Toolkit com.ibm.streams.speech2text 3.6.0
Specialized toolkits - release 4.3.1.0-i20200220 > com.ibm.streams.speech2text 3.6.0
General Information
This toolkit runs the rapid engine v4.8.1. The following rpms must be installed in order for this toolkit to work: atlas, atlas-devel, libsndfile and libsndfile-devel. Model files and model configuration files must be on the host that the Speech2Text operator is running from. The toolkit delivers standard narrow-band and broad-band en_US model files and configuration files in directory model:
- en_US.8kHz.pkg - narrow-band model file
- en_US.8kHz.basic.low_latency.pset - narrow-band basic configuration file (normalization estimated without buffering)
- en_US.8kHz.basic.attention.pset - narrow-band basic configuration file (buffered normalization, adds latency but improves accuracy)
- en_US.8kHz.diarization.low_latency.pset - narrow-band same as basic + enabling speaker labels
- en_US.8kHz.diarization.attention.pset - narrow-band same as basic + enabling speaker labels and (buffered normalization, adds latency but improves accuracy)
- en_US.16kHz.pkg - broad-band model file
- en_US.16kHz.basic.low_latency.pset - broad-band basic configuration file (normalization estimated without buffering)
- en_US.16kHz.diarization.low_latency.pset - broad-band same as basic + enabling speaker labels
For application patterns and helpful adapters, check out the open source streamsx.speech2Text repository: https://github.com/IBMStreams/streamsx.speech2text Only Watson model and config files built for this version of the toolkit are supported.
- Version
- 3.6.0
- Required Product Version
- 4.1.0.0