OmniGraph: Rich Representation and Graph Kernel Learning

TL;DR

OmniGraph is a comprehensive graph-based representation for NLP tasks that combines lexical, syntactic, and semantic information, enabling effective learning through graph kernels and outperforming benchmarks in finance and sentiment analysis.

Contribution

The paper introduces OmniGraph, a novel graph-based representation integrating multiple linguistic features and a convolution graph kernel learning method for improved NLP classification.

Findings

OmniGraph outperforms benchmarks in stock price prediction from news.

OmniGraph achieves high accuracy on a fine-grained sentiment corpus.

The method provides semantic insights across market sectors.

Abstract

OmniGraph, a novel representation to support a range of NLP classification tasks, integrates lexical items, syntactic dependencies and frame semantic parses into graphs. Feature engineering is folded into the learning through convolution graph kernel learning to explore different extents of the graph. A high-dimensional space of features includes individual nodes as well as complex subgraphs. In experiments on a text-forecasting problem that predicts stock price change from news for company mentions, OmniGraph beats several benchmarks based on bag-of-words, syntactic dependencies, and semantic trees. The highly expressive features OmniGraph discovers provide insights into the semantics across distinct market sectors. To demonstrate the method's generality, we also report its high performance results on a fine-grained sentiment corpus.