Representation biases in sentence transformers
Variants of the BERT architecture specialised for producing full-sentence representations often achieve better performance on downstream tasks than sentence embeddings extracted from vanilla BERT. However, there is still little understanding of what properties of inputs determine the properties of such representations. In this study, we construct several sets of sentences with pre-defined lexical and syntactic structures and show that SOTA sentence transformers have a strong nominal-participant-set bias: cosine similarities between pairs of sentences are more strongly determined by the overlap in the set of their noun participants than by having the same predicates, lengthy nominal modifiers, or adjuncts. At the same time, the precise syntactic-thematic functions of the participants are largely irrelevant.
Code (0)
등록된 구현이 없습니다.
Tasks
SentenceSentence EmbeddingsMethods 이 논문이 사용한 방법론
Similar Papers 제목 키워드 기반
Towards Debiasing Sentence Representations
As natural language processing methods are increasingly deployed in real-world scenarios such as healthcare, legal systems, and social science, it becomes necessary to recognize the role they potentially play in shaping …
Linguistic AcceptabilityNatural Language UnderstandingSentenceSentiment Analysis+1On Measuring Social Biases in Sentence Encoders
The Word Embedding Association Test shows that GloVe and word2vec word embeddings exhibit human-like implicit biases based on gender, race, and other social constructs (Caliskan et al., 2017). Meanwhile, research on lear…
SentenceWord EmbeddingsTransformer Grammars: Augmenting Transformer Language Models with Syntactic Inductive Biases at Scale
We introduce Transformer Grammars (TGs), a novel class of Transformer language models that combine (i) the expressive power, scalability, and strong performance of Transformers and (ii) recursive syntactic compositions, …
Inductive BiasLanguage ModelingLanguage ModellingSentenceInformation Representation Fairness in Long-Document Embeddings: The Peculiar Interaction of Positional and Language Bias
To be discoverable in an embedding-based search process, each part of a document should be reflected in its embedding representation. To quantify any potential reflection biases, we introduce a permutation-based evaluati…
Deep Representational Re-tuning using Contrastive Tension
Extracting semantically useful natural language sentence representations from pre-trained deep neural networks such as Transformers remains a challenge. We first demonstrate that pre-training objectives impose a signific…
Semantic SimilaritySemantic Textual SimilaritySentenceSentence Embedding+2