automatic speech recognition (ASR)

This project provides a client package and example scripts to access the alphaspeech pro ASR APIs.

WSJ: Wall Street Journal corpus from ARPA in 1992, 1994 (LDC93S6A, LDC94S13A).

Vystadial: English part of Vystadial CTS corpus from Prague [Korvas et al. 2014].

VoxForge: free, open-source ASR dataset of crowdsourced speech from voxforge.org.

VCTK: Voice Cloning Toolkit dataset from CSTR, Edinburgh [Veaux et al. 2013].

TIMIT: famous corpus of American English with phone-level transcriptions (LDC93S1).

TED-LIUM 3: Release 3 of TED talk corpus from LIUM (SLR51).

TED-LIUM 2: Release 2 of TED talk corpus from LIUM (SLR19).

TED-LIUM 1: Release 1 of TED talk corpus from LIUM (SLR7).

Tatoeba: Tatoeba Project of English sentences (https://tatoeba.org/eng).

synthcmd: Synthetic Speech Commands Dataset from Kaggle.

SWC: Spoken Wikipedia Corpus, crowdsourced speech of read Wikipedia articles.

Switchboard-1, release 2: famous ASR data set of CTS from the 1990s (LDC97S62).

ST-AEDS: Surfingtech American English Dataset of cellphone speech (SLR45).

speechcmd: Speech Commands Dataset from Google [Warden 2018].

Snips: SLU dataset from Snips (now part of Sonos) [Saade et al. 2019].

rt03: NIST 2003 Rich Transcription Evaluation Data for CTS (LDC2007S10).

RM: Resource Management v. 2.0 corpus from DARPA in the 1990s (LDC93S3A).

RedDots: corpus of short-dur utts from mobile apps (sites.google.com/site/thereddotsproject).

PDA: Personal Digital Assistant speech dataset from CMU.