Building Thematic Lexical Resources by Bootstrapping and Machine Learning

We discuss work in progress in the semi-automatic generation of \emph{thematic lexicons} by means of \emph{term categorization}, a novel task employing techniques from information retrieval (IR) and machine learning (ML). Specifically, we view the generation of such lexicons as an iterative process of learning previously unknown associations between terms and \emph{themes} (i.e.\ disciplines, or fields of activity). The process is iterative, in that it generates, for each $c_{i}$ in a set $C=\{c_{1},\ldots,c_{m}\}$ of themes, a sequence $L^{i}_{0}\subseteq L^{i}_{1}\subseteq \ldots \subseteq L^{i}_{n}$ of lexicons, bootstrapping from an initial lexicon $L^{i}_{0}$ and a set of text corpora $\Theta=\{\theta_{0},\ldots,\theta_{n-1}\}$ given as input. The method is inspired by \emph{text categorization}, the discipline concerned with labelling natural language texts with labels from a predefined set of themes, or categories. However, while text categorization deals with documents represented as vectors in a space of terms, we formulate the task of term categorization as one in which terms are (dually) represented as vectors in a space of documents, and in which terms (instead of documents) are labelled with themes. As a learning device, we adopt \emph{boosting}, since (a) it has demonstrated state-of-the-art effectiveness in a variety of text categorization applications, and (b) it naturally allows for a form of ``data cleaning``, thereby making the process of generating a thematic lexicon an iteration of generate-and-test steps