Is automatic linguistic profiling feasible in an ESL context?

Abstract

The objective of this thesis is to test if it is possible to design a program (Automatic Profiling) that can automatically generate a linguistic profile of a written interlanguage sample. The basic approach we take is illustrated in Figure 1, and detailed below. Interlanguage annotated constituent morphological. sample lexicon structure rules hierarchy PT Profile Figure 1: The basic architecture of Automatic Profiling As can be seen in Figure 1, AP is designed to create an accurate profile of a given sample without any intervention by the user of the system. In other words, AP details the key grammatical (and lexical) aspects of the sample, and it evaluates its status in terms of second language development. It is also designed to tackle the irregular interlanguage data produced by a second language learner. Specifically, in traditional computational linguistics the automatic analysis of learner data is considered to be very difficult, if not impossible, because learner data are seen to be too irregular. The approach taken in this thesis is based on extensive research on the acquisition of English as a second language. The regularities found in ESL acquisition serve as the basic point of reference for the interlanguage parser that has been constructed for this thesis. The basic steps of the procedure are quite straight-forward. The machine will simply take the written interlanguage sample and automatically annotate its lexicon. On this basis, it will generate constituent structures, and it will use morphological and syntactic developmental regularities and compare the regularities found in the data with the PT hierarchy. Once the position of the learner grammar within the PT hierarchy has been determined, a complete linguistic profile will be generated in real time. The work that is presented in this thesis derives in part from the tasks that follow from this rough outline of my approach to automatic linguistic profiling (AP). Further parts of the work presented here derive from the need to contextualise AP in the context of language testing, syllabus construction and the ESL classroom. In addition, AP has also been designed to be used as a research tool in corpus-based studies, and this capacity will also be presented.