METHOD AND APPARATUS FOR BROADCASTING SEARCH RESULT BASED ON ARTIFICIAL INTELLIGENCE

• US 20170255702A1
• Filed: 11/11/2016
• Published: 09/07/2017
• Est. Priority Date: 03/01/2016
• Status: Active Grant

• Current Assignee: Beijing Baidu Netcom Science and Technology Company Limited (Baidu Incorporated)
• Original Assignee: Beijing Baidu Netcom Science and Technology Company Limited (Baidu Incorporated)
• Inventors: LIU, Kai, FENG, Yang, YANG, Qin, LYU, Yajuan
• Granted Patent: US 10,810,272 B2
• Time in Patent Office: Days
• CPC Class Codes: G06F 16/2468 Fuzzy queries
• CPC Class Codes: G06F 16/35 Clustering; Classification
• CPC Class Codes: G06F 16/638 Presentation of query results
• CPC Class Codes: G06F 16/9535 Search customisation based on user profiles and personalisation
• CPC Class Codes: G06N 5/04 Inference or reasoning models
• CPC Class Codes: G06N 5/048 Fuzzy inferencing

Abstract

The present disclosure provides a method and an apparatus for broadcasting a search result based on artificial intelligence. The method includes: receiving a query sentence inputted by a user, and acquiring a plurality of candidate search results according to the query sentence; analyzing each candidate search result to determine a category of each candidate search result, in which the category includes a structured result and a rich-text result; acquiring intention information of the query sentence, and screening the plurality of candidate search results according to the intention information and the category of each candidate search result to obtain a screened search result; and generating text information corresponding to the screened search result, and broadcasting the text information.

Claims

1. A method for broadcasting a search result based on artificial intelligence, comprising:

receiving a query sentence inputted by a user, and acquiring a plurality of candidate search results according to the query sentence;

analyzing each candidate search result to determine a category of each candidate search result, wherein the category comprises a structured result and a rich-text result;

acquiring intention information of the query sentence, and screening the plurality of candidate search results according to the intention information and the category of each candidate search result to obtain a screened search result; and

generating text information corresponding to the screened search result, and broadcasting the text information.

2. The method according to claim 1, wherein analyzing each candidate search result to determine a category of each candidate search result comprises:

acquiring page content information corresponding to each candidate search result;

parsing the page content information and extracting a data characteristic corresponding to the page content information; and

determining the category of each candidate search result according to the data characteristic.

3. The method according to claim 1, wherein, if the category of each candidate search result is the structured result, generating text information corresponding to the screened search result comprises:

extracting a plurality of fields contained in the structured result;

calculating an importance degree of each field with respect to the structured result, and acquiring a candidate field meeting a preset condition according to the importance degree;

generating a plurality of candidate text information according to the candidate field; and

calculating a similarity of the query sentence with respect to each candidate text information, and sorting the plurality of candidate text information according to the similarity, and generating the text information corresponding to the screened search result according to a sorted result.

4. The method according to claim 3, further comprising:

performing a colloquial broadcast rewriting on the plurality of candidate text information.

5. The method according to claim 3, wherein calculating a similarity of the query sentence with respect to each candidate text information comprises:

calculating the similarity of the query sentence with respect to each candidate text information based on a generalized language model, wherein the similarity comprises at least one of a literal similarity, a semantic similarity and a broadcast style similarity.

6. The method according to claim 3, wherein sorting the plurality of candidate text information according to the similarity comprises:

sorting the plurality of candidate text information based on a preset ranking model, the preset ranking model comprising a Linear-Regression ranking model, and a Pair-Wise ranking model.

7. The method according to claim 1, wherein if the category of each candidate search result is the rich-text result, generating text information corresponding to the screened search result comprises:

performing a preprocessing on the rich-text result, the preprocessing comprising at least one of a text extraction, a text de-duplication, a keyword analysis, a text structural analysis;

extracting characteristics of a plurality of candidate sentences in the rich-text result based on a generalized language model, the characteristics comprising at least one of a semantic similarity of each candidate sentence with respect to the query sentence, an intention similarity of each candidate sentence with respect to the query sentence, position information of each candidate sentence in the rich-text result and keyword information contained in each candidate sentence;

calculating a score of each candidate sentence according to the characteristics;

sorting the plurality of candidate sentences according to the score, and generating the text information corresponding to the screened search result according to a sorted result.

8. The method according to claim 1, wherein broadcasting the text information comprises:

synthesizing the text information into speech information based on a TTS speech synthesis technology, and broadcasting the speech information.

9. An apparatus for broadcasting a search result based on artificial intelligence, comprising:

a processor; and

a memory configured to store an instruction executable by the processor;

wherein the processor is configured to;

receive a query sentence inputted by a user, and acquire a plurality of candidate search results according to the query sentence;

analyze each candidate search result to determine a category of each candidate search result, wherein the category comprises a structured result and a rich-text result;

acquire intention information of the query sentence, and screen the plurality of candidate search results according to the intention information and the category of each candidate search result to obtain a screened search result; and

generate text information corresponding to the screened search result, and broadcast the text information.

10. The apparatus according to claim 9, wherein the processor is configured to analyze each candidate search result to determine a category of each candidate search result, by acts of:

acquiring page content information corresponding to each candidate search result;

parsing the page content information and extract a data characteristic corresponding to the page content information; and

determining the category of each candidate search result according to the data characteristic.

11. The apparatus according to claim 9, wherein, if the category of each candidate search result is the structured result, the processor is configured to generate text information corresponding to the screened search result, by acts of:

extracting a plurality of fields contained in the structured result;

calculating an importance degree of each field with respect to the structured result, and acquiring a candidate field meeting a preset condition according to the importance degree;

generating a plurality of candidate text information according to the candidate field; and

calculating a similarity of the query sentence with respect to each candidate text information, and sorting the plurality of candidate text information according to the similarity, and generating the text information corresponding to the screened search result according to a sorted result.

12. The apparatus according to claim 11, wherein the processor is further configured to:

perform a colloquial broadcast rewriting on the plurality of candidate text information.

13. The apparatus according to claim 11, wherein the processor is configured to calculate a similarity of the query sentence with respect to each candidate text information, by acts of:

calculating the similarity of the query sentence with respect to each candidate text information based on a generalized language model, wherein the similarity comprises at least one of a literal similarity, a semantic similarity and a broadcast style similarity.

14. The apparatus according to claim 11, wherein the processor is configured to sort the plurality of candidate text information according to the similarity, by acts of:

sorting the plurality of candidate text information based on a preset ranking model, the preset ranking model comprising a Linear-Regression ranking model, and a Pair-Wise ranking model.

15. The apparatus according to claim 9, wherein if the category of each candidate search result is the rich-text result, the processor is configured to generate text information corresponding to the screened search result, by acts of:

performing a preprocessing on the rich-text result, the preprocessing comprising at least one of a text extraction, a text de-duplication, a keyword analysis, a text structural analysis;

extracting characteristics of a plurality of candidate sentences in the rich-text result based on a generalized language model, the characteristics comprising at least one of a semantic similarity of each candidate sentence with respect to the query sentence, an intention similarity of each candidate sentence with respect to the query sentence, position information of each candidate sentence in the rich-text result and keyword information contained in each candidate sentence;

calculating a score of each candidate sentence according to the characteristics;

sorting the plurality of candidate sentences according to the score, and generate the text information corresponding to the screened search result according to a sorted result.

16. The apparatus according to claim 9, wherein the processor is configured to broadcast the text information, by acts of:

synthesizing the text information into speech information based on a TTS speech synthesis technology, and broadcasting the speech information.