ontology library 204 , and returns it to the query preprocessing module 202 .

The query preprocessing module 202 sends the ontology concept vocabulary returned by the semantic ontology reasoning engine 203 to the traditional search module 205, and indicates to search according to the semantics. Searching according to semantics here refers to performing string matching according to the semantic concepts marked on the webpage when the webpage has been semantically annotated, rather than directly performing string matching on the content of the webpage itself.

The traditional search module 205 performs semantic search and sends the search results to the result return interface 206 . The result returning interface 206 returns the search result to the user.

It can be seen that the above-mentioned semantic search system matches the user's query keywords with the semantic concept vocabulary of the marked webpage.

To sum up, although the existing text retrieval technology can search for files containing query keywords, it cannot identify the semantic information of the searched files; and the existing semantic retrieval technology no longer performs keyword retrieval, resulting in The obtained files contain too many results that do not match the user's query information, and the matching efficiency based on user query keywords and semantic concept vocabulary is not satisfactory. Therefore, the search accuracy of the existing retrieval technology is not high.

Contents of the invention

In view of this, the main purpose of the embodiments of the present invention is to provide a semantic ontology-based retrieval system to improve search accuracy.

Another object of the embodiments of the present invention is to provide a semantic ontology-based retrieval method to improve search accuracy.

In order to achieve the above object, technical solution of the present invention is achieved in that way:

The embodiment of the present invention discloses a semantic ontology-based retrieval system, which includes:

Semantic ontology index database, used to save semantic ontology index;

The semantic ontology search processing unit is configured to obtain a text hit file list, and match the text hit file list with the semantic ontology index in the semantic ontology index database to obtain a document semantic classification table.

The embodiment of the present invention also discloses a semantic ontology-based retrieval method, which includes the following steps:

A. Obtain the files with established text indexes, and establish semantic ontology indexes for the obtained files;

B. Obtain a list of text hit files, perform semantic ontology index matching processing on the list of text hit files, and obtain a document semantic classification table.

Therefore, the semantic ontology-based retrieval system and method provided by the embodiments of the present invention have the following advantages: Firstly, a semantic ontology index is established for a file with an established text index, and when the user searches, the text index matching is performed on the text query information input by the user The text hit file list is obtained through processing, and then the semantic ontology index matching process is performed on the text hit file list to obtain the document semantic classification table, so that the text retrieval results have semantic classification information, and the search accuracy is improved.

Description of drawings

Fig. 1 is the structural block diagram of existing text search engine;

Fig. 2 is a structural block diagram of an existing semantic search system;

Fig. 3 is a structural block diagram of a semantic ontology-based retrieval system according to an embodiment of the present invention;

Fig. 4 is a flow chart of establishing a semantic ontology index by a semantic ontology index processing unit in an embodiment of the present invention;

Fig. 5 is a flow chart of the retrieval system of the embodiment of the present invention shown in Fig. 3 performing a search process for a user;

FIG. 6 is a schematic diagram of two resource descriptions defined by the embodiment of the present invention;

Fig. 7 is a schematic diagram of the results deduced from Fig. 6;

Fig. 8 is a relation diagram established for the semantic ontology vocabulary in the embodiment by the annotation ontology library in the embodiment of the present invention;

FIG. 9 is a relationship diagram of the semantic ontology vocabulary in FIG. 8 after reasoning.

Detailed ways

In order to make the purpose, technical solution and advantages of the present invention clearer, the present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments.

Fig. 3 is a structural block diagram of a semantic ontology-based retrieval system according to an embodiment of the present invention. As shown in Figure 3, the system includes: a search interface module 301, a document semantic classification rule engine 302, a search processing module 303, a semantic ontology reasoning engine 304, an annotation ontology library 305, an index database 306, an index processing module 307, and a file database 308 And network file capture module 309. Wherein, the search processing module 303 includes: a text search processing unit 310, a semantic ontology search processing unit 311 and a sorting processing unit 312; the index database 306 includes: a text index 315 and a semantic ontology index 316; the index processing module includes: a text index processing unit 313 and a semantic ontology index processing unit 314.

