Bilingual Learning of Multi-sense Embeddings with Discrete Autoencoders
arXiv:1603.09128
Abstract
We present an approach to learning multi-sense word embeddings relying both on monolingual and bilingual information. Our model consists of an encoder, which uses monolingual and bilingual context (i.e. a parallel sentence) to choose a sense for a given word, and a decoder which predicts context words based on the chosen sense. The two components are estimated jointly. We observe that the word representations induced from bilingual data outperform the monolingual counterparts across a range of evaluation tasks, even though crosslingual information is not available at test time.
arXiv comment11 pages, to appear at NAACL 2016
References in corpus (8)
- Efficient Estimation of Word Representations in Vector Space
- Natural Language Processing (almost) from Scratch
- An Autoencoder Approach to Learning Bilingual Word Representations
- Breaking Sticks and Ambiguities with Adaptive Skip-gram
- Multilingual Models for Compositional Distributed Semantics
- Multilingual Part-of-Speech Tagging: Two Unsupervised Approaches
- Embedding Word Similarity with Neural Machine Translation
- Learning to Represent Words in Context with Multilingual Supervision
Cited by in corpus (5)
- Learning Gender-Neutral Word Embeddings
- AspEm: Embedding Learning by Aspects in Heterogeneous Information Networks
- Semi Supervised Preposition-Sense Disambiguation using Multilingual Data
- A Simple Approach to Learn Polysemous Word Embeddings
- Towards Multi-Sense Cross-Lingual Alignment of Contextual Embeddings