Counts@IITK at SemEval-2021 Task 8: SciBERT Based Entity And Semantic Relation Extraction For Scientific Data

Proceedings of the 15th International Workshop on Semantic Evaluation (SemEval-2021), pages 1232–1238 Bangkok, Thailand (online), August 5–6, 2021. ©2021 Association for Computational Linguistics 1232 Counts@IITK at SemEval-2021 Task 8: SciBERT Based Entity And Semantic Relation Extraction For Scientific Data Akash Gangwar∗ Sabhay Jain∗ Shubham Sourav∗ Indian Institute of Technology Kanpur (IIT Kanpur)Ashutosh Modi {akashgnr, sabhayj, ssourav}@iitk.ac.in ashutoshm@cse.iitk.ac.in Abstract This paper presents the system for SemEval 2021 Task 8 (MeasEval). MeasEval is a novel span extraction, classification, and rela- tion extraction task focused on finding quan- tities, attributes of these quantities, and addi- tional information, including the related mea- sured entities, properties, and measurement contexts. Our submitted system, which placed fifth (team rank) on the leaderboard, consisted of SciBERT with [CLS] token embedding and CRF layer on top. We were also placed first in Quantity (tied) and Unit subtasks, second in MeasuredEntity, Modifier and Qualifies sub- tasks, and third in Qualifier subtask. 1 Introduction SemEval 2021 Task 8 (Harper et al. 2021) is a task for extracting entities and semantic relations between them from a corpus of scientific articles coming from different domains. Instead of just identifying quantities, the task gives more weigh- tage to parsing and extracting important semantic relations among the extracted entities. This is chal- lenging because texts are ambiguous, and incon- sistent, and extraction relies heavily on implicit knowledge. The results of this task can also be used for extractive scientific data summarization. Given a scientific text, the task is to identify the span of quantities, units, and other attributes of those quantities and related measured entities, prop- erties, and qualifiers, if any. The organizers have divided the task into five subtasks and submissions will be evaluated against all five sub-tasks1. 1. Quantity Extraction: For each paragraph of text, identify all Quantity spans. ∗ Authors equally contributed to this work. 1https://competitions.codalab.org/ competitions/25770 Input Sentence Quantity Extraction Model Unit Extraction Modifier classification Model MeasuredEntity and HasQuantity MeasuredProperty and HasProperty Qualifier and Qualifies Figure 1: Overview of our proposed approach 2. Unit Detection and Modifier Classification: For each identified Quantity, identify the Unit of measurement and classify additional value Modifiers (count, range, approximate, mean, etc.) that apply to the Quantity. 3. MeasuredEntity and MeasuredProperty Extraction: For each identified Quantity, identify the MeasuredEntity and Measured- Property associated with it. 4. Qualifier Extraction: Identify and mark the span of any Qualifier that is needed to record additional related context. 5. HasQuantity, HasProperty and Qualifies Extraction: Identify relationships between Quantity, MeasureEntity, MeasuredProperty, and Qualifier. We consider subtask 1 as an entity extraction task, and subtask 3, 4, and 5 are viewed as rela- tion extraction tasks. After extracting the quan- tities, other attributes (MeasuredEntity, Property, and Qualifier) related to those quantities need to be predicted. The directed graph in Figure 1 gives an overview of our proposed approach. The set of incoming edges to each node represents the input to the trained model (represented by node), and the label at each node represents the prediction made by the model. The task data …

Content truncated to 500 words