Method for automatically extracting merchant information based on deep learning

Abstract

The invention discloses a method for automatically extracting merchant information based on deep learning. The method comprises the steps: obtaining a merchant text, extracting merchant features, andgenerating a merchant information text data set; re-checking and correcting an extraction result; carrying out noise reduction processing on the manual correction result, labeling the merchant text according to the extracted merchant features, and completing the construction of a merchant information text data set; characters of the training corpus are mapped into index representation; constructing a word-index mapping table and a label-index mapping table, reading a word vector from a pre-trained word vector model, inputting the word vector into the word vector model as an initialization value, performing numeralization on a merchant text and an entity label into index representation through the mapping tables, filling the index representation into a fixed length, and submitting the fixedlength to a sequence labeling model for training; and the sequence annotation model performs prediction annotation on the merchant text to be tested, and finds out entity information existing in themerchant text. Manpower for manual one-by-one labeling is reduced, time cost is reduced, and high efficiency, automation and accuracy are achieved.

Description

technical field [0001] The invention relates to the field of telecommunications, in particular to a method for automatically extracting business information based on deep learning. Background technique [0002] In the content of the call, when the user asks for specific information about a certain company, such as detailed address information and company phone number, because the user says the company name to a large extent, there is a lack of state, only the abbreviation, omission of the address information status at the beginning of the company, and unclear The company's full name, etc., the system cannot directly capture the company name, and cannot effectively feed back information. However, in the content of the call, users will use a large number of characteristic words to supplement the content of the company. Using supplementary feature word information can make the system judge the company name more effectively. [0003] The existing difficulty is that there is no...

AI Technical Summary

Problems solved by technology

[0003] The existing difficulty is that there is no supplementary feature word in the business information in the customer service hotline, and it is necessary to manually characterize the business name

If the telecom operator needs to label the name of the business, it will need to spend a lot of manpower to label millions of business names. If the business name is changed later, it needs to be maintained and modified in time

Manually labeling merchant feature words one by one is highly accurate, but it will also consume a lot of manpower and time

Images