Back to results
US 10,762,150 B2Granted
Searching method and searching apparatus based on neural network and search engine
AssigneeBeijing Baidu Netcom Science Technology Co., Ltd.
InventorsXinyu Wang, Di Jiang, Lei Shi, Chen Li, Meng Liao, Jingzhou He
Life of this patent
Estimating the term…
Abstract
A searching method and a searching apparatus based on a neural network and a search engine are disclosed, the searching method including: acquiring a query and a pre-query input by a user; acquiring a plurality of search results according to the query; generating a target term vector representation according to the query, the pre-query and the plurality of search results based on an MLP; and forecasting the target term vector representation based on a semantic model of a deep neural network so as to acquire a plurality of s optimized search results corresponding to the query.
Claims17
A searching method based on a neural network, comprising:
- acquiring a query and a pre-query input by a user;
- acquiring a plurality of search results according to the query;
- generating a target term vector representation according to the query, the pre-query and the plurality of search results based on a Multi-layer Perceptron (MLP);
- and
- forecasting the target term vector representation based on a semantic model of a deep neural network so as to acquire a plurality of optimized search results corresponding to the query;
- wherein generating a target term vector representation according to the query, the pre-query and the plurality of search results based on an MLP comprises:
- acquiring an intersection set including one or more segments contained in both the query and the pre-query, a first difference set including one or more segments contained in the query but not contained in the pre-query, a second difference set including one or more segments contained in the pre-query but not contained in the query and a union set including one or more segments contained in the query or the pre-query;
- acquiring a set of one or more segments and a search identifying set corresponding to the plurality of search results, wherein the search identifying set is configured to indicate whether the plurality of search results are comprised in a click set of the pre-query, a non-click set of the pre-query or a non-display set of the pre-query;
- and
- generating the target term vector representation according to the query, the pre-query, the intersection set, the first difference set, the second difference set, the union set, the set of the one or more segments and the search identifying set corresponding to the plurality of search results based on the MLP.
7 more elements
A searching apparatus based on a neural network, comprising:
- one or more processors;
- a memory storing instructions executable by the one or more processors;
- wherein the one or more processors are configured to:
- acquire a query and a pre-query input by a user;
- acquire a plurality of search results according to the query;
- generate a target term vector representation according to the query, the pre-query and the plurality of search results based on a Multi-layer Perceptron (MLP);
- and
- forecast the target term vector representation based on a semantic model of a deep neural network so as to acquire a plurality of optimized search results corresponding to the query;
- wherein the one or more processors are configured to perform following acts to generate a target term vector representation according to the query, the pre-query and the plurality of search results based on an MLP:
- acquiring an intersection set including one or more segments contained in both the query and the pre-query, a first difference set including one or more segments contained in the query but not contained in the pre-query, a second difference set including one or more segments contained in the pre-query but not contained in the query and a union set including one or more segments contained in the query or the pre-query;
- acquiring a set of one or more segments and a search identifying set corresponding to the plurality of search results, wherein the search identifying set is configured to indicate whether the plurality of search results are comprised in a click set of the pre-query, a non-click set of the pre-query or a non-display set of the pre-query;
- and
- generating the target term vector representation according to the query, the pre-query, the intersection set, the first difference set, the second difference set, the union set, the set of the one or more segments and the search identifying set corresponding to the plurality of search results based on the MLP.
10 more elements
A search engine, comprising:
- the searching apparatus based on a neural network comprising:
- one or more processors;
- a memory storing instructions executable by the one or more processors;
- wherein the one or more processors are configured to:
- acquire a query and a pre-query input by a user;
- acquire a plurality of search results according to the query;
- generate a target term vector representation according to the query, the pre-query and the plurality of search results based on a Multi-layer Perceptron (MLP);
- and
- forecast the target term vector representation based on a semantic model of a deep neural network so as to acquire a plurality of optimized search results corresponding to the query;
- wherein the one or more processors are configured to perform following acts to generate a target term vector representation according to the query, the pre-query and the plurality of search results based on an MLP:
- acquiring an intersection set including one or more segments contained in both the query and the pre-query, a first difference set including one or more segments contained in the query but not contained in the pre-query, a second difference set including one or more segments contained in the pre-query but not contained in the query and a union set including one or more segments contained in the query or the pre-query;
- acquiring a set of one or more segments and a search identifying set corresponding to the plurality of search results, wherein the search identifying set is configured to indicate whether the plurality of search results are comprised in a click set of the pre-query, a non-click set of the pre-query or a non-display set of the pre-query;
- and
- generating the target term vector representation according to the query, the pre-query, the intersection set, the first difference set, the second difference set, the union set, the set of the one or more segments and the search identifying set corresponding to the plurality of search results based on the MLP.
11 more elements
Dependencies are inferred from the claim text.