Skip to main navigation Skip to search Skip to main content

An intention domain classification method based on concept map and language model

  • School of Computer Science and Technology, Harbin Institute of Technology

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

The intention domain classification is one of the fundamental issues in the dialogue system. In view of the diversity and variations of users inputs in the dialogue system and the difficulty of obtaining large-scale corpus for intention domain classification tasks, this paper proposes an intention domain classification method based on concept map and language model, which combines concept maps with machine learning models. Firstly, the concept map is used to carry out a higher level conceptual abstraction (i.e. knowledge)of the users' inputs. Then the most suitable abstraction for linguistic characteristics are selected by a language model, and ultimately it is concatenated with the original input as the new input for a fastText model to perform intent domain classification. The macro-F1 value of the method on the SMP2017-ECDT data corpus reaches 94.21.

Original languageEnglish
Title of host publicationICNC-FSKD 2018 - 14th International Conference on Natural Computation, Fuzzy Systems and Knowledge Discovery
EditorsZheng Xiao, Lipo Wang, Guoqing Xiao, Xiong Ning, Kenli Li, Maozhen Li
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages184-189
Number of pages6
ISBN (Electronic)9781538680971
DOIs
StatePublished - 2 Jul 2018
Externally publishedYes
Event14th International Conference on Natural Computation, Fuzzy Systems and Knowledge Discovery, ICNC-FSKD 2018 - Huangshan, Anhui, China
Duration: 28 Jul 201830 Jul 2018

Publication series

NameICNC-FSKD 2018 - 14th International Conference on Natural Computation, Fuzzy Systems and Knowledge Discovery

Conference

Conference14th International Conference on Natural Computation, Fuzzy Systems and Knowledge Discovery, ICNC-FSKD 2018
Country/TerritoryChina
CityHuangshan, Anhui
Period28/07/1830/07/18

Keywords

  • Concept Map
  • Dialogue Understanding
  • FastText
  • Intention Domain Classification
  • Knowledge Map
  • Language Model
  • Spoken Dialogue System

Fingerprint

Dive into the research topics of 'An intention domain classification method based on concept map and language model'. Together they form a unique fingerprint.

Cite this