Connectionist Temporal Classification(CTC) beam search decoding for C#(ctcdecode)

• 
• 
• 

ctcdecode is an implementation of CTC (Connectionist Temporal Classification) beam search decoding for C#. C++ code borrowed liberally from parlance ctcdecode parlance. It includes swappable scorer support enabling standard beam search, and KenLM-based decoding.
If you are new to the concepts of CTC and Beam Search, please visit the Resources section where we link a few tutorials explaining why they are needed.
Attention: Works only under Linux and Windows, because uses native libraries.

Installation

Building the C++ library requires gcc or clang. KenLM language modeling support is also optionally included, and enabled by default.

# get the code
git clone --recursive https://github.com/aleksandr-aleksashin/ctcdecode-csharp.git
cd ctcdecode-csharp

Compilation

To compile native library needs to switch into directory ./ctcdecode.

• For building Linux native library required g++. Run make clean and then make. This way works in WSL too.
• For building Windows native library required MinGW64. Run make clean and then make -f makefile-windows.

How to Use

using CTCBeamDecoder;
using CTCBeamDecoder.Models;

using var scorer = new DecoderScorer(labels);
var decoder = new CTCDecoder();
var result = decoder.Decode(output, scorer);

Inputs to CTCDecoder

• cutoffTopN Cutoff number in pruning. Only the top cutoffTopN characters with the highest probability in the vocab will be used in beam search. Default 40.
• cutoffProb Cutoff probability in pruning. 1.0 means no pruning. Default 1.
• beamWidth This controls how broad the beam search is. Higher values are more likely to find top beams, but they also will make your beam search exponentially slower. Furthermore, the longer your outputs, the more time large beams will take. This is an important parameter that represents a tradeoff you need to make based on your dataset and needs. Default 100.
• numProcesses Parallelize the batch using num_processes workers. You probably want to pass the number of cpus your computer has. Default 4.
• blankId This should be the index of the CTC blank token (probably 0). Default 0.
• logProbsInput If your outputs have passed through a softmax and represent probabilities, this should be false, if they passed through a LogSoftmax and represent negative log likelihood, you need to pass True. Default false.

Inputs to the Decode method

• output should be the output activations from your model. If your output has passed through a SoftMax layer, you shouldn't need to alter it (except maybe to transpose), but if your output represents negative log likelihoods (raw logits), you either need to pass it through an additional softmax or you can pass logProbsInput=False to the decoder. Shape: BATCHSIZE x N_TIMESTEPS x N_LABELS.
• scorer to be used in decoding.
• seqLens representing the sequence length of the items in the batch. Shape: BATCHSIZE.

Inputs to the OnlineDecode method

• output should be the output activations from your model. If your output has passed through a SoftMax layer, you shouldn't need to alter it (except maybe to transpose), but if your output represents negative log likelihoods (raw logits), you either need to pass it through an additional softmax or you can pass logProbsInput=False to the decoder. Shape: BATCHSIZE x N_TIMESTEPS x N_LABELS.
• states sequence of decoding states with lens equal to batch_size. Shape: BATCHSIZE.
• isEosS sequence of bool with lens equal to batch size. Should have False if havent pushed all chunks yet, and True if you pushed last cank and you want to get an answer. Shape: BATCHSIZE.
• seqLens representing the sequence length of the items in the batch. Shape: BATCHSIZE.

Inputs to DecoderScorer

• labels are the tokens you used to train your model. They should be in the same order as your outputs. For example if your tokens are the english letters and you used 0 as your blank token, then you would pass in list "_abcdefghijklmopqrstuvwxyz") as your argument to labels.
• modelPath is the path to your external kenlm language model(LM). Default null.
• alpha Weighting associated with the LMs probabilities. A weight of 0 means the LM has no effect. Default 0.
• beta Weight associated with the number of words within our beam. Default 0.

Inputs to DecoderState

• scorer associated scorer.

Resources

• Distill Guide to CTC
• Beam Search Video by Andrew Ng
• An Intuitive Explanation of Beam Search

License

Authored by: Aleksashin Aleksandr (aleksandr-aleksashin)
Developed in Dasha.AI Inc (Human-like conversational AI for developers)

This project is under MIT license. You can obtain the license copy here.