Gregory Garretson

Animacy encoding in English: why and how

We report on two recent medium-scale initiatives annotating present day English corpora for animacy distinctions. We discuss the relevance of animacy for computational linguistics, specifically generation, the annotation categories used in the two studies and the interannotator reliability for one of the studies.

Between the Humanist and the Modernist: Semi-automated Analysis of Linguistic Corpora

This paper promotes a semi-automated approach to corpus studies of discourse phenomena and other phenomena that do not easily lend themselves to computational methods. The approach involves the following components: (a) use of “linguistic proxies” for the phenomena under study, which allow finding and coding tokens in a corpus, (b) automated methods of identifying tokens and adding codes to them, and (c) manual analysis of the tokens, aided by appropriate software tools. In particular, the use of alternating passes of automated and manual analysis is advocated. These methods are illustrated through description of three sub-studies within a project examining the English possessive alternation conducted by the authors. Several advantages of a semi-automated approach are presented, including (a) an improved cycle of exploratory analysis, (b) high levels of accuracy coupled with reasonable levels of speed and consistency, (c) increased explicitness in coding methodology, and (d) the creation of reusable tools.

The computer as research assistant: a new approach to variable patterns in corpus data

This article advocates a particular type of semi-automated approach to working with corpus data termed “shared evaluation”, the central idea of which is that the computer takes over more of the work of sorting and classifying the data, while a subsequent pass by a human coder ensures the ultimate accuracy of the data selection and classification. The article begins with a discussion of the traditional approach to corpus data and the tools that are currently available. It then describes the shared evaluation approach and compares this to a typical concordancer-based approach. The article goes on to present SVEP, a computer program developed by the authors to implement this approach and offered freely to other researchers, describing the most significant aspects of the program and its use. A case study involving adjective complementation is then presented, including examples of how SVEP was used in the study and an evaluation of the accuracy the program achieved. The article ends with a discussion of the advantages and disadvantages of SVEP in particular (and some ways the program might be improved) and of semi-automated approaches such as shared evaluation in general.