Extended query method based on pseudo-correlation feedback of question and answer system

Abstract

The invention discloses an extended query method based on pseudo-correlation feedback of a question-answering system, which comprises a searcher and the question-answering system, and the question-answering system comprises a surface searcher and a deep searcher; the surface searcher comprises keyword query; the deep searcher comprises expansion query, historical query and causal query; the scoring method comprises the following steps that 1, character information needing to be inquired is input into a searcher, and at the moment, the searcher transmits the information to a question answering system; according to the extension query method based on pseudo-correlation feedback of the question and answer system, causal query is set, the key information causal relationship in the information is queried, the query accuracy is improved, meanwhile, the interactivity between the document and the key information is more accurately improved, the query effect is improved, meanwhile, secondary retrieval is set, and the query efficiency is improved. And the information in the document C is queried again, so that the information query accuracy is improved, the expansion task is queried, and the use experience of the user is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of extended query methods, and specifically to an extended query method based on pseudo-relevance feedback of a question-answering system. Background Art

[0002] Extended query is a technology in computer science and is used in the fields of information retrieval and natural language processing. Extended query is used to improve the recall rate of information retrieval. By adding new key information to the original query sentence to re-query, the search engine will first perform a retrieval on the query sentence input by the user. According to the retrieved documents, suitable key information is selected and added to the query sentence for re-retrieval, thereby finding more relevant documents. However, after the user submits a query, only the original query is extended by words such as synonyms. A question-answering system is an advanced form of an information retrieval system that can answer questions posed by users in natural language accurately and concisely in natural language. When performing extended query, a synonym table or a proper noun table, etc. will be prepared in advance. However, the existing extended query methods only utilize the information in the user's original query. However, due to the lack of information in a large number of original queries, this simple method cannot truly expand the user's search intention, affecting the user experience, and the quality and efficiency of the search are relatively low. Summary of the Invention

[0003] The purpose of the present invention is to provide an extended query method based on pseudo-relevance feedback of a question-answering system to solve the problem proposed in the above background art that only the information in the user's original query is utilized, but due to the lack of information in a large number of original queries, the user's search intention cannot be truly expanded, affecting the user experience, and the quality and efficiency of the search are relatively low.

[0004] To achieve the above purpose, the present invention provides the following technical solution: An extended query method based on pseudo-relevance feedback of a question-answering system, including a searcher and a question-answering system, and the question-answering system includes: a surface searcher and a deep searcher; The surface searcher includes keyword query; The deep searcher includes: extended query, historical query, and causal query; Among them, the scoring method includes the following steps: Step 1: Input the text information to be queried inside the searcher. At this time, the searcher transmits the information to the question-answering system; Step 2: The question-answering system extracts the key information in the text for the information, and transmits the key information of the text information to the surface searcher and the deep searcher respectively; Step 3: The surface searcher transmits the information to the keyword query, queries the keywords in the sentence through the keyword query, and finally integrates the queried information to form Document A; Step 4: The deep searcher transmits the information to the extended query, and the extended query searches for extended words similar to the key information; Step 5: The deep searcher transmits the information to the historical query, which can search for information within the searched database; Step 6: The deep searcher transmits the information to the causal query to search for causal information related to the keywords; Step 7: Integrate the information retrieved by the extended query, historical query, and causal query to form Document B; Step 8: Integrate the content within Document A and Document B to form a new Document C; Step 9: The searcher transmits the queried text information to Document C, performs a secondary retrieval of the content within Document C, and finally exports the query results.

[0005] Preferably, the question - answering system synchronously transmits the information to the surface searcher and the deep searcher respectively, and the surface searcher transmits the information to the keyword query.

[0006] With the above - mentioned technical solution, the surface searcher can search for the key information in the information, and the deep searcher performs an extended search on the key information of the information.

[0007] Preferably, the deep searcher synchronously transmits the information to the extended query, historical query, and causal query respectively, and Steps 4, 5, and 6 are carried out synchronously.

[0008] With the above - mentioned technical solution, the deep searcher performs multi - angle queries on the information of the key information from different directions, improving the search accuracy and increasing the query efficiency.

