A retrieval type artificial intelligence question and answer robot development method

Abstract

An artificial intelligence question and answer robot development method can perform unsupervised learning on question topics by uploading a dialogue corpus to a cloud, automatically cluster the question topics, and persist clustering results in various forms to serve as training data of a deep learning classifier. The deep learning classifier carries out topic classification on questions of a user, calculates similarity, compares the similarity with a threshold value, judges that the questions are learned if the similarity exceeds the threshold value, and retrieves and gives corresponding answers. In the training process, a visual interface is provided to measure the training quality, and the question classification accuracy can reach more than 95%. According to the method, some common questions can be answered through training, a manual corpus labeling process of the question and answer robot is not needed, hyper-parameter tuning can be carried out through a visual interface, the development efficiency and classification accuracy of the question and answer robot are greatly improved, and the method can be applied to practical applications such as intelligent customer service and chat assistants.

Description

technical field [0001] The invention relates to a method for developing a search-type artificial intelligence question answering robot. Background technique [0002] Question-answering bots are an important application in natural language processing. With the development of artificial intelligence, the implementation of question-answering robots has evolved from traditional rule-based matching to artificial intelligence-based question-answering robots. Compared with traditional rule-based matching, artificial intelligence-based question-answering robots have obvious advantages: 1. It is more versatile and robust to questions, and is closer to human habits. Second, there is no need to maintain a large number of rule bases, and the maintenance cost is lower. [0003] At present, there are mainly two ways to realize the question-answering robot of artificial intelligence, the question-answering robot based on the retrieval type and the question-answering robot based on the Se...

AI Technical Summary

Problems solved by technology

However, the deep neural network itself has the characteristics of "black box", the adjustment of hyperparameters is difficult, and a large number of questions are also required to be manually marked, and the development efficiency is low.

"Retrieval" question-answering robots based on deep neural networks are rarely reported in the literature

Images