English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

In this paper, we adopted a retrospective approach to examine and comparefive existing popular sentence encoders, i.e., Sentence-BERT, UniversalSentence Encoder (USE), LASER, InferSent, and Doc2vec, in terms of theirperformance on downstream tasks versus their capability to capture basicsemantic properties. Initially, we evaluated all five sentence encoders on thepopular SentEval benchmark and found that multiple sentence encoders performquite well on a variety of popular downstream tasks. However, being unable tofind a single winner in all cases, we designed further experiments to gain adeeper understanding of their behavior. Specifically, we proposed four semanticevaluation criteria, i.e., Paraphrasing, Synonym Replacement, AntonymReplacement, and Sentence Jumbling, and evaluated the same five sentenceencoders using these criteria. We found that the Sentence-Bert and USE modelspass 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 andjumbling criteria. These results suggest that although these popular sentenceencoders perform quite well on the SentEval benchmark, they still struggle tocapture some basic semantic properties, thus, posing a daunting dilemma in NLPresearch.

The Daunting Dilemma with Sentence Encoders: Success on Standard Benchmarks, Failure in Capturing Basic Semantic Properties