Keyword-based question and answer method and device and medium

Abstract

The invention discloses a keyword-based question and answer method and device and a medium, and the method comprises the steps: obtaining an online corpus, extracting keywords in the online corpus, screening out a target keyword matched with a training corpus from the obtained keywords, inputting the target keyword and the training corpus into a corpus training model together, and adjusting parameters of the model based on an output result of the model to obtain a semantic understanding model; after the semantic understanding model is obtained, obtaining a standard question corresponding to auser question according to the user question, thereby outputting an answer corresponding to the standard question, and completing a question and answer mode. Therefore, because the target keyword is used as an auxiliary and is used as an input sample together with the training corpus, the weight of the information matched with the keywords in the training corpus is improved, the attention of the semantic comprehension model to the keywords is actively improved, and the similar user questions are accurately recognized, so that more accurate answers can be output, and the user experience is improved.

Description

technical field [0001] The present application relates to the technical field of intelligent question answering, in particular to a keyword-based question answering method, device and medium. Background technique [0002] The rise of the intelligent question answering model has greatly liberated manual question answering, which can not save labor costs, and is easy to maintain, and has been widely used in various fields, such as taxation. [0003] The semantic understanding model actually represents the specific relationship between user questions (also called corpus) and standard questions (also called labels), and the model is usually trained based on the collected corpus through the corpus training model (deep learning model) Afterwards, for example, the user's question (input object) is input into the corpus training model, so as to obtain the standard question (output result). At present, in the training process of the corpus training model, only the collected corpus i...

AI Technical Summary

Problems solved by technology

At present, in the training process of the corpus training model, only the collected corpus is used as the training sample, and the influence of different words in the corpus on the output results is not distinguished, resulting in the inability to distinguish the corpus containing similar words, making the accuracy of the output results lower

For example, the two user questions are: "How to report personal tax final settlement and payment?" and "How to report personal tax final settlement and payment by myself?" If the output results are accurate, they should correspond to different standards Asked "How to operate the year-end final settlement of individual tax" and "Self-declaration process of individual tax year", but according to the current semantic understanding model, it cannot recognize the difference between the above two user questions, so it can only give the same standard ask

[0004] At present, in order to overcome the above problems, more corpus is usually used for training, but this method has little effect, and it is easy to cause model confusion, which reduces the accuracy of the model and leads to poor user experience.

Images