Self-learning the operation method of smart phone sales agent assistant

Abstract

The invention discloses a self-learning smart phone sales agent assistant, which includes the following steps: S1: the salesperson talks with the customer, records the dialogue voice through a machine, generates candidate answers, and presents the candidate items to the salesperson; S2: according to the salesperson's mouse operation Circumstances, such as mouse clicks on candidates, mouse clicks to expand the details of an item, the proportion of time the mouse stays on an item, and calculate the mouse operation relevance score; S3: Utilize a real-time speech recognition engine to transcribe what the salesperson said during the candidate display , its semantic similarity with the candidate is calculated by a generic semantic similarity model as a relevance score for sales adoption. The present invention updates training samples in real time through online learning, solves the problems of label data acquisition and model online in time, improves the overall performance of smart phone sales seat assistants, and is worthy of popularization.

Description

technical field [0001] The invention relates to the technical field of sales seat assistants, in particular to an operation method for self-learning smart phone sales seat assistants. Background technique [0002] During the working process of the sales representatives in the telephone sales call center, when faced with new business problems, new product advantages and selling points and other content issues, the sales representatives quickly input and search on the spot to obtain answers, answer questions, and introduce product advantages and selling points. By introducing real-time speech recognition technology and natural speech processing technology, the smart phone sales agent assistant transforms the traditional manual work of listening, understanding, recalling, and searching into the work of machines automatically providing candidate answers in real time, and then manually judging and selecting the results. However, the high-quality candidate answers provided by the ...

AI Technical Summary

Problems solved by technology

However, the high-quality candidate answers provided by the machine require a powerful question-answering model as a support, and the quality of the model depends largely on the quantity and quality of the labeled data. However, the cost of large-scale manual labeling data is very high, and because labeling requires a certain The time period, the new question answering model can not answer in real time, and the quality of the candidate answer depends on the quantity and quality of the labeled data, the cost of obtaining a large amount of manual labeling data is high, and because of the time period brought by the labeling, the latest question answering model cannot be timely online

Images