Task Oriented Dialogue System

Task-Oriented Dialogue (TOD) system conducts a conversation with a specific purpose and is increasingly necessary due to the emergence of artificial-intelligence speakers and virtual personal assistants. In general, a TOD system is composed of three main sections: a dialogue state tracking (DST) module to track the user’s purpose, a dialogue policy module (POL) to choose system actions like API calling or ending conversation, and a natural-language generation (NLG) module to produce a response to the user.

Task Oriented Dialogue System with Text Classification

Task Oriented Dialogue(TOD) Systems help users accomplish their specific goal such as finding a restaurant, booking movie tickets. Moreover, the TOD provides psychological counseling to the users. In that case, if TOD is combined with Text Classification, the system can detect strange symptoms while progressing the psychological counseling, and it is efficient.

Knowledge-Grounded Dialogue System

Knowledge-Grounded Dialogue system tries to have a engaging and informative conversation with users. With the given knowledge information, which can be formed as structured or unstructured, dialogue system pick or retrieve relative information using current user input as a query and tries to fuse selected knowledge into response generation. Due to this natural process of knowledge-grounded conversation, this problem is usually decomposed into selecting knowledge from a large pool of candidates and generating a response based on the selected knowledge and context.