# Ontology-Aware Token Embeddings for Prepositional Phrase Attachment

## TL;DR

This paper introduces context-sensitive, ontology-aware token embeddings based on WordNet synsets, which improve prepositional phrase attachment accuracy by modeling lexical ambiguity.

## Contribution

It presents a novel approach to embedding words as distributions over semantic concepts, enhancing syntactic disambiguation tasks.

## Key findings

- Context-sensitive embeddings improve PP attachment accuracy by 5.4 percentage points.
- Model achieves a 34.4% relative reduction in errors.
- Joint learning of concept embeddings and model parameters enhances performance.

## Abstract

Type-level word embeddings use the same set of parameters to represent all instances of a word regardless of its context, ignoring the inherent lexical ambiguity in language. Instead, we embed semantic concepts (or synsets) as defined in WordNet and represent a word token in a particular context by estimating a distribution over relevant semantic concepts. We use the new, context-sensitive embeddings in a model for predicting prepositional phrase(PP) attachments and jointly learn the concept embeddings and model parameters. We show that using context-sensitive embeddings improves the accuracy of the PP attachment model by 5.4% absolute points, which amounts to a 34.4% relative reduction in errors.