Embedded Files
This site lists the training data is permissible for the training of MT systems and language models for ASR.
This site lists the training data is permissible for the training of MT systems and language models for ASR.
Provided Data
Provided Data
ASR AM Training Data (English, German): sign this agreement, scan it and send to gretter@fbk.eu.
Other Permissible Data
Other Permissible Data
Parallel:
parallel corpora from the Wikipedia for en-{cs,de,fr,vi} pairs, provided by Krzsyztof Wołk of Polish-Japanese Academy of Information Technology
Monolingual:
TED
any subtitles of a TED or TEDx talk that is not listed as non-permissible.
LDC
LDC2011T11 Arabic Gigaword Fifth Edition
Miscellaneous:
Cantab Research baseline LM and text corpus kindly provided by Tony Robinson of Cantab Research (now considered compatible with IWSLT 2015)
Page updated
Google Sites
Report abuse