Learning to Compute Word Embeddings On the Fly

2017/06/01 by Dzmitry Bahdanau, Bahdanau, Dzmitry, Tom Bosc +9 · 1 citation
Computer Science · #Computation and Language (cs.CL) #FOS: Computer and information sciences #Machine Learning (cs.LG) #Natural Language Processing Techniques #Text Readability and Simplification #Topic Modeling

paper · pdf · doi:10.48550/arxiv.1706.00286

openalex publication_date 2017/06/01 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

Words in natural language follow a Zipfian distribution whereby some words are frequent but most are rare. Learning representations for words in the "long tail" of this distribution requires enormous amounts of data. Representations of rare words trained directly on end tasks are usually poor, requiring us to pre-train embeddings on external data, or treat all rare words as out-of-vocabulary words with a unique representation. We provide a method for predicting embeddings of rare words on the fly from small amounts of auxiliary data with a network trained end-to-end for the downstream task. We show that this improves results against baselines where embeddings are trained on the end task for reading comprehension, recognizing textual entailment and language modeling.

Citations

• Leveraging Lexical Resources for Learning Entity Embeddings in Multi-Relational Data
• Finding Function in Form: Compositional Character Models for Open Vocabulary Word Representation
• A large annotated corpus for learning natural language inference

Cited by

• Making Neural Machine Reading Comprehension Faster

Related

• Overcoming Poor Word Embeddings with Word Definitions
• Rare Words: A Major Problem for Contextualized Embeddings And How to Fix\n it by Attentive Mimicking
• Learning Semantic Representations for Novel Words: Leveraging Both Form and Context
• FRAGE: Frequency-Agnostic Word Representation
• Distilling Word Embeddings: An Encoding Approach