The network file grabbing module 309 is mainly responsible for grabbing webpages from the Internet, and saving the grabbed webpages into the file database 308 . The network file grabbing module 309 generally traverses the webpage space through a webpage grabbing program, such as "network robot" or "web spider", scans websites within a certain range of Internet Protocol (IP) addresses, and follows links on the network. Capture web files from one web page to another, and from one website to another.

The file database 308 is used to store files for user retrieval, including audio files, video files and text files. These files can be network files or non-network files. Each document in document database 308 has a unique document identification (DocID).

The index processing module 307 is mainly responsible for analyzing the files stored in the file database 308, extracting keywords of file content, eliminating duplicate files, etc., and establishing different types of index information for the files in the file database 308. The index processing module 307 includes a text index processing unit 313 and a semantic ontology index processing unit 314 .

The text index processing unit 313 is a traditional processing unit for establishing a text index, and establishes a text index by analyzing file content, extracting keywords and identification information of the file. Since the traditional text index building process is a mature prior art, it will not be repeated here.

The semantic ontology index processing unit 314 is responsible for establishing a semantic ontology index for the documents whose text index has been established. First, analyze the document that has established a text index to determine whether it contains semantic annotation information. If a file contains semantic annotation information, extract the relevant semantic annotation information and file identification information, and establish the semantic ontology index of the file.

The index database 306 is used to store the index information created by the index processing module 307 , that is, to store the text index 315 created by the text index processing unit 313 and the semantic ontology index 316 created by the semantic ontology index processing unit 314 .

The search processing module 303 is responsible for processing the user's query request, by matching the user's text query information with the index information of the file, and feeding back the files that meet the user's query condition to the user in a certain order. The search processing module 303 includes a text search processing unit 310 , a semantic ontology search processing unit 311 and a ranking processing unit 312 .

The text search processing unit 310 is responsible for matching the text query information input by the user with the text index 315 , and find out the text matching file identification information that meets the user query condition.

The semantic ontology search processing unit 311 is responsible for matching the text hit document identification information obtained by the text search processing unit 310 with the semantic ontology index 316, performing semantic classification on these text hit document identification information, and obtaining a document semantic classification table.

The annotation ontology library 305 and the semantic ontology reasoning engine 304 are responsible for performing semantic reasoning on the ontology concept vocabulary set in the document semantic classification table generated by the semantic ontology search processing unit 311 to obtain an extended semantic ontology vocabulary set. The annotation ontology library 305 stores the defined vocabulary of semantic ontology concepts and the relationship between semantic ontology concepts, and the semantic ontology reasoning engine 304 defines reasoning rules and performs reasoning operations.

The document semantic classification rule engine 302 triggers the semantic classification rules defined by itself according to the situation deduced by the semantic ontology reasoning engine 304, and extends and integrates the document semantic classification table.

The sorting processing unit 312 is responsible for sorting and optimizing the final results, that is, calculating the relevance and importance of the semantic document classification table obtained through a series of processing, such as text index matching, semantic ontology index matching and semantic reasoning extension, etc. And feed back the sorting of the searched files to the search interface module 301 according to the calculation result.

The search interface module 301 is responsible for the interactive operation between the system and the user, and forwards the text query information input by the user to the search processing module 303; and feeds back the sorting results of the sorting processing unit 312 to the user.

The text index 315 stored in the index database 306 includes a text forward index and a text inverted index. Table 1 is the text forward index table, and Table 2 is the text inverted index table, as shown in Table 1 and Table 2:

Table 1

Document ID (DocID) Key words 1 Paradise, music,... 2 application,... 3 application,... 4 Paradise, games,... ... ...

Table 2

Key words Document Identification Sequence (DocID) Heaven 1, 4,... application 2, 3,... ... ...

