LABDA at SemEval-2017 Task 10: Extracting Keyphrases from Scientific Publications by combining the BANNER tool and the UMLS Semantic Network

Proceedings of the 11th International Workshop on Semantic Evaluations (SemEval-2017), pages 947–950, Vancouver, Canada, August 3 - 4, 2017. c⃝2017 Association for Computational Linguistics LABDA at SemEval-2017 Task 10: Extracting Keyphrases from Scientific Publications by combining the BANNER tool and the UMLS Semantic Network Isabel Segura-Bedmar, Crist´obal Col´on-Ruiz, Paloma Mart´ınez Computer Science Department, Universidad Carlos III of Madrid Avd. Universidad, 30, Legan´es, 28911, Madrid, Spain isegura,ccolon,pmf@inf.uc3m.es Abstract This paper describes the system presented by the LABDA group at SemEval 2017 Task 10 ScienceIE, specifically for the subtasks of identification and classifica- tion of keyphrases from scientific articles. For the task of identification, we use the BANNER tool, a named entity recogni- tion system, which is based on condi- tional random fields (CRF) and has ob- tained successful results in the biomedi- cal domain. To classify keyphrases, we study the UMLS semantic network and propose a possible linking between the keyphrase types and the UMLS semantic groups. Based on this semantic linking, we create a dictionary for each keyphrase type. Then, a feature indicating if a to- ken is found in one of these dictionaries is incorporated to feature set used by the BANNER tool. The final results on the test dataset show that our system still needs to be improved, but the conditional ran- dom fields and, consequently, the BAN- NER system can be used as a first approxi- mation to identify and classify keyphrases. 1 Introduction In the era of big data, as it could not be otherwise, an enormous amount of scientific articles is avail- able. Although during the last few years search en- gines have provided significant improvements in information access, researches still have to spend much time exploring the huge number of articles published in their research fields. This laborious task could be reduced if search engines were able to answer common questions such as: which stud- ies have dealt with a specific TASK?, which stud- ies have explored a PROCESS? or which studies have employed such MATERIAL?. The automatic detection and classification of keyphrases (which describe tasks, processes and materials) as well as the extraction of their relations between them from scientific articles can support to find the answers to the previous questions. This task is very impor- tant, but has hardly been explored at the present time (Augenstein and Sgaard, 2017). The ScienceIE task at SemEval 2017 (Augenstein et al., 2017) aims the auto- matic extraction of keyphrases and their relations from scientific publications. The task consists of three subtasks: (1) the subtask A is focused on the identification of the keyphrases in a given article; (2) the subtask B is focused on the classification of keyphrases by one of the following types: MATERIAL, TASK, and PROCESS; and (3) the subtask C deals with the classification of the relationships between keyphrases by one of the following types: HYPONYM-OF, SYNONYM- OF, and NONE. For the evaluation of the task, the organizers have defined three different scenarios, which the participating teams can choose to submit their …

Content truncated to 500 words