DPR at SemEval-2021 Task 8: Dynamic Path Reasoning for Measurement Relation Extraction

Proceedings of the 15th International Workshop on Semantic Evaluation (SemEval-2021), pages 397–403 Bangkok, Thailand (online), August 5–6, 2021. ©2021 Association for Computational Linguistics 397 DPR at SemEval-2021 Task 8: Dynamic Path Reasoning for Measurement Relation Extraction Amir Pouran Ben Veyseh1, Franck Dernoncourt2, and Thien Huu Nguyen1 1 Department of Computer and Information Science, University of Oregon, Eugene, OR 97403, USA 2 Adobe Research, San Jose, CA, USA {apouranb,thien}@cs.uoregon.edu, franck.dernoncourt@adobe.com Abstract Scientific documents are replete with measure- ments mentioned in various formats and styles. As such, in a document with multiple quanti- ties and measured entities, the task of associ- ating each quantity to its corresponding mea- sured entity is challenging. Thus, it is neces- sary to have a method to efficiently extract all measurements and attributes related to them. To this end, in this paper, we propose a novel model for the task of measurement relation ex- traction (MRE) whose goal is to recognize the relation between measured entities, quantities and conditions mentioned in a document. Our model employs a deep translation-based archi- tecture to dynamically induce the important words in the document to classify the relation between a pair of entities. Furthermore, we in- troduce a novel regularization technique based on Information Bottleneck (IB) to filter out the noisy information from the induced set of im- portant words. Our experiments on the recent SemEval 2021 Task 8 datasets reveal the effec- tiveness of the proposed model. 1 Introduction One of the key indicators of scientific writing is the quantities description of various experiments and results. While the mentions of all measurements could provide a rigorous understanding of the topic, it might make the reading and automatic process- ing of the text more difficult. As such, designing effective methods to recognize the mentions of mea- surements and also the conditions in which they are valid is necessary. According to the definition of the SemEval 2021 Task 8 (Harper et al., 2021), a measurement might consist of the following com- ponents: (i) Measure Entity: A span referring to an entity that one of its properties has been mea- sured and its value is provided in the document; (ii) Measured Property: A span referring to the charac- teristics of an entity that has been measured; (iii) [ME1] samples [/ME1] have been generated with Coronin and Dystrophin proteins. In the filtration experiments, some of them with a [PR1] diameter[/PR1] [QT1] less than 2 mm [/QT1] have been filtered out using [QT2] 200-degree [/QT2] filtering [ME2] radiation [/ME2], resulting in [QT3] 20% [/QT3] [ME3] utilization [/ME3]. These results are obtained in a [QL1] dry climate [/QL1]. Figure 1: A document annotated with the measured en- tities (i.e., [ME]), quantity (i.e., [QT]), measured prop- erty (i.e., [PR]) and qualifier (i.e., [QL]) (best viewed in color). Quantity: A span in the document that refers to a value and possibly it comes with a unit; and (iv) Qualifier: A span referring to a condition in which more information about the Quantity, Measured Property or Measured Entity is …

Content truncated to 500 words