Above 95% accuracy on broadcast news content
On-the-fly language models adaptation
Segmentation of acoustic flow into coherent segments (music, microphone vstelephony speech, music + speech)
Speaker diarization(segmentation, tracking, gender detection)
Speaker identification (requires specific model training for the related speakers)
Language detection
Complete transcription of the speech parts into text
Confidence scores at the word and sentence level
Punctuation
