Vec2Sent: Probing Sentence Embeddings with Natural Language Generation
arXiv:2011.00592
Abstract
We introspect black-box sentence embeddings by conditionally generating from them with the objective to retrieve the underlying discrete sentence. We perceive of this as a new unsupervised probing task and show that it correlates well with downstream task performance. We also illustrate how the language generated from different encoders differs. We apply our approach to generate sentence analogies from sentence embeddings.
arXiv commentAccepted for publication in COLING 2020
References in corpus (9)
- Unsupervised Learning of Sentence Embeddings using Compositional n-Gram Features
- SentEval: An Evaluation Toolkit for Universal Sentence Representations
- Fine-grained Analysis of Sentence Embeddings Using Auxiliary Prediction Tasks
- BPEmb: Tokenization-free Pre-trained Subword Embeddings in 275 Languages
- Evaluation of sentence embeddings in downstream and linguistic probing tasks
- Concatenated Power Mean Word Embeddings as Universal Cross-Lingual Sentence Representations
- MoverScore: Text Generation Evaluating with Contextualized Embeddings and Earth Mover Distance
- SBERT-WK: A Sentence Embedding Method by Dissecting BERT-based Word Models
- Situating Sentence Embedders with Nearest Neighbor Overlap