As can be seen from the above two tables, the text forward index uses the file identifier as the key value to establish the mapping relationship between the file identifier and the keyword; while the text inverted index uses the keyword as the key value to establish the keyword and file The mapping relationship between identifiers.

Similarly, the semantic ontology index 315 stored in the index database 306 includes a semantic ontology forward index and a semantic ontology inverted index. Table 3 is the semantic ontology forward index table, and Table 4 is the semantic ontology inverted index table, as shown in Table 3 and Table 4:

table 3

Document ID (DocID) Semantic ID 1 Pop music 2 classical music

3 novel 4 Computer Games 5 Pop music ... ...

Table 4

Semantic ID Document Identification Sequence (DocID) Pop music 1, 5, ... classical music 2,... novel 3.... Computer Games 4.... ... ...

Semantic Ontology Forward Index uses document identifiers as key values to establish the mapping relationship between document identifiers and semantic identifiers; while Semantic Ontology Inverted Index uses semantic identifiers as key values to establish the mapping relationship between semantic identifiers and document identifiers.

FIG. 4 is a flow chart of establishing a semantic ontology index 316 by the semantic ontology index processing unit 314 in the embodiment of the present invention. The establishment process of the semantic ontology index is carried out on the basis of the text index established by the text index processing unit, and the execution trigger condition is that the text index processing unit 313 has established a text index for a certain file. Referring to Figure 4, the establishment process of the semantic ontology index includes the following steps:

In step 401 , the semantic ontology index processing unit 314 first reads the document that has been processed by the text index processing unit 313 and has established a text index.

In step 402, the semantic ontology index processing unit 314 judges whether the read file is marked with a semantic mark. If the file is marked with semantic tags, go to step 403; otherwise, end the process of establishing a semantic ontology index for the file.

The difference between semantically annotated documents and non-semantically annotated documents is that semantically annotated documents establish ontology-concept mapping information. For example, if a file is identified as 9, and the content of the web page with the URL http://grids.ucs.indiana.edu/ptliupages/publications/index.html mainly describes the matters that need to be paid attention to when doing research, then the web page can be Labeled as "Research (Research)" concept. Some of the existing semantic annotation information is in the form of annotations, and some are embedded in web pages in the form of XML packages. In this example, a semantic annotation information expressed in annotation form is given, which is annotated with Stanford University's text annotation tool OntoMat:

<html>

<head>

<! --<rdf:RDF xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#"

xmlns:daml="http://www.daml.org/2001/03/daml+oil#"

xmlns="http://annotation.semanticweb.org/iswc/iswc.daml#"

<Research rdf:about=″http://grids.ucs.indiana.edu/ptliupages/publications/index.html″

</rdf:RDF>

-->

<title>Community Grids Publications</title>

This example indicates that the content of the webpage http://grids.ucs.indiana.edu/ptliupages/publications/index.html is mainly about "Research". For webpages marked with OntoMat tools, the semantic annotation information is placed in the annotation information in the Html head, starting with <rdf:RDF and ending with </rdf:RDF>. Therefore, when the semantic ontology index processing unit 314 detects that the semantic annotation information starts with <rdf:RDF and ends with </rdf:RDF>, it determines that the webpage file has been annotated by semantic tags.

In step 403, the semantic ontology index processing unit 314 reads the semantic annotation information of the document.

In this embodiment, the semantic ontology index processing unit 314 reads the semantic annotation information of the webpage whose file ID is 9, that is, reads the comment information in the Html header. Table 5 is the format table for extracting semantic annotation information, as shown in Table 5:

table 5

Document ID (DocID) Semantic annotation information 9 <rdf:RDF xmlns:rdf="http://www.w3.org/1999/02/22-rdf-svntax-ns#"xmlns:daml=http://www.daml.org/2001/03/daml+oil#″xmlns="http://annotation.semanticweb.org/iswc/iswc.daml#"<Researchrdf:about="http://grids.ucs.indiana.edu/ptliupages/publications/index.html"</rdf:RDF> ... ...

