Dialogue state rewriting and reply generating method and system, electronic device and storage medium

An electronic device, comprising:

• at least one processor;
• and
• a memory communicatively connected with the at least one processor;
• wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform a dialog method for an intelligent dialog system, wherein the dialog method comprises, implemented at an intelligent dialog system, the steps of:
• receiving dialog information from a user, extracting a corresponding intent based on the dialog information of the user, acquiring the corresponding dialog state based on the intent, expressing the received dialog information in the form of a vector by using a pre-trained expression model, and inputting the vectorized dialog information into a pre-trained rewriting model to rewrite the corresponding dialog state based on the received dialog information of the user;
• determining to-be-used dialog action information based on the dialog information of the user and the dialog state by using a pre-trained policy model;
• and
• generating a reply statement based on the dialog information of the user and the dialog action information by using a pre-trained reply generation model,
• wherein the step of acquiring the corresponding dialog state based on the intent comprises:
• detecting whether the intent is a new intent;
• retrieving a corresponding intent unit from an intent knowledge graph if the intent is a new intent, wherein the intent unit is of a three-layer frame unit structure, comprising a first layer which is a name of an entity or a task, a second layer which is classification information related to the entity or task of the first layer, and a third layer which comprises common expression statements around features of the second layer, and wherein the intent knowledge graph comprises a plurality of intent units and is obtained by mining information of historical dialogs with all users, and a guide item in each intent unit serve as an edge connecting different intent units;
• creating the corresponding dialog state based on the intent unit;
• and
• retrieving the dialog state corresponding to the intent from a historical state library of the user if the intent is not a new intent.

A non-transitory computer readable storage medium with computer instructions stored thereon, wherein the computer instructions are used for causing a computer to perform a dialog method for an intelligent dialog system, wherein the dialog method comprises, implemented at an intelligent dialog system, the steps of:

• receiving dialog information from a user, extracting a corresponding intent based on the dialog information of the user, acquiring the corresponding dialog state based on the intent, expressing the received dialog information in the form of a vector by using a pre-trained expression model, and inputting the vectorized dialog information into a pre-trained rewriting model to rewrite the corresponding dialog state based on the received dialog information of the user;
• determining to-be-used dialog action information based on the dialog information of the user and the dialog state by using a pre-trained policy model;
• and
• generating a reply statement based on the dialog information of the user and the dialog action information by using a pre-trained reply generation model,
• wherein the step of acquiring the corresponding dialog state based on the intent comprises:
• detecting whether the intent is a new intent;
• retrieving a corresponding intent unit from an intent knowledge graph if the intent is a new intent, wherein the intent unit is of a three-layer frame unit structure, comprising a first layer which is a name of an entity or a task, a second layer which is classification information related to the entity or task of the first layer, and a third layer which comprises common expression statements around features of the second layer, and wherein the intent knowledge graph comprises a plurality of intent units and is obtained by mining information of historical dialogs with all users, and a guide item in each intent unit serve as an edge connecting different intent units;
• creating the corresponding dialog state based on the intent unit;
• and
• retrieving the dialog state corresponding to the intent from a historical state library of the user if the intent is not a new intent.