Implementing a Sense Tagger in a General Architecture for Text Engineering

II II II II II II | II il II II II II II II II II | I! | Implementing a Sense Tagger in a General Architecture for Text Engineering Hamish Cunningham, Mark Stevenson and Yorick Wilks Department of Computer Science, University of Sheffield, Regent Court, 211 Portobello Street, Sheffield S1 4DP, UK {hamish, marks, yorick}@dcs.shef.ac.uk http://www, des. shef. ac. uk/research/groups/nlp/gate/ Abstract We describe two systems: GATE (General Ar- chitecture for Text Engineering), an architec- ture to aid in the production and delivery of language engineering systems which signifi- cantly reduces development time and ease of reuse in such systems. We also describe a sense tagger which we implemented within the GATE architecture, and which achieves high accuracy (92% of all words in text to a broad semantic level). We used the implementation of the sense tagger as a real-world task on which to evaluate the usefulness of the GATE architecture and identified strengths and weak- nesses in the architecture. 1 Introduction This paper is about two things: a novel hybrid sense tagger for unrestricted text (Wilks and Stevenson, 1997), and the experience of developing this sys- tem within GATE - a General Architecture for Text Engineering (Cunninham et al., 1997; Cunningham, Wilks, and Gaizauskas, 1996a). We hope you can forgive this mild schizophrenia - we feel that both topics are relevant to the sub- ject of new methods in NLP: the first because both the problem of arriving at methods for sense tag- ging and of tuning those methods to specific domains and lexical resources is an increasingly active topic in the field (Basili, Della Rocca, and Pazienza, 1997; Harley and Gleanon, 1997); the second because work on NLP components shares a heap of problems with other language processing work to do with reusabil- ity, data visualisation, and software-level robustness and efficiency that, we feel, are best solved by the provision of a inclusive and general architecture and development environment for the field. We begin with a review of the general concept behind GATE (section 2), then describe the practi- calities of the system that are relevant to the sense tagging system we have developed (section 3). Next we discuss the sense tagging problem (section 4), and then our system (section 5). Finally we look at the experience of developing the tagger within the architecture (section 6), and draw out some lessons for the future (section 7). 2 GATE - the concept GATE is an architecture and development environ- ment for research and development workers in NLP and Language Engineering 1. It is an architecture in the sense that it specifies a macro-level organisa- tionai pattern for the various components and data resources that make up a language processing (ac- tually at present only text processing) system (Shaw and Garlan, 1996). It is also a development envi- ronment that adds a rich set of graphical tools to the architecture enabling the developer to easily in- tegrate new processing componemts, to manage flow of control between components, to visualise …

Content truncated to 500 words