A Proposition Bank of Urdu

A Proposition Bank of Urdu Maaz Anwar Nomani♣, Riyaz Ahmad Bhat♣, Dipti Misra Sharma♣, Ashwini Vaidya♥, Martha Palmer♠, and Tafseer Ahmed♦ FC Kohli Center on Intelligent Systems (KCIS), IIIT-H, Hyderabad, India♣ Indian Institute of Technology, New Delhi, India♥ University of Colorado, Boulder, CO 80309 USA♠ DHA Suffa University, Karachi, Pakistan♦ {maaz.anwar, riyaz.bhat}@research.iiit.ac.in, dipti@iiit.ac.in, ashwini.vaidya@gmail.com, martha.palmer@colorado.edu tafseer@dsu.edu.pk Abstract This paper describes our efforts for the development of a Proposition Bank for Urdu, an Indo-Aryan language. Our primary goal is the labeling of syntactic nodes in the existing Urdu dependency Treebank with specific argument labels. In essence, it involves annotation of predicate argument structures of both simple and complex predicates in the Treebank corpus. In this paper, we describe the overall process of building the PropBank of Urdu. We discuss various statistics pertaining to the Urdu PropBank and the issues which the annotators encountered while developing the PropBank. We also discuss how these challenges were addressed to successfully expand the PropBank corpus. While reporting the Inter-annotator agreement between the two annotators, we show that the annotators share similar understanding of the annotation guidelines and of the linguistic phenomena present in the language. Keywords: PropBank, Semantic Role Labelling, Treebanking 1. Introduction The last decade has seen the development of several corpora with semantic role annotation in addition to syntactic anno- tation because they offer rich data for empirical research, building language understanding systems and other natural language processing (NLP) applications. Proposition Bank (from now on, PropBank) (Kingsbury and Palmer, 2003) is a corpus in which the arguments of each verb predicate (simple or complex) are marked with their semantic roles. PropBank for English establishes a layer of semantic representation for Penn Treebank which is already annotated with Phrase structure trees. Captur- ing the semantics through predicate-argument structure in- volves quite a few challenges peculiar to each predicate type because the syntactic notions in which the verb’s ar- guments and adjuncts are realized can vary based on the senses. Our report on PropBanking an Indian language, Urdu, spo- ken in major parts of India and Pakistan, describes the an- notation process of labeling the predicate-argument struc- ture on top of an Urdu Dependency Treebank. The need for such a resource arises from the fact that while the Urdu Treebank does provide syntactico-semantic labels for some adjuncts like location and time, it fails to categorize the ar- guments which a verb can take depending upon subject and object pertaining to the verb. Adding a semantic layer to the Treebank will help in resolving and addressing this re- quirement. Urdu is spoken primarily in Northern and Southern India. The language’s Indo-Aryan word base has been enriched by borrowing from Persian and Arabic. Urdu is written right- to-left in an extension of the Persian alphabet, which is it- self an extension of Arabic writing. Urdu is associated with the Nastaliq style of Persian calligraphy, whereas Arabic is Ashwini Vaidya was at University of Colorado, Boulder, USA when the work on Urdu PropBank was done. generally written in …

Content truncated to 500 words