Publication Details

Query-Based Keyphrase Extraction from Long Documents

DOČEKAL Martin and SMRŽ Pavel. Query-Based Keyphrase Extraction from Long Documents. In: The International FLAIRS Conference Proceedings. 2022, vol. 2022. Jensen Beach: LibraryPress@UF, 2022, pp. 1-4. ISSN 2334-0762. Available from: https://journals.flvc.org/FLAIRS/article/view/130737
Czech title
Extrakce klíčových frází z dlouhých dokumentů založená na dotazech
Type
conference paper
Language
english
Authors
URL
Keywords

keyphrase,keyword,long documents,query-based keyphrase extraction,BERT,transformer

Abstract

Transformer-based architectures in natural language processing force input size limits that can be problematic when long documents need to be processed. This paper overcomes this issue for keyphrase extraction by chunking the long documents while keeping a global context as a query defining the topic for which relevant keyphrases should be extracted. The developed system employs a pre-trained BERT model and adapts it to estimate the probability that a given text span forms a keyphrase. We experimented using various context sizes on two popular datasets, Inspec and SemEval, and a large novel dataset. The presented results show that a shorter context with a query overcomes a longer one without the query on long documents.

Published
2022
Pages
1-4
Journal
The International FLAIRS Conference Proceedings, vol. 2022, no. 35, ISSN 2334-0762
Proceedings
The International FLAIRS Conference Proceedings
Series
2022
Conference
The 35th International FLAIRS Conference, Hutchinson Island, Jensen Beach, Florida, USA, US
Publisher
LibraryPress@UF
Place
Jensen Beach, US
DOI
EID Scopus
BibTeX
@INPROCEEDINGS{FITPUB12744,
   author = "Martin Do\v{c}ekal and Pavel Smr\v{z}",
   title = "Query-Based Keyphrase Extraction from Long Documents",
   pages = "1--4",
   booktitle = "The International FLAIRS Conference Proceedings",
   series = "2022",
   journal = "The International FLAIRS Conference Proceedings",
   volume = 2022,
   number = 35,
   year = 2022,
   location = "Jensen Beach, US",
   publisher = "LibraryPress@UF",
   ISSN = "2334-0762",
   doi = "10.32473/flairs.v35i.130737",
   language = "english",
   url = "https://www.fit.vut.cz/research/publication/12744"
}
Back to top