Report on the Conald Workshop on Learning from Text and the Web

An increasing fraction of the world's information and data is now represented in textual form. For example, the World Wide Web, online news feeds, and other Internet sources contain a tremendous volume of textual information. The goal of the CONALD workshop on Learning from Text and the Web was to explore computer methods for automatically extracting, clustering and classifying information from text and hypertext sources. The workshop included ten oral paper presentations, an organized discussion by a panel of distinguished researchers, and a handful of other contributed papers. The workshop provided a good survey of the state of the art in machine learning methods applied to text processing tasks. The presented work involved a wide array of learning approaches, including nite-state-machine induction HD, MMK], neural networks that can accept advice from users SER], relational learning methods Moo, SC], statistical clustering algorithms GS, Hof, LV, YPC], boosting methods ADW], algorithms for learning with hierarchical classes Hof, MG], and active learning methods LT, NM]. A principal limitation of many of these approaches is that they do not directly reeect attempts to develop formal models of the text phenomenon of interest. The research presented at the workshop also spanned a broad range of application tasks, in-presentation of documents in information retrieval systems GS, Hof], collaborative ltering dVN], lexicon learning GBGH], query reformulation KK], text generation Rad] and analysis of the statistical properties of text MA]. In short, the state of the art in learning from text and the web is that a broad range of methods are currently being applied to many important and interesting tasks. There remain numerous open research questions, however. Broadly, the goals of the work presented at the workshop fall into two overlapping categories: (i) making textual information available in a structured format so that it can be used for complex queries and problem solving, and (ii) assisting users in nding, organizing and managing information represented in text sources. As an example of research aimed at the former goal, Muslea, Minton and Knoblock MMK] have developed an approach to learning wrappers for semi-structured Web sources, such as restaurant directories. Their method is able to induce extraction rules from small numbers of labeled examples. These learned extraction rules are then applied so that Web pages can be treated like structured databases. As an example of work geared toward the latter goal, Shavlik and Eliassi-Rad SER] have developed an approach to …