SIRIUS-LTG-UiO at SemEval-2018 Task 7: Convolutional Neural Networks with Shortest Dependency Paths for Semantic Relation Extraction and Classification in Scientific Papers

Proceedings of the 12th International Workshop on Semantic Evaluation (SemEval-2018), pages 805–810 New Orleans, Louisiana, June 5–6, 2018. ©2018 Association for Computational Linguistics SIRIUS-LTG-UiO at SemEval-2018 Task 7: Convolutional Neural Networks with Shortest Dependency Paths for Semantic Relation Extraction and Classification in Scientific Papers Farhad Nooralahzadeh, Lilja Øvrelid, Jan Tore Lønning Department of Informatics University of Oslo, Norway {farhadno,liljao,jtl}@ifi.uio.no Abstract This article presents the SIRIUS-LTG-UiO system for the SemEval 2018 Task 7 on Se- mantic Relation Extraction and Classification in Scientific Papers. First we extract the short- est dependency path (sdp) between two enti- ties, then we introduce a convolutional neural network (CNN) which takes the shortest de- pendency path embeddings as input and per- forms relation classification with differing ob- jectives for each subtask of the shared task. This approach achieved overall F1 scores of 76.7 and 83.2 for relation classification on clean and noisy data, respectively. Further- more, for combined relation extraction and classification on clean data, it obtained F1 scores of 37.4 and 33.6 for each phase. Our system ranks 3rd in all three sub-tasks of the shared task. 1 Introduction Relation extraction and classification can be de- fined as follows: given a sentence where entities are manually annotated, we aim to identify the pairs of entities that are instances of the seman- tic relations of interest and classify them based on a pre-defined set of relation types. A range of dif- ferent approaches have been applied to solve this task in previous work. Conventional classification approaches have made use of contextual, lexical and syntactic features combined with richer lin- guistic and background knowledge such as Word- Net and FrameNet (Hendrickx et al., 2010; Rink and Harabagiu, 2010). Recently, the re-emergence of deep neural net- works provides a way to develop highly auto- matic features and representations to handle com- plex interpretation tasks. These approaches have yielded impressive results for many different NLP tasks. The use of deep neural networks for re- lation classification has been investigated in sev- eral recent studies (Socher et al., 2012; Lin et al., 2016; Zhou et al., 2016). Convolutional neural networks (CNNs) have been effectively applied to extract lexical and sentence level features for re- lation classification (Zhang and Wang, 2015; Lee et al., 2017; Nguyen and Grishman, 2015). How- ever, these works consider whole sentences or the context between two target entities as input for the CNN. Such representations suffer from irrelevant sub-sequences or clauses when target entities oc- cur far from each other or there are other target entities in the same sentence. To avoid negative effects from irrelevant chunks or clauses and cap- ture the relation between two entities, Xu et al. (2015a); Liu et al. (2015) and Xu et al. (2015b) employ a CNN to learn more robust and effec- tive relation representations from the shortest de- pendency path (sdp) between two entities. The sdp between two entities in the dependency graph captures a condensed representation of the infor- mation required to assert a relationship between two …

Content truncated to 500 words