Computational Linguistics is the journal of The Association for Computational Linguistics published by The MIT Press.

Computational Linguistics
Special Issue on Semantic Role Labeling

The general problem of interpreting text involves the determination of the semantic relations among the entities and the events they participate in. Given a sentence, one formulation of the task consists of detecting basic event structures such as "who" did "what" to "whom", "when" and "where". From a linguistic point of view, a key component of the task corresponds to identifying the semantic arguments filling the roles of the sentence predicates. These predicates are mainly lexicalized by verbs but also by some verb nominalizations and adjectives. Typical predicate semantic arguments include Agent, Patient, and Instrument; semantic roles may also be found as adjuncts (e.g., Locative, Temporal, Manner, and Cause). The related tasks of determining the semantic relations among nouns and their modifiers, as well as prepositions and their arguments, are clearly important for text interpretation as well, and indeed often draw on similar role labels.

As with many areas in computational linguistcs (CL) and Natural Language Processing (NLP), work has proceeded for decades on manually created semantic grammars and other resources for supporting text interpretation (e.g., [Hirst 1987], [Pustejovsky 1995], [Copestake and Flickinger 2000]). This body of research has supported deep semantic analysis of language input, but has the drawbacks typical of such approaches in requiring intensive manual labor, often restricted to narrow domains. The growth of statistical machine learning methods, addressing these issues of the knowledge acquisition bottleneck, were for many years limited in this area to related problems of learning subcategorization frames [Briscoe and Carroll 1997] or classifying verbs according to argument structure properties [Merlo and Stevenson 2001] [Schulte im Walde 2000], due to the lack of appropriate resources to support such methods in labeling semantic roles of arguments.

Recently, however, the compilation and manual annotation with semantic roles of medium-large corpora – the PropBank, NomBank, and FrameNet initiatives – has enabled the development of statistical approaches specifically for the task of semantic role labeling (SRL). SRL, especially focused on the labeling of verbal arguments and adjuncts, has become a well-defined task with a substantial body of work and comparative evaluation (e.g., see [Gildea and Jurafsky 2002], [Surdeanu et al. 2003], [Xue and Palmer 2004], [Pradhan et al. 2005], CoNLL Shared Task in 2004 and 2005, Senseval-3). The identification of such event frames holds potential for significant impact in many NLP applications, as suggested by the following works on Information Extraction [Surdeanu et al. 2003], Question Answering [Narayanan and Harabagiu 2004], Summarization [Melli et al. 2005], and Machine Translation [Boas 2002]; as well, work on noun modifier relations has been encouraging for related NLP tasks (e.g., [Moldovan and Badulescu 2005], [Rosario and Hearst 2004]). Although the use of SRL systems in real-world applications has so far been limited, the outlook is promising over the next several years for a spread of this type of analysis to a range of applications requiring some level of semantic interpretation. Moreover, the problem represents an excellent framework to perform research on CL and NLP techniques for acquiring and exploiting semantic relations among the different components of the structured output to be constructed.

Topics

The call for papers of this special issue invites submissions of articles describing novel and challenging work and results in Semantic Role Labeling (SRL) and its applications, with emphasis on the evaluation of qualitative and quantitative aspects that give a deep insight on the SRL task and, in general, on the syntactico-semantic processing of natural language. The range of topics to be covered includes, but is not limited to:

• Novel statistical and machine learning approaches and architectures for SRL
• Study of the relevant information/knowledge for the task
• Learning from small training sets
• Unsupervised models for SRL
• Scalability of the state-of-the-art systems
• How to make systems robust against annotation errors
• Inclusion of deep semantic information and relations
• Generalization to new corpora and to new unseen frames
• Knowledge-based approaches to SRL and comparison to the statistical approach
• Combination of systems and approaches, specially addressing the integration of knowledge-based and statistical views
• Study of the relation between the syntactic and semantic layers for SRL characterization and system development
• Applications of SRL (e.g., in domains such as Q&A, MT, Summarization, etc.)
• Evaluation: new metrics for direct evaluation and indirect evaluations through applications
• Development of corpora and resources for the task
• SRL for languages other than English

Accepted Articles

The following articles have been selected for publication in the Special Issue:

• Kristina Toutanova, Aria Haghighi, Christopher D. Manning, A Global Joint Model for Semantic Role Labeling
• Alessandro Moschitti, Daniele Pighin, Roberto Basili, Tree Kernels for Semantic Role Labelling
• Nianwen Xue, Labeling Chinese Predicates with Semantic Roles
• Vasin Punyakanok, Dan Roth, Wen-tau Yih, The Importance of Syntactic Parsing and Inference in Semantic Role Labeling
• Sameer S. Pradhan, Wayne Ward, James H. Martin, Towards Robust Semantic Role Labeling

Important Dates

Call for papers: 15 March 2006
Submission of articles: 15 July 2006
Preliminary decisions to authors: 23 February 2007
Submission of revised articles: 30 April 2007
Final decisions to authors: 15 June 2007
Final versions due from authors: 31 July 2007
Publication: end of 2007

Guest Editors

Lluís Màrquez, Technical University of Catalonia
Kenneth C. Litkowski,  CL Research
Suzanne Stevenson,  University of Toronto
Xavier Carreras, Massachusetts Institute of Technology

Guest Editorial Board

Tim Baldwin, University of Melbourne
Katrin Erk, Saarland University
Roxana Girju, University of Illinois at Urbana-Champaign
Kadri Hacioglu, University of Colorado
Sanda Harabagiu, University of Texas at Dallas
Valia Kordoni, Saarland University
Namhee Kwon, ISI - University of Southern California
Mirella Lapata, University of Edinburgh
Christopher Manning, Stanford University
Paola Merlo, University of Geneva
Rada Mihalcea, University of North Texas
Dan Moldovan, University of Texas at Dallas
Alessandro Moschitti, University of Rome Tor Vergata
Hwee Tou Ng, National Universtity of Singapore
Xue Nianwen, University of Colorado
Nicolas Nicolov, Umbria Inc.
Sameer Pradhan, BBN Technologies
James Pustejovsky, Brandeis University
Maarten de Rijke, University of Amsterdam
Dan Roth, University of Illinois at Urbana-Champaign
Sabine Schulte im Walde, Saarland University
Mihai Surdeanu, Technical University of Catalonia
Cynthia Thompson, PricewaterhouseCoopers
Kristina Toutanova, Microsoft Research
Scott Wen-tau Yih, Microsoft Research