TheanoLM - An Extensible Toolkit for Neural Network Language Modeling
This toolkit addresses the need for efficient and extensible language modeling tools for researchers in speech recognition, though it is incremental as it builds on existing Theano-based methods.
The authors introduced TheanoLM, a flexible toolkit for neural network language modeling that achieved significant improvements over back-off n-gram models in Finnish and English conversational speech recognition tasks, with training times an order of magnitude shorter than existing toolkits.
We present a new tool for training neural network language models (NNLMs), scoring sentences, and generating text. The tool has been written using Python library Theano, which allows researcher to easily extend it and tune any aspect of the training process. Regardless of the flexibility, Theano is able to generate extremely fast native code that can utilize a GPU or multiple CPU cores in order to parallelize the heavy numerical computations. The tool has been evaluated in difficult Finnish and English conversational speech recognition tasks, and significant improvement was obtained over our best back-off n-gram models. The results that we obtained in the Finnish task were compared to those from existing RNNLM and RWTHLM toolkits, and found to be as good or better, while training times were an order of magnitude shorter.