Information search method, device, apparatus and computer-readable medium

Patent metadata

Field	Value
Publication number	US-11977589-B2
Application number	US-201915733467-A
Country	US
Kind code	B2
Filing date	Mar 6, 2019
Priority date	Jun 20, 2018
Publication date	May 7, 2024
Grant date	May 7, 2024

Abstract

An information search method is provided. The method includes: searching for webpages related to a search request through a search engine; extracting respective texts related to the search request from respective webpages and splicing the texts to obtain a spliced text; obtaining a text segment from the spliced text; and sending the obtained text segment to the search engine, to display the obtained text segment in an information search result through the search engine. The present disclosure can bring great advantages to the search engine in terms of user experience and interaction, and can satisfy user requirements for a function of an intelligent question and answer. Through present disclosure, it is beneficial to directly presenting a short text with higher relevance to the search request to the user, thereby saving time in screening information for the user.

First claim

What is claimed is:

1. An information search method, implemented by circuits for implementing functions, comprising:
receiving, through a search engine, a search request:
searching, through the search engine, for webpages related to the search request in a database;
receiving, through a paragraph extracting module, the search request and articles in the webpages related to the search request returned by the search engine:
calculating, through the paragraph extracting module, a similarity between the search request and each paragraph in the articles in the webpages related to the search request, selecting a paragraph with A highest similarity from each article to obtain a plurality of selected paragraphs, and splicing the plurality of selected paragraphs end to end to obtain one spliced text;
inputting the search request and the spliced text into a machine reading comprehension model, obtaining from the machine reading comprehension model a continuous text segment related to the search request, wherein the obtained continuous text segment is extracted from the spliced text and is a part of the spliced text; and
sending, by a processor, the obtained continuous text segment to the search engine, to display the obtained continuous text segment in an information search result through the search engine,
wherein the machine reading comprehension model is obtained by training a deep artificial neural network algorithm with training data;

wherein training the machine reading comprehension model includes:
using a question in the training data as a search request input into the machine reading comprehension model, and using an original fragment in the training data as an expected output of the machine reading comprehension model, the original fragment in the training data being annotated in advance as an answer to the question;
wherein the training data includes Chinese corpus and English corpus obtained by: collecting a number of user search requests and corresponding web pages or articles respectively returned by search engines, and labeling a text fragment in paragraphs of a corresponding returned web page or article corresponding to a user search request as an answer to the user search request.

2. The information search method according to claim 1 , wherein searching for webpages related to a search request through the search engine comprises:
ranking all articles in the webpages in a descending order, according to relevance degrees of the articles with the search request; and
selecting N articles with the highest relevance degrees with the search request from a ranking result, wherein N is a preset integer.

3. The information search method according to claim 1 , wherein calculating a similarity between the search request and each paragraph in the articles in the webpages related to the search request comprises:
constructing a similarity calculation model for a paragraph and the search request; and
calculating the similarity between each paragraph in the articles in the webpages related to the search request with and the search request through the similarity calculation model.

4. An information search device, implemented by circuits for implementing functions, comprising: one or more processors; and
a storage device configured for storing one or more programs, wherein the one or more programs are executed by the one or more processors to enable the one or more processors to:
receive, through a search engine, a search request:
search, through the search engine, for webpages related to the search request in a database;
receive, through a paragraph extracting module, the search request and articles in the webpages related to the search request returned by the search engine:
calculate, through the paragraph extracting module, a similarity between the search request and each paragraph in the articles in the webpages related to the search request, select a paragraph with a highest similarity from each article to obtain a plurality of selected paragraphs, and splice the plurality of selected paragraphs end to end to obtain one spliced text;
input the search request and the spliced text into a machine reading comprehension model, obtain from the machine reading comprehension model a continuous text segment related to the search request, wherein the obtained continuous text segment is extracted from the spliced text and is a part of the spliced text; and
send the obtained continuous text segment to the search engine, to display the obtained continuous text segment in an information search result through the search engine,
wherein the machine reading comprehension model is obtained by training a deep artificial neural network algorithm with training data;

wherein training the machine reading comprehension model includes:
using a question in the training data as a search request input into the machine reading comprehension model, and using an original fragment in the training data as an expected output of the machine reading comprehension model, the original fragment in the training data being annotated in advance as an answer to the question:
wherein the training data includes Chinese corpus and English corpus obtained by: collecting a number of user search requests and corresponding web pages or articles respectively returned by search engines, and labeling a text fragment in paragraphs of a corresponding returned web page or article corresponding to a user search request as an answer to the user search request.

5. The information search device according to claim 4 , wherein the one or more programs are executed by the one or more processors to enable the one or more processors further to:
rank all articles in the webpages in a descending order, according to relevance degrees of the articles with the search request; and
select N articles with the highest relevance degrees with the search request from a ranking result, wherein N is a preset integer.

6. The information search device according to claim 4 , wherein the one or more programs are executed by the one or more processors to enable the one or more processors further to:
construct a similarity calculation model for a paragraph and the search request; and
calculate the similarity between each paragraph in the articles in the webpages related to the search request and the search request through the similarity calculation model.

7. A non-transitory computer-readable storage medium, in which a computer program is stored, implemented by circuits for implementing functions, wherein the computer program, when executed by a processor, causes the processor to perform operations comprising:
receiving, through a search engine, a search request:
searching, through the search engine, for webpages related to the search request in a database;
receiving, through a paragraph extracting module, the search request and articles in the webpages related to the search request returned by the search engine;
calculating, through the paragraph extracting module, a similarity between the search request and each paragraph in the articles in the webpages related to the search request, selecting a paragraph with a highest similarity from each article to obtain a plurality of selected paragraphs, and splicing the plurality of selected paragraphs end to end to obtain one spliced text;
inputting the search request and the spliced text into a machine reading comprehension model, obtaining from the machine reading comprehension model a continuous text segment related to the search request, wherein the obtained continuous text segment is extracted from the spliced text and is a part of the spliced text; and
sending the obtained continuous text segment to the search engine, to display the obtained continuous text segment in an infor

Assignees

• Baidu online network technology beijing co ltd

Inventors

• Fan Miao
• Yu Jinxing
• Dou Zehao
• Sun Mingming
• Li Ping
• Wang Haifeng

Classifications

• G06N3/09

  Supervised learning · CPC title

• G06F16/953 Primary

  Querying, e.g. by the use of web search engines · CPC title

• G06N5/04

  Inference or reasoning models · CPC title

• G06N20/00

  Machine learning · CPC title

• G06N3/08

  Learning methods · CPC title