Text Type Structure and Logical Document Structure

Text Type Structure and Logical Document Structure Hagen Langer Justus-Liebig-Universit¨at/ Universit¨at Osnabr¨uck hagen.langer@web.de Harald L¨ungen Justus-Liebig-Universit¨at Gießen, Germany luengen@uni-giessen.de Petra Saskia Bayerl Justus-Liebig-Universit¨at Gießen, Germany bayerl@uni-giessen.de Abstract Most research on automated categorization of doc- uments has concentrated on the assignment of one or many categories to a whole text. However, new applications, e.g. in the area of the Semantic Web, require a richer and more fine-grained annotation of documents, such as detailed thematic informa- tion about the parts of a document. Hence we in- vestigate the automatic categorization of text seg- ments of scientific articles with XML markup into 16 topic types from a text type structure schema. A corpus of 47 linguistic articles was provided with XML markup on different annotation layers repre- senting text type structure, logical document struc- ture, and grammatical categories. Six different fea- ture extraction strategies were applied to this corpus and combined in various parametrizations in differ- ent classifiers. The aim was to explore the contribu- tion of each type of information, in particular the logical structure features, to the classification ac- curacy. The results suggest that some of the topic types of our hierarchy are successfully learnable, while the features from the logical structure layer had no particular impact on the results. 1 Introduction Our project Semantics of Generic Document Struc- tures is concerned with the text type structure of sci- entific articles and its relations to document gram- mars and markup. One of its goals is to explore the feasibility of an automatic categorization of text segments of scientific articles which are an- notated with logical structure tags (e.g. section, paragraph, appendix) into topic types such as background, researchTopic, method, and results de- fined in a hierarchical text type schema. The schema representing the text type structure (or, thematic structure) of scientific articles is shown in Figure 1 (explained more fully in section 2). It is as- sumed that authors to some degree adhere to such a schema when creating a logical structure for their documents for instance by XML markup, and that therefore such markup bears clues as to the thematic structure of scientific articles. content evidence background framework method_evd problem answers theory_frm concepts framework_R method_evd_R data dataCollection dataAnalysis results interpretation conclusions textual researchTopic rationale othersWork background_R resource Figure 1: Text type schema Automatic document classification is an ad- vanced field within computational linguistics with many practical applications and has yielded a wealth of standard methods and tools over the past 10 years. Many systems have been developed on the Reuters-21578 corpus containing 21578 English newpaper articles manually categorized into 135 topic categories, e.g. (Zu et al., 2003). Besides newspaper articles, some approaches have treated the automatic categorization of (HTML) hypertexts available on the W3 into universal text topic sets such as the ones used in the LookSmart and Yahoo web directories (Dumais and Chen, 2000). An ap- proach that focuses solely on research papers is CiteSeer (Giles et al., 1998), the online digital li- brary of scientific articles that can …

Content truncated to 500 words