Multi-domain evaluation framework for named entity recognition tools
- Submitting institution
-
Birmingham City University
- Unit of assessment
- 11 - Computer Science and Informatics
- Output identifier
- 11Z_OP_D0003
- Type
- D - Journal article
- DOI
-
10.1016/j.csl.2016.10.003
- Title of journal
- Computer speech and language
- Article number
- -
- First page
- 34
- Volume
- 43
- Issue
- -
- ISSN
- 0885-2308
- Open access status
- Deposit exception
- Month of publication
- -
- Year of publication
- 2017
- URL
-
https://www.sciencedirect.com/science/article/pii/S0885230815300504
- 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
-
-
- Research group(s)
-
-
- Citation count
- 6
- Proposed double-weighted
- No
- Reserve for an output with double weighting
- No
- Additional information
- A significant challenge which directly impacts the performance of Natural language processing (NLP) techniques is domain diversity. This paper developed a standard, expandable and flexible framework to analyse and test NLP tools across various domains. The represented text varies in many aspects including taxonomies, length, formality and format. The results demonstrated clearly the performance inconsistency in terms of precision and recall among tools whenever domain changes. The swing in performance due to domain change that is demonstrated in this work, has urged the need for adaptive and transfer learning for NLP processing, which became very active areas of research recently.
- Author contribution statement
- -
- Non-English
- No
- English abstract
- -