Integrating Paraphrasing into the FRANK QA System

Nick Ferguson, Liane Guillou, Kwabena Nuamah, Alan Bundy

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract / Description of output

We present a study into the ability of paraphrase generation to increase the variety of natural language queries that the Frank Query Answering system can answer. We choose an English-French backtranslation model to generate paraphrases, which we test using a small challenge dataset. We conclude that this method is not useful for improving the variety of natural language queries that Frank can answer. Based on our observations, we recommend future work in the following directions: (1) allowing the ability to specify a form to paraphrase an input into; (2) constrained paraphrasing to avoid loss of information about query intent; and (3) the need for an automatic evaluation metric which captures semantic similarity, allows syntactic variation, and rewards preservation of query intent.
Original languageEnglish
Title of host publicationProceedings of the 3rd Human-Like Computing Workshop (HLC 2022)
EditorsAlan Bundy, Denis Mareschal
PublisherCEUR Workshop Proceedings (
Number of pages6
Publication statusPublished - 2 Oct 2022
EventThe 3rd International Workshop on Human-Like Computing 2022 - Windsor, United Kingdom
Duration: 28 Sept 202230 Sept 2022
Conference number: 3

Publication series

NameHuman-Like Computing Workshop 2022
PublisherCEUR Workshop Proceedings
ISSN (Electronic)1613-0073


WorkshopThe 3rd International Workshop on Human-Like Computing 2022
Abbreviated titleHLC 2022
Country/TerritoryUnited Kingdom
Internet address

Keywords / Materials (for Non-textual outputs)

  • Question Answering
  • Paraphrasing
  • Backtranslation


Dive into the research topics of 'Integrating Paraphrasing into the FRANK QA System'. Together they form a unique fingerprint.

Cite this