Keyphrase Extraction

KP20k is a large-scale scholarly articles dataset with 528K articles for training, 20K articles for validation and 20K articles for testing. Source: Keyphrase Prediction With Pre-trained Language Model Image Source: https://arxiv.org/pdf/1704.06879.pdf

We describe the SemEval task of extracting keyphrases and relations between them from scientific documents, which is crucial for understanding which publications describe which processes, tasks and materials. Although this was a new task, we had a total of 26 submissions across 3 evaluation scenarios. We expect the task and the findings reported in this paper to be relevant for researchers working on understanding scientific content, as well as the broader knowledge base population and information extraction communities.

KPTimes is a large-scale dataset of news texts paired with editor-curated keyphrases. Source: KPTimes: A Large-Scale Dataset for Keyphrase Generation on News Documents

Paper: Improved automatic keyword extraction given more linguistic knowledge Doi: 10.3115/1119355.1119383

A dataset for benchmarking keyphrase extraction and generation techniques from long document English scientific papers. The dataset has high quality and consists of 2,000 scientific papers from the Computer Science domain published by ACM. Each paper has its keyphrases assigned by the authors and verified by the reviewers. Different parts of papers, such as title and abstract, are separated, enabling extraction based on the part of an article's text. The content of each paper is converted from PDF to plain text. The pieces of formulae, tables, figures and LaTeX mark up were removed automatically. Link: https://huggingface.co/datasets/midas/krapivin

The dataset was constructed by first finding suitable publications and then collecting keyphrases from manual annotators. Google SOAP API was used to find documents using variants of the query “keywords general terms filetype:pdf”. Over 250 of these PDF documents were downloaded for further processing. Documents were then manually restricted to scientific conference papers, with a length range of 4-12 pages. The PDF documents were then converted to plain text using the PDF995 software suite (as it handled two-columned text better than other programs tried). At the end of this process, 211 documents in plain text format were selected which were converted successfully without problems. The authors then recruited student volunteers from our department to participate in manual keyphrase assignments. Each volunteer was given three PDF files (with author-assigned keyphrases hidden) to assign keyphrases to.

Dataset Description EUROPA is a dataset designed for training and evaluating multilingual keyphrase generation models in the legal domain. It consists of legal judgments from the Court of Justice of the European Union (EU) and includes instances in all 24 official EU languages. Key Features: Multilingual: Covers 24 official EU languages. Domain-Specific: Focuses on legal documents. Source: Derived from Court of Justice of the European Union judgments. - Curated by: N3 team - Languages: French, German, English, Italian, Dutch, Greek, Danish, Portuguese, Spanish, Swedish, Finnish, Lithuanian, Estonian, Czech, Hungarian, Latvian, Slovenian, Polish, Maltese, Slovak, Romanian, Bulgarian, Croatian, Irish - License: MIT License ### Dataset Sources - Paper: https://arxiv.org/abs/2403.00252 ## Dataset Structure - celex_id: CELEX identifier inherited from CJEU. Different translated versions of the same judgment share the same celex_id. If you wish to set a unique identifier for each instance, you can concatenate lang and celex_id values; - lang: ISO 639-1 language code; - input: judgment transcription or translation; - keyphrases: reference keyphrases drafted by the CJEU. As explained in our paper, the dataset is split chronologically for assessing temporal generalization of models: - training set: 1957 to 2010 (131 076 instances); - validation set: 2011 to 2015 (63 373 instances); - test set: 2016 to 2023 (90 508 instances). ## Citation @article{salaun2024europa, title={EUROPA: A Legal Multilingual Keyphrase Generation Dataset}, author={Sala{\"u}n, Olivier and Piedboeuf, Fr{\'e}d{\'e}ric and Le Berre, Guillaume and Hermelo, David Alfonso and Langlais, Philippe}, journal={arXiv preprint arXiv:2403.00252}, year={2024} }

Dataset contains CS/Math articles abstracts (in Russian) obtained from two online sources. For each article publication year, journal name, authors, title, keyphrases and abstract are provided. Morozov, D. A., Glazkova, A. V., Tyutyulnikov, M. A., & Iomdin, B. L. (2023). Keyphrase generation for abstracts of the Russian-language scientific articles. NSU Vestnik. Series: Linguistics and Intercultural Communication, 21(1), 54-66.