Skip to main content

Vocabulary of natural language processing

Search from vocabulary

Concept information

applied NLP > human-computer interaction > task-oriented dialogue

Preferred term

task-oriented dialogue  

Definition

  • Interactions between a user and a computer system focused on a specific task.

Synonym(s)

  • TOD

Note

  • Task-oriented dialogue is often decomposed into three sub-tasks: (1) dialogue state tracking (DST) for tracking user’s belief state; (2) dialogue policy learning (POL) for deciding which system action to take; (3) natural language generation (NLG) for generating dialogue response. (Su et al., Multi-Task Pre-Training for Plug-and-Play Task-Oriented Dialogue System, 2022)

Example

  • Task-oriented dialogue systems have recently attracted growing attention and achieved substantial progress. (Guo, Shuang, Li, Wang & Liu, 2022)
  • These datasets encompass 11000 task-oriented dialogues for shopping scenarios with photorealistic scenes spurring the development of numerous multi-modal methods tailored to conversational agents. (Osebe, Wanigasekara, Gueudre, Tran, Sharma, Yang, Hu, Ruan, Barut & Su, 2024)
  • TOD has explicit goals (e.g. restaurant reservation) and many conversational interactions like belief states and database information making language understanding and policy learning more complex than those chit-chat scenarios. (Zeng, He, Wang, Zeng, Wang, Xian & Xu, 2023)
  • We present PolyResponse a conversational search engine that supports task-oriented dialogue. (Henderson, Vulic, Casanueva, Budzianowski, Gerz, Coope, Spithourakis, Wen, Mrkšic & Su, 2019)

In other languages

URI

http://data.loterre.fr/ark:/67375/8LP-Z8W8G16Z-X

Download this concept:

RDF/XML TURTLE JSON-LD Last modified 5/21/24