Unlock instant, AI-driven research and patent intelligence for your innovation.

Table retrieval method based on deep learning

A deep learning and table technology, applied in the field of table retrieval based on deep learning, can solve the problems of reduced representation vector representation ability, unused table statistical features, low level of complex and difficult sample retrieval effect, etc., to achieve strong adaptability, improve The effect of model accuracy

Active Publication Date: 2021-12-03
南京云问网络技术有限公司
View PDF4 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0003] Many current table retrieval systems use deep learning neural network structures for table content information, such as BiLSTM, BERT, etc., but do not use the unique statistical features for tables, which reduces the representation ability of representation vectors, and for complex The retrieval effect level of difficult samples is still low

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Table retrieval method based on deep learning
  • Table retrieval method based on deep learning
  • Table retrieval method based on deep learning

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0040] The present invention will be further illustrated below in conjunction with the accompanying drawings and specific embodiments. This embodiment is implemented on the premise of the technical solution of the present invention. It should be understood that these embodiments are only used to illustrate the present invention and are not intended to limit the scope of the present invention.

[0041] Such as figure 1 As shown, the embodiment of the present invention provides a table retrieval method based on deep learning, including:

[0042]Step 1. Receive the query statement q input by the user, and load the corpus collection of rows, columns and cells of all tables T in the database , each feature information It is a list set composed of one row, one column or cell contents of the table, i and m are natural numbers, m>3, and 1≤i≤m; simultaneously load the background information of all tables in the database , where a table Contains row, column, cell information fea...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

The invention discloses a table retrieval method based on deep learning. The method comprises the steps of receiving a query statement q input by a user, loading a feature information set F of rows, columns and cells of all tables in a database and background information C of all tables in the database, and then starting an inference process. According to the method, a RoBERTa pre-training model with a better effect than BERT is adopted, and statistical features are added on the basis of an original deep learning model for feature fusion, so that similarity information on a statistical level is utilized during similarity calculation, and comprehensiveness and accuracy are achieved; and meanwhile, during training, a training method of combining BM25 and adding difficult sample training is adopted, so that the trained model has higher adaptive capacity to error-prone samples, and the model precision is improved.

Description

technical field [0001] The invention relates to the technical field of table retrieval, in particular to a table retrieval method based on deep learning. Background technique [0002] The development of information technology continues to promote the transformation of Internet technology. Data tables and knowledge graphs are commonly used storage forms for structured knowledge bases. Therefore, how to quickly retrieve the most relevant information in tables is important for tasks such as intelligent search and question answering. Meaning, while improving user efficiency and experience. [0003] Many current table retrieval systems use deep learning neural network structures for table content information, such as BiLSTM, BERT, etc., but they do not use the unique statistical features for tables, which reduces the representation ability of representation vectors, and for complex The retrieval effect level of difficult samples is still low. Contents of the invention [0004...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
Patent Type & Authority Applications(China)
IPC IPC(8): G06K9/62G06F16/332G06F16/33
CPCG06F16/332G06F16/3331G06F18/22G06F18/253G06F18/214
Inventor 杜振东
Owner 南京云问网络技术有限公司