# One Representation per Word - Does it make Sense for Composition?

## TL;DR

This paper examines whether word sense disambiguation is necessary before composition, finding that simple composition methods can effectively recover sense information without explicit disambiguation.

## Contribution

It demonstrates that single-vector models can perform comparably or better than multi-sense models in composition tasks, challenging the need for explicit sense disambiguation.

## Key findings

- Single-sense vector models perform as well or better than multi-sense models.
- Simple composition functions like addition can recover sense-specific information.
- Single-sense models are effective despite less granular representations.

## Abstract

In this paper, we investigate whether an a priori disambiguation of word senses is strictly necessary or whether the meaning of a word in context can be disambiguated through composition alone. We evaluate the performance of off-the-shelf single-vector and multi-sense vector models on a benchmark phrase similarity task and a novel task for word-sense discrimination. We find that single-sense vector models perform as well or better than multi-sense vector models despite arguably less clean elementary representations. Our findings furthermore show that simple composition functions such as pointwise addition are able to recover sense specific information from a single-sense vector model remarkably well.