Server stt-standalone-worker Endpoints

• '/transcribe':
  • methods='POST'
  • inputs='wavFile'
  • output=audio transcription
• '/check':
  • methods='GET'
  • inputs=None
  • output=1 if the service is on
• '/stop':
  • methods='POST'
  • inputs=None
  • output=None

Config:

• Environment variables:
  • SERVICE_PORT
• Volumes:
  • /path/to/the/AM/directory:/opt/models/AM
  • /path/to/the/LM/directory:/opt/models/LM

Parameters

AM config file (decode.cfg)

• decoder: Acoustic model used for decoding (dnn2, dnn3)
• ampath: Path to the AM directory
• lmPath: Path to the language model generation directory (this option is only required by the model service)
• lmOrder: Ngrams used to generate the language model (this option is only required by the model service)
• min_active: Min active tokens
• max_active: Max active tokens
• beam: Decoding beam
• lattice_beam: Beam used in lattice generation
• acwt: Acoustic likelihoods scale
• frame_subsampling_factor : related to chain models (dnn3)

AM directory:

• final.mdl
• conf
  • ivector_extractor.conf
  • mfcc.conf
  • online_cmvn.conf
  • splice.conf
  • online.conf
• ivector_extractor
  • final.dubm
  • final.ie
  • final.mat
  • global_cmvn.stats
  • online_cmvn.conf
  • splice_opts

Nb: use relative path in config files example: decode.cfg >ampath: online/ >lmPath: online/LM_gen/

Nb: ivector_extractor.conf must include only the extraction parameters without the file dependecies > - --num-gselect=5 >- --min-post=0.025 >- --posterior-scale=0.1 >- --max-remembered-frames=1000 >- --max-count=100

LM directory:

• HCLG.fst
• words.txt