Cross Framework Meaning Representation Parsing

Additional Funding Sources

The research described was supported by Boise State University.

Abstract

Many natural language processing tasks can benefit from syntactic and semantic information. While there are several choices to representing the syntax and semantics of a sentence using a graph structure, the recent 2019 CoNLL Shared Task brings together five approaches into a single representation. Our approach to the CoNLL shared task is to model the words in the sentences as nodes and use a neural network with an attention layer to determine the graph structure. Our preliminary results have shown promise in this task.

Comments

T85

This document is currently not available here.

Share

COinS
 

Cross Framework Meaning Representation Parsing

Many natural language processing tasks can benefit from syntactic and semantic information. While there are several choices to representing the syntax and semantics of a sentence using a graph structure, the recent 2019 CoNLL Shared Task brings together five approaches into a single representation. Our approach to the CoNLL shared task is to model the words in the sentences as nodes and use a neural network with an attention layer to determine the graph structure. Our preliminary results have shown promise in this task.