Method of multi-round session framework based on cold start

Abstract

The invention discloses a multi-round session framework method based on cold start, which is characterized by comprising the steps of ontology setting, topic tracker employment, dialogue flow management, natural language understanding, dialogue state tracking, strategy action execution and natural language generation. The method comprises the following steps: step 1, ontology setting, wherein thecontent of the ontology setting comprises topics, intentions, slot positions, triggering rules of the topics, the intentions and the slot positions and robot answering verbal skills; 2, topic trackerempolyment, wherein a pre-training text classification model is adopted to judge the type of a user question, and a topic tracker is used as a gate control to only allow a task-type question to entera subsequent multi-round session process; 3, dialogue flow management, wherein natural language understanding, dialogue state tracking, strategy action and natural language generation processing are sequentially performed on the task type questions, and system answers are generated and sent to the user. According to the method disclosed in the invention, the requirement of direct cold start of themulti-round intelligent session robot without training corpus can be met, and the pluggable and ready-to-use multi-round task type dialogue function is realized through the setting.

Description

technical field [0001] The invention relates to the field of intelligent conversation systems, in particular to a cold start-based multi-round conversation framework method. Background technique [0002] With the development of AI technology, intelligent conversation technology has been vigorously developed. Task-based multi-round conversation is a technology that can conduct multi-round communication conversations with users based on domain-specific tasks. Usually the multi-turn session system is implemented by a pipeline structure. The user's question utterance is processed sequentially through the pipeline structure of natural language understanding NLU, dialogue state tracking DST, policy action POLICY and natural language generation NLG, and a system reply is obtained. NLU is a natural language understanding module, which is used to understand the user's intention and slot; the DST dialogue state tracking and POLICY policy learning module learns the user's state and t...

AI Technical Summary

Problems solved by technology

These modules can be learned and implemented by deep learning models, and can also be implemented by rules. However, in practical applications, because task-based conversational systems in different fields need to be configured with slots and pipeline structure models or methods related to different fields and intentions , only a single pipeline structure is not enough to meet the needs of commercial multi-round conversation business for out-of-the-box use, and it is difficult to cold start without corpus

[0003] Comparative document CN202010019770-a multi-round dialogue intelligent customer service system based on proper word correction and cold start, which involves technical features such as intention recognition, slot recognition, and multi-round conversation management, but the existing problem is cold start of slot recognition It is based on dictionary matching. If it is a number or date type, it is difficult to enumerate all combinations of numbers. The intent recognition needs to be trained with a model. The intent is non-cold start, and a complete cold start cannot be achieved.

[0004] Comparative document CN201811202675-a method for real

Images