A generalised hybrid architecture for NLP

Willis, Alistair; Yang, Hui and De Roeck, Anne (2012). A generalised hybrid architecture for NLP. In: Workshop on Innovative Hybrid Approaches to the Processing of Textual Data, 23 Apr 2012, Avignon, France, Association for Computational Linguistics, pp. 97–105.

URL: http://www.aclweb.org/anthology/W/W12/W12-0513.pdf

Abstract

Many tasks in natural language processing require that sentences be classified from a set of discrete interpretations. In these cases, there appear to be great benefits in using hybrid systems which apply multiple analyses to the test cases. In this paper, we examine a general principle for building hybrid systems, based on combining the results of several, high precision heuristics. By generalising the results of systems for sentiment analysis and ambiguity recognition, we argue that if correctly combined, multiple techniques classify better than single techniques. More importantly, the combined techniques can be used in tasks where no single classification is appropriate.

Viewing alternatives

Item Actions

Export

About

Recommendations