drsphelps at SemEval-2022 Task 2: Learning idiom representations using BERTRAM

TL;DR

This paper presents a method to improve idiom sentence embeddings by integrating BERTRAM-generated idiom embeddings into a BERT transformer, enhancing idiomaticity detection performance.

Contribution

The authors introduce a novel approach combining BERTRAM with BERT to generate better idiom representations for multilingual idiomaticity detection.

Findings

Enhanced idiom embedding quality improves task performance

Embedding quality is highly sensitive to input context quality

Method outperforms baseline models on SemEval-2022 Task 2

Abstract

This paper describes our system for SemEval-2022 Task 2 Multilingual Idiomaticity Detection and Sentence Embedding sub-task B. We modify a standard BERT sentence transformer by adding embeddings for each idioms, which are created using BERTRAM and a small number of contexts. We show that this technique increases the quality of idiom representations and leads to better performance on the task. We also perform analysis on our final results and show that the quality of the produced idiom embeddings is highly sensitive to the quality of the input contexts.