IEEE Proceedings of 4 th International Conference on Intelligent Human Computer Interaction, Kharagpur, India, December 27-29, 2012 978-1-4673-4369-5/12/$31.00 ©2012 IEEE A Comparative Study of Native (L1) and Nonnative (L2) English Speech Strategy for Discourse Prosodic Organization Shambhu Nath Saha Center for Educational Technology Indian Institute of Technology Kharagpur, India shambhuju@gmail.com Shyamal Kr. Das Mandal Center for Educational Technology Indian Institute of Technology Kharagpur, India sdasmandal@cet.iitkgp.ernet.in Abstract— English speech of L1 English and L1 Bengali speakers of the same discourse is analyzed by applying hierarchical discourse prosody framework HPG (Hierarchy of Prosodic Phrase Group) in order to find the between-and within- group similarities and differences in speech planning of L1 (native) and L2 (non-native) English speakers for prosodic organization at discourse level. The analysis reveals that the speech rate of L1 speakers is higher than speech rate of L2 speakers; L2 speakers contain more break boundary than that of the L1 speakers at every discourse prosodic level in the organization, which exhibit the fact that L2 speakers use more intermediate chunking units and larger scale planning units than that of L1 speakers. Between-group differences are also found through the analysis of phrase component at prosodic phrase level (PPh) and accent component at prosodic word level (PW). These findings can be attributed to L2 speakers’ improper phrasing, improper word level prominence and ambiguous difference between content words and function words in discourse prosodic organization. Index Terms— Discourse level, Prosodic organization, Segmental features, Suprasegmental features, Language Transfer, Speech prosody, Prosodic word, Prosodic phrase. I. INTRODUCTION English is an international language for communication and its importance grows continuously day by day throughout the world. In large number of countries, English is studied and spoken as second language, so understanding the range of variation present in the English spoken in the world today is fundamental issue for the development of English language education as well as spoken language science and technology. Combining native and nonnative speakers, in India more people who speak or understand English than any other country in the world. Thus research in Indian English dialects from a multidisciplinary perspective is urgently needed to address issues in communication, learning and technology. Research in linguistics can investigate and analyze the range of variation present in Indian English dialects; research in speech science can implement linguistics findings into the development of language pedagogy as well as into ICT tools and environments tailored to the requirements of Indian speaker populations. The Asian English Speech cOrpus Project (AESOP) [1], a multi-national collaboration of linguists, psychologists, speech scientists, technologists and educators was already started to collect and compare L2 English speech corpora from as many Asian countries as possible in order to derive a set of core properties common to all varieties of Asian English, as well as to discover features that are particular to individual dialects. The major research goal of AESOP is not to normalize Asian English to any particular ENL standard, but instead to catalog and predict similarities and differences among the verities of English found across Asia. Since India is a multilingual country we have so many variety of Indian English based on the local language and dialect. So there is a need of huge research in speech science for implementing the linguistic finding into the development of language pedagogy for second language acquisition. A learner of nonnative language (L2) tends to retain the prosody of his/her native language (L1) and transfer it into L2 when speaking. This cross-linguistic influence is referred to as language transfer. Language transfer occurs at both the segmental and suprasegmental levels [2]. Segmental features are important for correct word pronunciation and suprasegmental features convey important information about linguistics and information structures. Negative transfer in suprasegmental features may significantly impede intelligibility and comprehensibility of L2 speech and hamper communication. Current research disproves the idea that second language speech becomes less comprehensible because of different from native pronunciation. Stress is one of the important suprasegmental features for second language acquisition. Lexical stress deals with the part-of-speech of a word. Also, stress changes may occur for different inflectional forms of a given word. On the other hand utterance-level stress can indicate the focus, which helps to represent the information structure of a discourse by distinguishing between given and new information, or back ground and foreground information. English contains stressed and unstressed syllables and is referred to as stress-timed language with quasi-uniform durations between consecutive stressed syllables [3]. English pitch contour also has specific linguistic and paralinguistic importance. On the other hand Bengali is syllable-timed language where all syllables, whether stressed or unstressed, tend to occur at regular time interval and have quasi-uniform