[0009] Preferably, Step 4 and Step 3 are carried out synchronously, the surface searcher and the deep searcher use a unified database, and the historical query uses a historical search database.

[0010] With the above - mentioned technical solution, the surface searcher and the deep searcher simultaneously search for the content within the database.

[0011] Preferably, both the surface searcher and the deep searcher use the TextRank algorithm to retrieve the information within the database files.

[0012] With the above - mentioned technical solution, TextRank can extract the key information within the file and stably extract the key information within the database.

[0013] Preferably, the searcher uses TextRank to retrieve the information within Document C and exports the retrieved information.

[0014] With the above technical solution, the searcher uses TextRank to search document C to implement secondary search and increase the accuracy of information search.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: An extended query method for pseudo-relevance feedback based on a question-and-answer system: 1. Causal queries are set to query the causal relationships of key information in the information, improve the accuracy of the query, and at the same time more precisely improve the interactivity between the document and the key information, improve the query effect. At the same time, secondary retrieval is set to query the information inside document C again, improve the accuracy of information query, and the query expansion task to improve the user experience; 2. A deep searcher and a surface searcher are set, which can separately retrieve information in different directions, increase the direction of information retrieval. At the same time, through historical queries, the information inside the database previously used by the user can be extracted to better understand the user's search intent, improve the quality and accuracy of the search. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a schematic diagram of the query process structure of the method of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0017] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0018] Please refer to Figure 1 , the present invention provides a technical solution: An extended query method for pseudo-relevance feedback based on a question-and-answer system, including a searcher and a question-and-answer system, and the question-and-answer system includes: a surface searcher and a deep searcher; The query information is input inside the searcher. At this time, the searcher transfers the information to the inside of the question-and-answer system. The question-and-answer system retrieves the text inside the information, extracts the key information in the information, and transmits the keywords to the inside of the surface searcher and the deep searcher respectively through the question-and-answer system. At this time, through the surface searcher and the deep searcher, the information of the key information is queried from multiple different directions, expanding the angle of query information, realizing the interactivity between information, and at the same time realizing the diversity of information, making the query result closer to the user's search intent, increasing the efficiency of information query, expanding the query direction of information, and being able to more comprehensively understand the user's search intent and improve the user experience.

[0019] The surface searcher includes keyword queries; The deep searcher includes: extended queries, historical queries, and causal queries; Among them, the scoring method includes the following steps: Step 1: Input the text information to be queried inside the searcher. At this time, the searcher transmits the information to the question-and-answer system; Step 2: The question-and-answer system extracts the key information in the text, and transmits the key information of the text information to the surface searcher and the deep searcher respectively; Step 3: The surface searcher transmits the information to the keyword query, queries the keywords in the statement through the keyword query, and finally integrates the queried information to form Document A; Step 4: The deep searcher transmits the information to the extended query, and the extended query searches for extended words similar to the key information; Step 5: The deep searcher transmits the information to the historical query, which can search the information inside the searched database; Step 6: The deep searcher transmits the information to the causal query to search for causal information related to the keyword; Step 7: Integrate the information searched by the extended query, historical query, and causal query to form Document B; Step 8: Integrate the content inside Document A and Document B to form a new Document C; Step 9: The searcher transmits the queried text information to Document C, performs a secondary search inside Document C, and finally exports the query results; After the information is transmitted to the surface searcher and the deep searcher, the surface searcher transmits the key information to the keyword query, queries the key information itself, expands the key information, extracts the corresponding information, and integrates the extracted information to form Document A; At the same time, the deep searcher transmits the key information to the extended query, historical query, and causal query simultaneously. The extended query can search for extended words such as synonyms and similar words of the key information. At the same time, the causal query searches for information related to the key information and having a causal relationship to increase the diversity of information. The historical query searches for historical files in the searched historical folder to better understand the user's search purpose. Finally, the query files generated by the extended query, historical query, and causal query are integrated, and the files generated by the three are merged into Document B. Then, Document B and Document A are stored as Document C. At this time, the searcher transmits the queried text information to the inside of Document C, and queries the files inside Document C through the searcher to export the more critical information to form the query results.

