Language Model Action Set

This example is not available for the Lua programming language.

This section contains Python code. For more information about coding in Python, see Getting Started with SAS Viya for Python and SAS Viya: System Programming Guide.

The following code shows how you can use a step-by-step approach to end-to-end speech recognition in CAS:

 # step 1: import the SAS Scripting Wrapper for Analytics Transfer library
 from swat import *

 # step 2: start a CAS session
 s = CAS(host, port)

 # step 3: load CAS action sets
 s.loadactionset("audio")
 s.loadactionset("deepLearn")
 s.loadactionset("langModel")

 # step 4: load audio data for test set
 s.audio.loadaudio(path   = "test.listing",
                 casOut = {"name": "test_audio", "replace": True},
                 )

 # step 5: extract MFCC features from audio files for test set
 # Note: Values of the "frameExtractionOptions", "melBanksOptions", "mfccOptions"
 #       and "featureScalingMethod" parameters must be the SAME as those used to
 #       train the RNN acoustic model.
 s.audio.computefeatures(
     table                  = "test_audio",
     casOut                 = {"name": "test_dlscore_input", "replace": True},
     audioColumn            = "_audio_",
     copyvars               = ["_path_"],
     frameExtractionOptions = {"frameShift": 10, "frameLength": 25, "dither": 0.0},
     melBanksOptions        = {"nBins": 40},
     mfccOptions            = {"nCeps": 40},
     featureScalingMethod   = "STANDARDIZATION",
     nOutputFrames          = 1500,
 )

 # step 6: import the pretrained RNN acoustic model if it's not in-memory
 s.table.loadtable(path = "acoustic_model.sashdat",
                 casOut = {"name": "asr", "replace": True},
                 )
 s.table.loadtable(path = "acoustic_model_weights.sashdat",
                 casOut = {"name": "pretrained_weights", "replace": True},
                 )
 s.table.loadtable(path = "acoustic_model_weights_attr.sashdat",
                 casOut = {"name": "pretrained_weights_attr", "replace": True},
                 )
 s.table.attribute(task = "ADD",
                 name = "pretrained_weights",
                 attrtable = "pretrained_weights_attr"
                 )

 # step 7: use pretrained RNN model to score acoustic features of test set
 s.dlScore(table          = "test_dlscore_input",
         model          = "asr",
         initweights    = "pretrained_weights",
         layerImageType = 'WIDE',
         copyvars       = ['_path_'],
         casOut         = {"name": "test_rnn_scores", "replace": True},
         )

 # step 8: import an n-gram language model
 s.langModel.lmImport(table  = "language_model.sashdat",
                     casOut = "language_model"
                     )

 # step 9: decode RNN scores
 s.langModel.lmDecode(table          = "test_rnn_scores",
                     langModelTable = "language_model",
                     blankLabel     = " ",
                     spaceLabel     = "&",
                     copyvars       = ['_path_'],
                     casOut         = "test_results"
                     )

 # step 10: save the results
 s.save(table = "test_results",
     name  = "test_results.csv",
     replace = True
     )

This example is not available for the R programming language.