Semantic recognition method based on word segmentation

Abstract

The invention discloses a semantic recognition method based on word segmentation, and the method comprises the steps: calling each engine by a control and scheduling engine, loading each scene moduleby a word segmentation engine, carrying out word segmentation on a question sentence, and generating a keyword and a matched named entity; calling historical records in a database by a semantic analysis engine, matching the matched named entity with a command word template according to the time weight of the historical records, and extracting the command word template; and the knowledge base retrieval engine extracts related information from the database to present according to the command template with the highest time weight. The system is high in stability and reliability, high in recognition rate and good in man-machine interaction experience, all modules in the system can be deployed in a multi-point mode and are telescopic, and expansibility is high. The question of the user can be intelligently combined into the situation to match the most concerned problem of the user, and the problem that the user wants to know can be mined as much as possible to recommend the problem.

Description

technical field [0001] The invention relates to a semantic recognition method, in particular to a distributed semantic recognition method applied to the Internet financial industry. Background technique [0002] With the development of information technology, traditional search engines can no longer meet people's needs for information acquisition and retrieval, and intelligent customer service systems based on natural language have emerged as the times require. Internet financial enterprises have a huge user base, complex information, and high real-time requirements. How to quickly and accurately recognize the semantics of natural language input by users is the key to intelligent customer service. Traditional semantic understanding usually uses keyword-based precise or fuzzy matching methods, which have the following problems for large-scale financial data and scenarios: [0003] (1) The search speed is slow. Due to the many scenarios in the financial industry and the compl...

AI Technical Summary

Problems solved by technology

[0003] (1) The search speed is slow. Due to the many scenarios in the financial industry and the complexity of the business, if the keyword-based matching method is used, the number of named entities will be very large. On a scale of one million, to cycle search in such a huge keyword library, Not only is the efficiency poor and consumes computer resources, but also the response speed is slow and the human-computer interaction experience is poor

[0004] (2) The maintenance cost of the keyword thesaurus is high. If the keyword-based matching method is adopted, the named entity library is the key to determining the recognition rate, so a lot of manpower must be spent on the maintenance and update of the named entity library

[0005] (3) The recognition rate is low. Because users generally use voice input, there are inevitably accent problems. For example, "Tongzhou Electronics" is said to be "Tong Zou Electronics", and "Jiulong Electric Power" is said to be " Jiunong Power", and many stocks have aliases, such as "Dawn shares" actually refer to "N Daun", so if the matching method based on keywords is used, the matching will fail

[0006] (4) Unable to understand the user's intentions in combination with the user's context. For example, the user asks "Vanke's market" in the first sentence, and "What about Wanda" in the second sentence. If the keyword-based matching method is used, it is impossible to understand the user. The intention of the second sentence is to ask about Wanda’s market situation

[0007] It can be seen that in the financial field, semantic understanding based on keyword matching cannot effectively solve the problem of semantic recognition in the financial field

Images