[0020] The question-and-answer system synchronously transmits information to the surface searcher and the deep searcher respectively. The surface searcher transmits the information to keyword query; The surface searcher transmits the key information to keyword query, searches for the original meaning of the key information pair through keyword query, collects the internal content after the search, and the collected files form Document A.

[0021] The deep searcher synchronously transmits the information to extended query, historical query, and causal query respectively, and Steps 4, 5, and 6 are carried out synchronously; The extended query and the causal query simultaneously retrieve the files inside the database, extract the files containing the key information, and at the same time the historical query queries the historical files and extracts the files inside the historical files.

[0022] Step 4 and Step 3 are carried out synchronously. The surface searcher and the deep searcher adopt a unified database, and the historical query adopts a historical search database; When the extended query and the causal query query the database, the keyword query also queries the internal files of the database.

[0023] Both the surface searcher and the deep searcher adopt the TextRank algorithm, which can retrieve the information inside the database files; Both the surface searcher and the deep searcher adopt the TextRank algorithm inside to query the database.

[0024] The surface searcher and the deep searcher query the files inside the database through TextRank and export the queried files.

[0025] The searcher uses TextRank to retrieve the information inside Document C, exports the retrieved information, generates the final query result, and performs a secondary search, which can improve the accuracy of the result.

[0026] Working principle: When using this extended query method of pseudo-relevance feedback based on a question-and-answer system, extended query, historical query, and causal query are set up to expand the direction of querying files, improve the connectivity and interactivity between files, and can more accurately extract the files inside the database, improve the retrieval efficiency. A searcher and Document C are set up to perform a secondary search on the queried file Document C, making the exported query structure more in line with the user's query intention, improving the search quality, and increasing the overall practicality.

[0027] Although embodiments of the present invention have been shown and described, those of ordinary skill in the art will appreciate that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. An extended query method based on pseudo-relevance feedback of a question-answering system, comprising a search engine and a question-answering system, wherein the question-answering system comprises: Surface searchers and deep searchers; Surface searchers include keyword queries; Deep searchers include: extended query, historical query and causal query; The scoring method includes the following steps: Step 1: Enter the text information to be searched in the search engine, and the search engine transmits the information to the question-answering system; Step 2: The question-answering system extracts key information from the text and transmits the key information of the text to the surface searcher and the deep searcher respectively; Step 3: The surface searcher passes the information to the keyword query, searches for the keywords in the sentence through the keyword query, and finally integrates the searched information to form document A; Step 4: The deep searcher transmits the information to the expanded query, and the expanded query searches for expanded words similar to the key information; Step 5: The deep searcher passes the information to the historical query, which can search the information inside the searched database; Step 6: The deep searcher passes the information to the causal query and searches for causal information related to the keyword; Step 7: Integrate the information retrieved by the extended query, historical query and causal query to form document B; Step 8: Integrate the contents of document A and document B to form a new document C; Step 9: The search engine transmits the queried text information to document C, performs a secondary search within document C, and finally exports the query results.

2. The method for expanding query based on pseudo-relevance feedback of a question-answering system according to claim 1, characterized in that: The question-answering system synchronously transmits the information to the surface searcher and the deep searcher respectively, and the surface searcher transmits the information to the keyword query.

3. The method for expanding query based on pseudo-relevance feedback of a question-answering system according to claim 1, characterized in that: The deep searcher transmits information to the extended query, historical query and causal query synchronously respectively, and steps 4, 5 and 6 are performed synchronously.

4. The method for expanding query based on pseudo-relevance feedback of a question-answering system according to claim 1, characterized in that: The step 4 and step 3 are performed simultaneously, the surface searcher and the deep searcher use a unified database, and the historical query uses a historical search database.

5. The method for expanding query based on pseudo-relevance feedback of a question-answering system according to claim 1, characterized in that: The surface searcher and the deep searcher both use the TextRank algorithm to retrieve information inside the database file.

6. The method for expanding query based on pseudo-relevance feedback of a question-answering system according to claim 1, characterized in that: The search engine uses TextRank to retrieve the information in the document C and exports the retrieved information.

Images