Roger Garside

Comparing Corpora using Frequency Profiling

This paper describes a method of comparing corpora which uses frequency profiling. The method can be used to discover key words in the corpora which differentiate one corpus from another. Using annotated corpora, it can be applied to discover key grammatical or word-sense categories. This can be used as a quick way in to find the differences between the corpora and is shown to have applications in the study of social differentiation in the use of English vocabulary, profiling of learner English and document analysis in the software engineering process.

Corpus annotation : linguistic information from computer text corpora

List of contributors Preface 1. Introducing corpus annotation Geoffrey Leech 2. Grammatical tagging Geoffrey Leech 3. Syntactic annotation: treebanks Geoffrey Leech and Elizabeth Eyes 4. Semantic annotation Andrew Wilson and Jenny Thomas 5. Discourse annotation: anaphoric relations in corpora Roger Garside, Steve Fligelstone and Simon Botley 6. Further levels of annotation Geoffrey Leech, Anthony McEnery and Martin Wynne 7. A hybrid grammatical tagger: CLAWS4 Roger Garside and Nicholas Smith 8. How to generalise the task of annotation Steve Fligelstone, Mike Pacey and Paul Rayson 9. Improving a tagger Nicholas Smith 10. Retageting a tagger Fernando Sanchez Leon and Amalio F.Nieto-Serrano 11. The use of syntactic annotation tools: partial and full parsing Jeremy Bateman, Jean Forrest, and Tim Willis 12. Higher-level annotation tools Roger Garside and Paul Rayson 13. A corpus/annotation toolbox Anthony McEnery and Paul Rayson 14. A corpus-based grammar tool Anthony McEnery, John Paul Baker and John Hutchinson 15. The exploitation of multilingual annotated corpora for term extraction Anthony McEnery, Jean-Marc Lange, Michael Oakes and Jean Veronis 16. Cross-linguistic guidelines for the annotation of corpora Peter Kahrel, Ruthanna Barnett and Geoffrey Leech 17. Consistency and accuracy in correcting automatically-tagged corpora John Paul Baker Appendix I: Sources for further information (WWW and e-mail addresses) Appendix II: Abbreviations and acronyms Appendix III: Specimen annotation practices: the C7 and C5 tagsets Bibliography Index

REVERE: Support for Requirements Synthesis from Documents

Documents are important sources of system requirements. This is particularly true of domains that are document-centric in terms of their operational and development processes. For system evolution in organisations that have been subject to organisational change and loss of organisational memory, documents may be the major source of key requirements. Hence, systems engineers often face a daunting task of synthesising crucial requirements from a range of documents that include standards, interview transcripts and legacy specifications. The goal of REVERE was to investigate support for this task which has been described as document archaeology (Robertson S. and Robertson J. Mastering the Requirements Process. Reading, MA, Addison-Wesley, 1999). This paper describes the resulting REVERE toolset, its utility for document archaeology and for other tasks that have emerged in the course of our experiments with the toolset.

The REVERE Project: Experiments with the Application of Probabilistic NLP to Systems Engineering

Despite natural languages well-documented shortcomings as a medium for precise technical description, its use in software-intensive systems engineering remains inescapable. This poses many problems for engineers who must derive problem understanding and synthesise precise solution descriptions from free text. This is true both for the largely unstructured textual descriptions from which system requirements are derived, and for more formal documents, such as standards, which impose requirements on system development processes. This paper describes experiments that we have carried out in the REVERE project to investigate the use of probabilistic natural language processing techniques to provide systems engineering support.

Measuring MWE Compositionality Using Semantic Annotation

This paper reports on an experiment in which we explore a new approach to the automatic measurement of multi-word expression (MWE) compositionality. We propose an algorithm which ranks MWEs by their compositionality relative to a semantic field taxonomy based on the Lancaster English semantic lexicon (Piao et al., 2005a). The semantic information provided by the lexicon is used for measuring the semantic distance between a MWE and its constituent words. The algorithm is evaluated both on 89 manually ranked MWEs and on McCarthy et als (2003) manually ranked phrasal verbs. We compared the output of our tool with human judgments using Spearmans rank-order correlation coefficient. Our evaluation shows that the automatic ranking of the majority of our test data (86.52%) has strong to moderate correlation with the manual ranking while wide discrepancy is found for a small number of MWEs. Our algorithm also obtained a correlation of 0.3544 with manual ranking on McCarthy et als test data, which is comparable or better than most of the measures they tested. This experiment demonstrates that a semantic lexicon can assist in MWE compositionality measurement in addition to statistical algorithms.

A probabilistic parser

The UCREL team at the University of Lancaster is engaged in the development of a robust parsing mechanism, which will assign the appropriate grammatical structure to sentences in unconstrained English text. The techniques used involve the calculation of probabilities for competing structures, and are based on the techniques successfully used in tagging (i.e. assigning grammatical word classes) to the LOB (Lancaster-Oslo/Bergen) corpus.The first step in the parsing process involves dictionary lookup of successive pairs of grammatically tagged words, to give a number of possible continuations to the current parse. Since this lookup will often not be able unambiguously to distinguish the point at which a grammatical constituent should be closed, the second step of the parsing process will have to insert closures and distinguish between alternative parses. It will generate trees representing these possible alternatives, insert closure points for the constituents, and compute a probability for each parse tree from the probability of each constituent within the tree. It will then be able to select a preferred parse or parses for output.The probability of a grammatical constituent is derived from a bank of manually parsed sentences.

Assisting Requirements Recovery from Legacy Documents

Business change is often accompanied by loss of continuity of experience. This has serious implications for the adaptation of an organisation’s software since people with detailed knowledge of either the software or business processes may be unavailable to inform its adaptation. In many cases organisational memory will persist principally in the form of documents such as requirements specifications, operating procedures and regulatory standards. These offer an important resource for informing what features of the software are redundant, need to be retained or can be reused. Exploiting this resource poses formidable problems, however, since it is often incomplete, poorly structured, poorly maintained and voluminous. This paper proposes that tools exploiting probabilistic natural language-processing techniques offer the potential to ease these problems. Such tools are available, mature and have been proven in other domains.