Скачать или смотреть FarSpeech: Arabic Natural Language Processing for Live Arabic Speech.

FarSpeech, QCRI’s combined Arabic
speech recognition, natural language processing (NLP), and di-
alect identification pipeline. It features modern web technolo-
gies to capture live audio, transcribes Arabic audio, NLP pro-
cesses the transcripts, and identifies the dialect of the speaker.
For transcription, we use QATS, which is a Kaldi-based ASR
system that uses Time Delay Neural Networks (TDNN). For
NLP, we use a SOTA Arabic NLP toolkit that employs var-
ious deep neural network and SVM based models. Finally,
our dialect identification system uses multi-modality from both
acoustic and linguistic input. FarSpeech 1 presents different
screens to display the transcripts, text segmentation, part-of-
speech tags, recognized named entities, diacritized text, and the
identified dialect of the speech.

try it here: https://farspeech.qcri.org/