In step 404, the semantic ontology index processing unit 314 extracts semantic ontology concept vocabulary from the read semantic annotation information, and establishes a semantic ontology index.

In this embodiment, the semantic ontology index processing unit 314 invokes the relevant RDF document processing application programming interface (API), extracts the semantic ontology concept vocabulary "Research" from the semantic annotation information, establishes the semantic ontology forward index of the webpage 9, and simultaneously Converted into semantic ontology inverted index, as shown in Table 6 and Table 7. Table 6 is the semantic ontology forward index of webpage 9, and table 7 is the semantic ontology inverted index of webpage 9, as shown in Table 6 and Table 7:

Table 6

Document ID (DocID) Semantic ID 9 Research

Table 7

Semantic ID Document ID (DocID) Research 9

Step 405 , the semantic ontology index processing unit 314 saves the established semantic ontology forward index and semantic ontology inverted index into the index database 306 , that is, the content of the semantic ontology index 316 is formed.

Before establishing the semantic ontology index, the reason why the processing steps of the text index processing unit 313 must first be performed is that when the user searches, the files matching the text query information input by the user must first be queried, and then the semantic ontology index matching is performed on these files deal with. The processing steps of the text index processing unit 313 ensure that each file with a text index and semantic information has corresponding semantic ontology index information in the semantic ontology index 316, thereby avoiding the problem of directly reading files from the file database 308. The case where the document generated by semantic ontology index matching has semantic ontology index but no text index.

Fig. 5 is a flow chart of the retrieval system of the embodiment of the present invention shown in Fig. 3 performing a search process for a user, as shown in Fig. 5, comprising the following steps:

Step 501 , the search interface module 301 obtains the text query information input by the user, and sends it to the search processing module 303 . In this embodiment, it is assumed that the query information input by the user is "paradise".

Step 502 , the search processing module 303 receives the text query information sent by the search interface module 301 , performs segmentation preprocessing on it, and then sends the segmented query keywords to the text search processing unit 310 .

The specific process of segmentation processing has been described in existing related documents describing search engines, and will not be repeated here. In this embodiment, the text query information "paradise" is segmented and preprocessed and the result is the keyword "paradise".

Step 503 , the text search processing unit 310 matches the segmented query keywords with the text inverted index, and sends the matched text hit file list to the semantic ontology search processing unit 311 .

After receiving the query keyword, the text search processing unit 310 sends request information for reading the text inverted index to the index database 306, and the index database 306 returns the text inverted index in the text index 315 according to the request. The text search processing unit 310 matches the user query keyword "paradise" with the text inverted index, obtains a series of web page file identifiers containing the keyword—the text hit file identifier list, and sends the text hit file list to the semantic ontology The search processing unit 311 performs processing.

For simplicity, it is assumed in this embodiment that only 20 files are indexed. Table 8 is the text inverted index table returned to the text search processing unit 310 by the index database 306, as shown in Table 8:

Table 8

Key words file ID sequence ... ... application 01011100100011011010 Heaven 11011011111110001011 ... ...

In Table 8, each row corresponds to a keyword and the file identification sequence in which the keyword appears. Among them, the total number of binary digits of the file identification sequence is 20, indicating the total number of files indexed, each binary digit represents a file, and the position number of the binary digit is the same as the file identification serial number, that is, the first binary digit indicates that the identification serial number is 1 , the second binary bit indicates the file with the identification number 2, and so on. If a certain binary bit is 0, it means that the corresponding keyword does not appear in the corresponding file, and if it is 1, it means that the corresponding keyword appears in the corresponding file.

The text search processing unit 310 matches the user query keyword "paradise" to the "paradise" keyword in Table 8, takes out the following file identification sequence, that is, the text hit file list 11011011111110001011, and sends it to the semantic ontology search processing unit 311. The file whose binary bit is 1 in the text hit file list is the hit file.

