Sentence Representations via Gaussian Embedding
This work addresses the problem of representing asymmetric sentence relationships for natural language processing applications, offering an incremental improvement over existing methods.
The paper tackles the limitation of point-based sentence embeddings in capturing asymmetric relationships by proposing GaussCSE, a Gaussian distribution-based contrastive learning framework. It achieves comparable performance on natural language inference tasks and enables estimation of entailment relations, which point representations struggle with.
Recent progress in sentence embedding, which represents the meaning of a sentence as a point in a vector space, has achieved high performance on tasks such as a semantic textual similarity (STS) task. However, sentence representations as a point in a vector space can express only a part of the diverse information that sentences have, such as asymmetrical relationships between sentences. This paper proposes GaussCSE, a Gaussian distribution-based contrastive learning framework for sentence embedding that can handle asymmetric relationships between sentences, along with a similarity measure for identifying inclusion relations. Our experiments show that GaussCSE achieves the same performance as previous methods in natural language inference tasks, and is able to estimate the direction of entailment relations, which is difficult with point representations.