Large Multi-lingual, Multi-level and Multi-genre Annotation Corpus

Large Multi-lingual, Multi-level and Multi-genre Annotation Corpus Xuansong Li, Martha Palmer, Nianwen Xue, Lance Ramshaw, Mohamed Maamouri, Ann Bies, Kathryn Summerville Conger, Stephen Grimes, Stephanie Strassel Linguistic Data Consortium, University of Pennsylvania Philadelphia, PA 19104 USA Department of Linguistics and Computer Science, University of Colorado Boulder, CO 80309 USA Computer Science Department, Brandeis University Waltham, MA 02453 USA Raytheon BBN Technologies Cambridge, MA 02138 USA Email:{xuansong,maamouri,bies,sgrimes,strassel}@ldc.upenn.edu{martha.palmer,kathryn.conger@colorado.edu xuen@brandeis.edu lramshaw@bbn.com Abstract High accuracy for automated translation and information retrieval calls for linguistic annotations at various language levels. The plethora of informal internet content sparked the demand for porting state-of-art natural language processing (NLP) applications to new social media as well as diverse language adaptation. Effort launched by the BOLT (Broad Operational Language Translation) program at DARPA (Defense Advanced Research Projects Agency) successfully addressed the internet information with enhanced NLP systems. BOLT aims for automated translation and linguistic analysis for informal genres of text and speech in online and in-person communication. As a part of this program, the Linguistic Data Consortium (LDC) developed valuable linguistic resources in support of the training and evaluation of such new technologies. This paper focuses on methodologies, infrastructure, and procedure for developing linguistic annotation at various language levels, including Treebank (TB), word alignment (WA), PropBank (PB), and co-reference (CoRef). Inspired by the OntoNotes approach with adaptations to the tasks to reflect the goals and scope of the BOLT project, this effort has introduced more annotation types of informal and free-style genres in English, Chinese and Egyptian Arabic. The corpus produced is by far the largest multi-lingual, multi-level and multi-genre annotation corpus of informal text and speech. Keywords: machine translation; parallel aligned Treebank; word alignment; PropBank; co-reference 1. Introduction Annotated corpora constitute a crucial resource for developing NLP systems and facilitating linguistic analysis of languages. With the rapid growth of the internet, NLP technologies are challenged with an avalanche of unstructured user-generated data. Off-the-shelf tools, trained on venerable formal data, perform worse when applied to new social media data. Various research and development efforts have been invested in this new domain -- massive informal and unstructured data. In this line, Ryan Cotterell and Chris Callison-Burch (2014) created a multi- dialect and multi-genre corpus of informal text via Amazon’s Mechanical Turk services. With a crowdsourcing approach, Jha et al. (2010) built a prepositional phrase attachment corpus of informal and noisy blog text. Owoputi et al. (2013) created part-of- speech tagged data for informal and online conversational Twitter text. The OntoNotes corpus (Weischedel et al. 2013) is a collaborative effort between BBN Technologies, Brandeis University, the University of Colorado, the University of Pennsylvania, and the University of Southern California's Information Sciences. The OntoNotes corpus comprises integrated annotation of multiple levels in various genres and in three languages (English, Chinese, and MSAArabic), providing structural information (syntax and predicate argument structure) and shallow semantics (word sense and coreference). Table 1: BOLT Annotation Data To leverage overall NLP technologies to a new horizon, DARPA initiated the BOLT program, encouraging development and adaptation of systems …

Content truncated to 500 words