Similarly, if the text query information entered by the user is "paradise application", the keyword "paradise" and the keyword "application" will be obtained after segmentation and preprocessing, so as long as they match "paradise" and " Apply the two keywords, and perform an AND operation on the subsequent file identification sequence to get the result 01011000100010001010, where the binary bit is 1, which means that the two keywords "paradise" and "application" appear in the corresponding file at the same time.

In step 504, after the semantic ontology search processing unit 311 obtains the list of text hit files, it first judges whether to perform semantic ontology inverted index matching processing.

The basis for judging by the semantic ontology search processing unit 311 is the number of text hit files, if the number of hit files is greater than a certain threshold, the semantic ontology inverted index matching process is performed, and step 505 is performed; otherwise, the semantic ontology forward For index matching processing, step 506 is executed. The threshold value can be stored in the semantic ontology search processing unit 311 as a predefined value, or it can be a value dynamically adjusted by the retrieval system according to statistical laws or other conditions.

After receiving the text hit file list 11011011111110001011, the semantic ontology search processing unit 311 accumulates and calculates to obtain 14 numbers of 1s in the binary sequence, that is, the number of text hit files is 14. Assuming that the threshold value is 10, since 14 is greater than 10, the semantic ontology inverted index matching process is performed. If the threshold value is 15, since 14 is less than 15, the semantic ontology forward index matching process is performed.

Step 505 , the semantic ontology search processing unit 311 performs semantic ontology inverted index matching processing on the files in the text hit file list to obtain a document semantic classification table.

First, the semantic ontology search processing unit 311 sends a request message for reading the semantic ontology inverted index to the index database 306 . The index database 306 returns the semantic ontology inverted index according to the request. The semantic ontology search processing unit 311 sequentially reads each record in the semantic ontology inverted index, performs an intersection operation on the file identification sequence in the record and the text hit file list, that is, performs a bitwise AND operation on the two binary sequences, and then uses the operation The result covers the corresponding file identification sequence in the semantic ontology inverted index table. Finally, the records whose intersection is empty are filtered out, and the original semantic ontology inverted index table becomes the document semantic classification table. Execute step 507.

Table 9 is the semantic ontology inverted index table returned to the semantic ontology search processing unit 311 by the index database 306 in this embodiment, as shown in table 9:

Table 9

Semantic ID file ID sequence Pop music 01011010110001100000 Computer Games 10100101000100001011 classical music 00010000001011000001 novel 100000000000000000110 athletic star 00000000000000010000

In Table 9, it is assumed that the 20 indexed documents only involve five semantic ontology concepts, that is, there are five kinds of semantic identifiers in all documents. The document identification sequence after each semantic identification indicates the occurrence of the ontology concept in 20 documents. Its representation method is the same as the file identification sequence in the text inverted index, each binary bit represents a file, and the position number of the binary bit is the same as the identification number of the file. If a binary bit is 0, it means that the corresponding ontology concept is not marked in the corresponding file, and if it is 1, it means that the corresponding ontology concept is marked. For example, the file identification sequence of pop music is 01011010110001100000, which means that files with file IDs of 2, 4, 5, 7, 9, 10, 14, and 15 are labeled as pop music, reflecting that the content of these files is related to pop music.

The semantic ontology search unit 311 reads each file identification sequence in the semantic ontology inverted index shown in Table 9, performs a bitwise AND operation with the text hit file list 11011011111110001011, and stores the operation result in the corresponding file identification sequence in Table 9 position, and overwrite the original file identification sequence, and finally filter out the semantic identification items whose intersection is empty, that is, the semantic identification items with all zeros in the operation result, and generate a document semantic classification table. Table 10 is the generated document semantic classification table, as shown in Table 10:

Table 10

Semantic ID file ID sequence Pop music 01011010110000000000 Computer Games 10000001000100001011 classical music 00010000001010000001 novel 100000000000000000010

In this way, the text hit file list 11011011111110001011 is semantically classified.

Step 506 , the semantic ontology search processing unit 311 performs semantic ontology forward index matching processing on the files in the text hit file list to obtain a document semantic classification table.

