Towards an ACL Anthology Corpus with Logical Document Structure. An Overview of the ACL 2012 Contributed Task

Proceedings of the ACL-2012 Special Workshop on Rediscovering 50 Years of Discoveries, pages 88–97, Jeju, Republic of Korea, 10 July 2012. c⃝2012 Association for Computational Linguistics Towards an ACL Anthology Corpus with Logical Document Structure An Overview of the ACL 2012 Contributed Task Ulrich Schäfer DFKI Language Technology Lab Campus D 3 1 D-66123 Saarbrücken, Germany ulrich.schaefer@dfki.de Jonathon Read, Stephan Oepen Department of Informatics Universitetet i Oslo 0316 Oslo, Norway {jread|oe}@ifi.uio.no Abstract The ACL 2012 Contributed Task is a com- munity effort aiming to provide the full ACL Anthology as a high-quality corpus with rich markup, following the TEI P5 guidelines— a new resource dubbed the ACL Anthology Corpus (AAC). The goal of the task is three- fold: (a) to provide a shared resource for ex- perimentation on scientific text; (b) to serve as a basis for advanced search over the ACL Anthology, based on textual content and cita- tions; and, by combining the aforementioned goals, (c) to present a showcase of the benefits of natural language processing to a broader au- dience. The Contributed Task extends the cur- rent Anthology Reference Corpus (ARC) both in size, quality, and by aiming to provide tools that allow the corpus to be automatically ex- tended with new content—be they scanned or born-digital. 1 Introduction—Motivation The collection of the Association for Computational Linguistics (ACL) Anthology began in 2002, with 3,100 scanned and born-digital1 PDF papers. Since then, the ACL Anthology has become the open ac- cess collection2 of scientific papers in the area of Computational Linguistics and Language Technol- ogy. It contains conference and workshop proceed- ings and the journal Computational Linguistics (for- merly the American Journal of Computational Lin- guistics). As of Spring 2012, the ACL Anthol- 1The term born-digital means natively digital, i.e. prepared electronically using typesetting systems like LATEX, OpenOffice, and the like—as opposed to digitized (or scanned) documents. 2http://aclweb.org/anthology ogy comprises approximately 23,000 papers from 46 years. Bird et al. (2008) started collecting not only the PDF documents, but also providing the textual con- tent of the Anthology as a corpus, the ACL Anthol- ogy Reference Corpus3 (ACL-ARC). This text ver- sion was generated fully automatically and in differ- ent formats (see Section 2.2 below), using off-the- shelf tools and yielding somewhat variable quality. The main goal was to provide a reference cor- pus with fixed releases that researchers could use and refer to for comparison. In addition, the vision was formulated that manually corrected ground- truth subsets could be compiled. This is accom- plished so far for citation links from paper to paper inside the Anthology for a controlled subset. The focus thus was laid on bibliographic and bibliomet- ric research and resulted in the ACL Anthology Net- work (Radev et al., 2009) as a public, manually cor- rected citation database. What is currently missing is an easy-to-process XML variant that contains high-quality running text and logical markup from the layout, such as section headings, captions, footnotes, italics etc. In prin- ciple this could be derived from LATEX …

Content truncated to 500 words