JANA: An Arabic human-human dialogues corpus

We present JANA, a multi-genre corpus of Arabic dialogues labeled for Arabic Dialogues Language Understanding (ADLU) at the utterance level. This paper describes progress in a development of the human-human dialogue corpus of Arabic spontaneous Spoken Dialogues (SD) and Instant Massages (IM). We collected dialogues from different genre call centers such as Banks, nights, and Mobile Network providers; these dialogues consist of transcribed phone calls and instant messages for inquiries regarding providing service from call centers. In addition, the annotation schema and manually turns segmentation are described. The collected data consist of approximately 3001 turns with average 6.7 words per turn, contains 4725 utterances with average 4.3 words per utterance, and 20311 words; and it will be made freely available to academic and nonprofit research.