# Learning to Compute Word Embeddings On the Fly

arXiv: https://arxiv.org/abs/1706.00286 · HTML: https://mlsanity.com/paper/1706.00286/

## Key facts

- Cited by 12 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 1 Jun 2017.

## Built on by (most cited first, 12 of 12)

- [Dynamic Language Models for Continuously Evolving Content](https://mlsanity.com/paper/2106.06297/) (2021)
- [Relational Memory Augmented Language Models](https://mlsanity.com/paper/2201.09680/) (2022)
- [Learning Disentangled Representations for Natural Language Definitions](https://mlsanity.com/paper/2210.02898/) (2022)
- [ChatNVD: Advancing Cybersecurity Vulnerability Assessment with Large Language Models](https://mlsanity.com/paper/2412.04756/) (2024)
- [WSPAlign: Word Alignment Pre-training via Large-Scale Weakly Supervised Span Prediction](https://mlsanity.com/paper/2306.05644/) (2023)
- [LexSubCon: Integrating Knowledge from Lexical Resources into Contextual Embeddings for Lexical Substitution](https://mlsanity.com/paper/2107.05132/) (2021)
- [Exploring Dimensionality Reduction Techniques in Multilingual Transformers](https://mlsanity.com/paper/2204.08415/) (2022)
- [Lacking the embedding of a word? Look it up into a traditional dictionary](https://mlsanity.com/paper/2109.11763/) (2021)
- [Updater-Extractor Architecture for Inductive World State Representations](https://mlsanity.com/paper/2104.05500/) (2021)
- [Automatic Construction of Sememe Knowledge Bases via Dictionaries](https://mlsanity.com/paper/2105.12585/) (2021)
- [API-Miner: an API-to-API Specification Recommendation Engine](https://mlsanity.com/paper/2212.07253/) (2022)
- [Reconsidering Degeneration of Token Embeddings with Definitions for Encoder-based Pre-trained Language Models](https://mlsanity.com/paper/2408.01308/) (2024)

## Abstract (arXiv)

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.