Intent recognition data processing system

Abstract

The present invention relates to an intent recognition data processing system, which realizes: step S1, acquiring user query, and performing preprocessing to obtain the first word segmentation list {Q of user query 1 , Q 2 ,...Q M}, where M is the number of user query word segmentation, Q i For the i-th word segment, initialize i=1, execute step S2; step S2, based on Q i Retrieve the knowledge map, judge whether there is corresponding label information, if it exists, set Q i ’=Q i +preset delimiter+T i +Preset delimiter, among them, T i for Q i Corresponding label information, otherwise, set Q i ’=Q i ;Step S3, determine whether i is less than M, if so, set i=i+1, return to step S2, otherwise, based on all Q i 'Generate the second participle list {Q 1 ’, Q 2 ’,…Q M ’}; step S4, put {Q 1 ’, Q 2 ’,…Q M ’} into an input vector, which is input into the intent classification model to generate an intent recognition result. The invention improves the accuracy of intention recognition.

Description

technical field [0001] The present invention relates to the field of computer technology, in particular to an intention recognition data processing system. Background technique [0002] With the rapid development of artificial intelligence, intent recognition is particularly important in many application scenarios, such as speech recognition, intelligent question answering and other application scenarios. Existing intent recognition technologies are mainly based on scenarios such as chatting, but the intent system for vertical fields is not perfect. Taking the field of civil aviation as an example, the intent recognition of existing airports and airlines is mainly abstracted from the knowledge base used by customer service. The corpus is limited, and overlapping and confusion between intents occur from time to time. It is impossible to achieve clear intent boundaries and scenarios Comprehensive coverage. The intention recognition of some existing technologies is mainly bas...

AI Technical Summary

Problems solved by technology

Taking the field of civil aviation as an example, the intent recognition of existing airports and airlines is mainly abstracted from the knowledge base used by customer service. The corpus is limited, and overlapping and confusion between intents occur from time to time. It is impossible to achieve clear intent boundaries and scenarios comprehensive coverage

The intention recognition of some existing technologies is mainly based on rules and lacks flexibility. Only when the questions entered by the user can hit the key rules, the system can give the user accurate intention recognition results, and the user's questions are diverse and the expression is not standardized. and features such as typos

At this time, it is difficult for rule-based methods to accurately identify user intent

In addition, the existing intention recognition technology still lacks the integration of basic knowledge in the vertical field. Taking the civil aviation field as an example, the civil aviation industry has many professional vocabulary, including airline company names, airport names, abbreviations, and aliases; even flight dynamics The current intention recognition system has very little integration of these background information and civil aviation knowledge map information, so the accuracy of intention recognition is low

Images