Learning Crosslingual Word Embeddings without Bilingual Corpora

Abstract

Crosslingual word embeddings represent lexical items from different languages in the same vector space, enabling transfer of NLP tools. However, previous attempts had expensive resource requirements, difficulty incorporating monolingual data or were unable to handle polysemy. We address these drawbacks in our method which takes advantage of a high coverage dictionary in an EM style training algorithm over monolingual corpora in two languages. Our model achieves state-of-the-art performance on bilingual lexicon induction task exceeding models using large bilingual corpora, and competitive results on the monolingual word similarity and cross-lingual document classification task.

References in corpus (4)

• Transition-Based Dependency Parsing with Stack Long Short-Term Memory
• BilBOWA: Fast Bilingual Distributed Representations without Word Alignments
• Multilingual Models for Compositional Distributed Semantics
• Swivel: Improving Embeddings by Noticing What's Missing

Cited by in corpus (8)

• A Survey Of Cross-lingual Word Embedding Models
• Word Translation Without Parallel Data
• Unsupervised Cross-Modal Alignment of Speech and Text Embedding Spaces
• Semantic Specialisation of Distributional Word Vector Spaces using Monolingual and Cross-Lingual Constraints
• Automatic Identification of Indicators of Compromise using Neural-Based Sequence Labelling
• Cross-Lingual Sentiment Analysis Without (Good) Translation
• Multi-lingual Common Semantic Space Construction via Cluster-consistent Word Embedding
• Inverted Bilingual Topic Models for Lexicon Extraction from Non-parallel Data