First, the semantic ontology search processing unit 311 sends a request message to the index database 306 to read the forward index of the semantic ontology. Table 11 is that index database 306 returns semantic ontology forward index table according to the request of semantic ontology search processing unit 311, as shown in table 11:

Table 11

File ID Semantic ID 1 computer games, novels 2 Pop music

3 Computer Games 4 pop music, classical music 5 Pop music 6 Computer Games 7 Pop music 8 Computer Games 9 Pop music 10 Pop music 11 classical music 12 Computer Games 13 classical music 14 pop music, classical music 15 Pop music 16 athletic star 17 Computer Games 18 novel 19 computer games, novels 20 computer games, classical music

The semantic ontology search processing unit 311 converts the text hit file list 11011011111110001011 into specific file identifiers: 1, 2, 4, 5, 7, 8, 9, 10, 11, 12, 13, 17, 19, 20, and each A file identifier matches the corresponding record in the semantic ontology forward index as a query condition, and a semantic ontology forward index containing only these file identifiers is obtained. Table 12 is the semantic ontology forward index table obtained through the above process, as shown in Table 12:

Table 12

File ID Semantic ID 1 computer games, novels 2 Pop music 4 pop music, classical music 5 Pop music

7 Pop music 8 Computer Games 9 Pop music 10 Pop music 11 classical music 12 Computer Games 13 classical music 17 Computer Games 19 computer games, novels 20 computer games, classical music

Finally, take each semantic ontology concept that appears in Table 12 as a key value, count the document identifiers that appear in this key value, complete the conversion from forward index to inverted index, and generate a document semantic classification table. Table 13 is the document semantic classification table obtained through the above process, as shown in Table 13:

Table 13

Semantic ID file ID sequence Pop music 01011010110000000000 Computer Games 10000001000100001011 classical music 00010000001010000001 novel 100000000000000000010

Then step 507 is executed.

The reason why it is divided into semantic ontology inverted index matching processing and semantic ontology forward index matching processing is that efficiency is considered. Because in the process of semantic ontology inverted index matching processing, it is necessary to use the text hit file list to match each record in the semantic ontology inverted index in turn, and perform an intersection operation. This process of full table scanning semantic ontology inverted index , which is computationally expensive. Therefore, when the number of text hit files is small, performing semantic ontology forward index matching processing can reduce the amount of calculation. But no matter which matching method is used, the resulting document semantic classification table is the same, that is, Table 13 is the same as Table 10.

Step 507, the semantic ontology search processing unit 311 utilizes the semantic ontology reasoning engine 304, the annotation ontology library 305 and the document semantic classification rule engine to reason the semantic vocabulary in the document semantic classification table, expand the semantic classification table according to the reasoning results, and The extended document semantic classification table is sent to the sorting processing unit 312 .

After the semantic ontology search processing unit 311 executes the semantic ontology index matching operation, it first sends the semantic ontology concept vocabulary in the document semantic classification table to the semantic ontology reasoning engine 304 for semantic reasoning. Semantic ontology inference engine 304 generates an RDF document representing the relationship between semantic ontology words according to semantic ontology concepts and their relationships defined in ontology annotation library 305 and its own inference rules, and returns them to semantic ontology search processing unit 311 . Then, the semantic ontology search processing unit 311 matches the RDF document with the trigger conditions in the semantic classification rules defined in the document semantic classification rule engine 302, judges which semantic classification rules need to be triggered, and triggers the corresponding rules to generate the reasoning-extended The semantic taxonomy of documents. Finally, the extended semantic document classification table is sent to the sorting processing unit 312 .

