Intent Based Utterance Segmentation for Multi IntentNLU

Loading...
Thumbnail Image

Persistent link to this item

Statistics
View Statistics

Journal Title

Journal ISSN

Volume Title

Title

Intent Based Utterance Segmentation for Multi IntentNLU

Published Date

2019-08

Publisher

Type

Thesis or Dissertation

Abstract

Natural Language Understanding(NLU) is a process of converting the user utterance to a dialog-act after identifying domain, intent and slots from the utterance. User utterances can either contain a single intent or could express multiple intents. Building an NLU module for multi-intent utterances is a huge challenge as traditional state-of-the-art NLU modules do not differentiate between single and multi intent utterances thereby converting them to a single semantic frame which results in reduced performance. In this paper, we introduce a intent based utterance segmenter to split user utterances if each segmented clause corresponds to a different intent. Our experiments evaluate the performance of the utterance segmenter not only on the utterances from movie-ticket booking domain and restaurant reservation domain used for training but also on a new taxi ordering domain. We show that the total number of utterances that are parsed by a utterance segmenter enabled NLU surpass the utterances parsed by traditional NLU.

Keywords

Description

University of Minnesota M.S. thesis. August 2019. Major: Computer Science. Advisors: Joseph Konstan, Serguei Pakhomov. 1 computer file (PDF); vii, 36 pages.

Related to

Replaces

License

Series/Report Number

Funding information

Isbn identifier

Doi identifier

Previously Published Citation

Other identifiers

Suggested citation

Sethupat Radhakrishna, Arun. (2019). Intent Based Utterance Segmentation for Multi IntentNLU. Retrieved from the University Digital Conservancy, https://hdl.handle.net/11299/208959.

Content distributed via the University Digital Conservancy may be subject to additional license and use restrictions applied by the depositor. By using these files, users agree to the Terms of Use. Materials in the UDC may contain content that is disturbing and/or harmful. For more information, please see our statement on harmful content in digital repositories.