Universal configurable compatible multi-model domain-migratable multi-round dialogue system

Abstract

The invention discloses a universal configurable, compatible multi-model and domain-migratable multi-round dialogue system. The system comprises: a configurable module used for carrying out self-defined configuration according to professional knowledge of a current domain; a dialogue understanding module used for carrying out language coding on user statements through dialogue understanding, and carrying out intention recognition and joint extraction of slot value pairs by replacing a compatible dialogue understanding model; a dialogue state tracking module used for acquiring current multi-round historical dialogue state information; a strategy learning module used for performing action decision through the dialogue strategy model; a dialogue generation module used for generating dialoguesaccording to corresponding actions; and an online training module used for serving as a user simulator to be provided for developers, performing online iterative updating of the model and generatingavailable data. The whole architecture module of the system is universal, the system adapts to different application scenes, all field parameters are universal and configurable, corpora and dialogue understanding can be accepted at will, and a dialogue strategy model can be flexibly expanded and replaced.

Description

technical field [0001] The invention relates to the technical fields of information technology and data business, in particular to a multi-round dialog system that is universally configurable, compatible with multiple models, and domain transferable. Background technique [0002] At present, the task-oriented multi-round dialogue system is mainly based on the pipeline architecture, and its purpose is generally to complete the design of a series of internal modules for tasks in a specific field. Due to the relevance of its purpose, a large number of internal rules are associated with specific applications, resulting in the unity of the system , cannot be applied in multiple domains, fixed business rules are mixed with the system, specified fields cannot be changed, and the internal model is unique and cannot be updated. The coordination between modules is fragile, and changing the internal model structure requires a lot of labor costs. The fields of the system Immobilization,...

AI Technical Summary

Problems solved by technology

Moreover, its model is fixed, and it is impossible to continuously iteratively update the existing model, resulting in the inability to improve the model effect. The change requires a lot of logic transformation, and the fields need to be redefined, which consumes a lot of labor costs.

Images