The Daunting Dilemma with Sentence Encoders: Success on Standard Benchmarks, Failure in Capturing Basic Semantic Properties
This highlights a critical limitation in NLP for researchers and practitioners, showing that current models may not generalize beyond benchmarks, which is an incremental finding as it builds on existing evaluation methods.
The study evaluated five popular sentence encoders on standard benchmarks and found they perform well on downstream tasks, but all failed to capture basic semantic properties like antonym replacement and sentence jumbling, revealing a significant gap in their semantic understanding.
In this paper, we adopted a retrospective approach to examine and compare five existing popular sentence encoders, i.e., Sentence-BERT, Universal Sentence Encoder (USE), LASER, InferSent, and Doc2vec, in terms of their performance on downstream tasks versus their capability to capture basic semantic properties. Initially, we evaluated all five sentence encoders on the popular SentEval benchmark and found that multiple sentence encoders perform quite well on a variety of popular downstream tasks. However, being unable to find a single winner in all cases, we designed further experiments to gain a deeper understanding of their behavior. Specifically, we proposed four semantic evaluation criteria, i.e., Paraphrasing, Synonym Replacement, Antonym Replacement, and Sentence Jumbling, and evaluated the same five sentence encoders using these criteria. We found that the Sentence-Bert and USE models pass the paraphrasing criterion, with SBERT being the superior between the two. LASER dominates in the case of the synonym replacement criterion. Interestingly, all the sentence encoders failed the antonym replacement and jumbling criteria. These results suggest that although these popular sentence encoders perform quite well on the SentEval benchmark, they still struggle to capture some basic semantic properties, thus, posing a daunting dilemma in NLP research.