Improving Dialogue Act Classification for Spontaneous Arabic Speech and Instant Messages at Utterance Level

Improving Dialogue Act Classification for Spontaneous Arabic Speech and Instant Messages at Utterance Level AbdelRahim A. Elmadany1, Sherif M. Abdou2, Mervat Gheith3 1Department of Computer Science, Deanship of Community Service and Continuing Education, Jazan University, KSA 2Department of Information Technology, Faculty of Computers and Information, Cairo University, Egypt 3Department of Computer Science, Institute of Statistical Studies and Research (ISSR), Cairo University, Egypt aelmadany@jazanu.edu.sa, sh.ma.abdou@gmail.com, mervat_gheith@yahoo.com Abstract The ability to model and automatically detect dialogue act is an important step toward understanding spontaneous speech and Instant Messages. However, it has been difficult to infer a dialogue act from a surface utterance because it highly depends on the context of the utterance and speaker linguistic knowledge; especially in Arabic dialects. This paper proposes a statistical dialogue analysis model to recognize utterance’s dialogue acts using a multi-classes hierarchical structure. The model can automatically acquire probabilistic discourse knowledge from a dialogue corpus were collected and annotated manually from multi-genre Egyptian call-centers. Extensive experiments were conducted using Support Vector Machines classifier to evaluate the system performance. The results attained in the term of average F-measure scores of 0.912; showed that the proposed approach has moderately improved F-measure by approximately 20%. Keywords: Dialogues Language Understanding, Dialogue Acts Classification, Spoken Dialogues, Instant Messages, Natural Language Understanding 1. Introduction The most important and difficult part in human-computer interaction system “i.e. Dialogue System” is understanding what the user needs? This task is called "language understating component" or somewhere "Dialogue Acts (DAs) classification." DAs classification task is labeling the speaker’s intention in producing a particular utterance with short words; the DAs terminology approximately is the equivalent of the speech act of Searle (1969), and DAs is different based on dialogue systems domains (Elmadany et al., 2015b). Since 1999, the research in DAs area has increased after spoken dialogue systems been a commercial reality. Hence, the development of dialogue systems has focused on some of the conversational roles such acts which can perform because it is closely linked to the field of computational linguistics (Stolcke et al., 2000). DAs is used practically in many live dialogue systems such as Airline Travel Information Systems such as ATIS (Seneff et al., 1991), DARPA (Pellom et al., 2001), and VERBMOBIL (Wahlster, 2000). Recently, the development of dialogue interaction systems has gained considerable attention, but most of the resources and systems are built so far tailored to English and other Indo-European languages. The development of the dialogue systems for other languages as Arabic is required. So, the Arabic dialogue acts classification’s task has gained focus because it is a key player in Arabic language understanding to building these systems. The motivation of this paper comes from the point of view “building automatic language understanding component for Egyptian dialect dialogues”. The paper focuses on inquiry–answer dialogues from the call-centers domain because it receives or transmits a large volume of information inquiries from customers. In this research, we have selected Customer-Service entities from Banks, Flights, and Mobile Networks Operators call- centers. In this paper, we are referring …

Content truncated to 500 words