Cluster-based prediction of user ratings for stylistic surface realisation
- Submitting institution
-
Heriot-Watt University
- Unit of assessment
- 11 - Computer Science and Informatics
- Output identifier
- 15000271
- Type
- E - Conference contribution
- DOI
-
10.3115/v1/E14-1074
- Title of conference / published proceedings
- Proceedings of the 14th Conference of the European Chapter of the Association for Computational Linguistics
- First page
- 702
- Volume
- -
- Issue
- -
- ISSN
- -
- Open access status
- -
- Month of publication
- April
- Year of publication
- 2014
- URL
-
-
- Supplementary information
-
-
- Request cross-referral to
- -
- Output has been delayed by COVID-19
- No
- COVID-19 affected output statement
- -
- Forensic science
- No
- Criminology
- No
- Interdisciplinary
- No
- Number of additional authors
-
4
- Research group(s)
-
-
- Citation count
- -
- Proposed double-weighted
- No
- Reserve for an output with double weighting
- No
- Additional information
- This work was one of the pioneering papers for personalised NLG. This unique data-driven method of clustering for surface realisation contributed significantly to the dawn of Conversational AI by deep learning [Serban et al. 2015]. It is influential as it describes a method of bootstrapping to new users with no or little data, e.g. visitors new to the home using a smart speaker. This paper demonstrates high rigour in its clustering method and evaluation through an extensive on-line evaluation of generated sentences from over 240 subjects and the method is shown to be effective across multiple datasets and domains.
- Author contribution statement
- -
- Non-English
- No
- English abstract
- -