In this embodiment, the semantic ontology search processing unit 311 sends the four semantic ontology concept words in Table 10 or Table 13, pop music, computer game, classical music, and novel, to the semantic ontology reasoning engine 304 for reasoning. The reasoning principle of the semantic ontology reasoning engine 304 is: according to the representation form of the RDF triples of the resources, reasoning is performed according to the defined reasoning rules. The expression form of RDF triple is: (subject, predicate, individual). For example, two resource descriptions as shown in FIG. 6 are defined: Shenzhen 601 belongs to Guangdong 602; Guangdong 602 belongs to China 603. At the same time, an inference rule is defined as: (?a, belongs to, ?b), (?b, belongs to, ?c)→(?a, belongs to, ?c). The meaning expressed by this inference rule is: if a belongs to b, and b belongs to c, then it can be inferred that a belongs to c. Therefore, the result shown in Figure 7 can be deduced from the relationship shown in Figure 6: Shenzhen 601 belongs to China 603.

Assume that the relationship shown in Figure 8 is established for the four ontology concepts of the present embodiment in the annotation ontology library 305: the parent class of popular music 801 is popular music 802, and the parent class of popular music 802 and classical music 803 is music 804; The parent category of novel 805 is literature 806; the parent category of computer game 807 is game. Then, the RDF relationship of the four ontology concepts obtained after reasoning with inference rules is shown in Figure 9: the parent category of pop music 801 and classical music 803 is music 804; the parent category of novel 805 is literature 806; the parent category of computer game 807 is Class 808 for games. Its RDF triple output format is:

(pop, parent, music)

(classical music, parent, music)

(fiction, parent genre, literature)

(computer game, parent, game)

Such a semantic classification rule is defined in the document semantic classification rule engine 302: if there is a common individual in multiple triples, and the predicate is "parent class", then a new document classification is added to the document semantic classification table, and the category name is The name of the individual and the file identification sequence are the union of the file identification sequences corresponding to the subject words in multiple triplets, that is, the result sequence of the bitwise OR operation. Table 14 is the above-mentioned semantic classification rule table, as shown in Table 14:

Table 14

Triggering conditions perform an operation Exists multiple (?X1, parent, ?Y)(?X2, parent, ?Y)... A record is added to the document semantic classification table. The semantic identifier of this record is ? Y, the file ID sequence is ? X1,? The union of the file identification sequences corresponding to X1, ... … …

After semantic reasoning processing and expanding the integrated document semantic classification table according to the semantic classification rules. Table 15 is the extended document semantic classification table, as shown in Table 15:

Table 15

Semantic ID file ID sequence Pop music 01011010110000000000 Computer Games 10000001000100001011 classical music 00010000001010000001 novel 100000000000000000010 music 01011010111010000001

Step 508, the sorting processing unit 312 calculates the relevance and importance of the files in the document semantic classification table after semantic reasoning, then sorts the files according to the calculation results, and finally sends the sorted results and document semantic classification information to to the search interface module 301.

In step 509, the search interface module 301 feeds back the received sorting results and semantic classification information as search results to the user.

The above descriptions are only preferred embodiments of the present invention, and are not intended to limit the protection scope of the present invention.

Claims (24)

1. A retrieval system based on semantic ontology, characterized in that the system comprises: Semantic ontology index database, used to save semantic ontology index; The semantic ontology search processing unit is configured to obtain a text hit file list, and match the text hit file list with the semantic ontology index in the semantic ontology index database to obtain a document semantic classification table. 2. The system according to claim 1, characterized in that the system further comprises a semantic ontology index processing unit, configured to obtain documents for which text indexes have been established, and establish semantic ontology indexes for the obtained documents. 3. The system of claim 2, further comprising: A text index processing unit, configured to establish a text index for the file; Text index database for saving text indexes; The text search processing unit is configured to match the user's text query information with the text index in the text index database to obtain a list of text hit files. 4. The system of claim 1, 2 or 3, further comprising: The semantic ontology reasoning engine performs semantic reasoning on the semantic ontology vocabulary set in the document semantic classification table according to the semantic ontology vocabulary set in the annotation ontology library and the relationship between the semantic ontology vocabulary, and obtains the extended semantic ontology vocabulary set; Annotation ontology library, used to save semantic ontology vocabulary set and relationship between semantic ontology vocabulary; The document semantic classification rule engine is used to save the semantic classification rules, and trigger the corresponding semantic classification rules according to the extended semantic ontology vocabulary set deduced by the semantic ontology reasoning engine, and extend and integrate the document semantic classification table to obtain the extended The semantic taxonomy of documents. 5. The system according to claim 4, further comprising a sorting processing unit, configured to sort the files in the extended document semantic classification table. 6. The system according to claim 5, further comprising a search interface module, configured to send the user's text query information to the text search processing unit; Feedback to users. 7. The system according to claim 3, wherein the system further comprises a file database, which is used to store files for use by the semantic ontology index processing unit to establish a semantic ontology index and the text index processing unit to establish a text index . 8. The system according to claim 7, further comprising a network file capture module, configured to capture network files from the Internet and store them in the file database. 9. The system according to claim 3, wherein the text index includes a text forward index and a text inverted index; and the semantic ontology index includes a semantic ontology forward index and a semantic ontology inverted index. 10. The system according to claim 1, wherein the semantic ontology search processing unit performs semantic ontology forward index matching processing or semantic ontology inverted index matching processing on the text hit file list. 11. The system according to claim 10, wherein the semantic ontology search processing unit, when the number of text hit files in the text hit file list is greater than a threshold value, performs semantic ontology inverted index matching processing, Otherwise, carry out semantic ontology forward index matching processing. 12. The system according to claim 11, wherein the threshold is a predefined fixed value or a dynamically adjustable value. 13. The system according to claim 3, wherein the text search processing unit and the semantic ontology search processing unit are integrated in a search processing module; the text index processing unit and the semantic ontology index processing unit are integrated in a In the index processing module: the text index database and the semantic ontology index database are integrated into an index database. 14. The system according to claim 5, wherein the text search processing unit, semantic ontology search processing unit and sorting processing unit are integrated into one search processing module. 15. A search method based on semantic ontology, characterized in that the method comprises the following steps: A. Obtain the files with established text indexes, and establish semantic ontology indexes for the obtained files; B. Obtain a list of text hit files, perform semantic ontology index matching processing on the list of text hit files, and obtain a document semantic classification table. 16. The method of claim 15, wherein, It further includes before step A, the step of establishing a text index for the files in the file database; Before step B, it further includes a step of performing text index matching processing on the user's text query information to obtain a text hit file list. 17. The method according to claim 15 or 16, further comprising the steps of: C. Perform semantic reasoning on the semantic ontology vocabulary set in the document semantic classification table to obtain an extended semantic ontology vocabulary set; D. According to the deduced extended semantic ontology vocabulary set, perform an extended integration operation on the document semantic classification table to obtain an extended document semantic classification table. 18. The method according to claim 17, further comprising: a step of sorting the files in the extended document semantic classification table. 19. The method as claimed in claim 15, characterized in that, setting up the semantic ontology index described in the step A is setting up a semantic ontology forward index and setting up a semantic ontology inverted index; described in the step B, the text hit file list is carried out Semantic ontology index matching processing is to perform semantic ontology inverted index matching processing or semantic ontology forward index matching processing. 20. The method according to claim 15, further comprising before step B: in step B, performing semantic ontology inverted index matching processing on the text hit file list, or performing semantic ontology forward index matching processing judgment steps. 21. The method according to claim 20, characterized in that, the step of judging is: when the number of text hit files in the text hit file list is greater than a threshold value, carry out semantic ontology inverted index matching processing in step B , otherwise in step B, carry out semantic ontology forward index matching processing. 22. The method according to claim 21, wherein the threshold value is a predefined fixed value or a dynamically adjustable value. 23. The method according to claim 16, wherein said establishing a text index is establishing a text forward index and establishing a text inverted index; said performing text index matching processing on the user's text query information is performing text index matching. Inverted index matching processing or text forward index matching processing. 24. The method according to claim 16, further comprising: a step of establishing a file database before said establishing a text index.

Images