ahmadmardeni1
commented
3 years ago Good afternoon and thank you for submitting your topic suggestion. Your topic form has been entered into our queue and should be reviewed (for approval) as soon as a content moderator is finished reviewing the ones in the queue before it.
lalith1403
Topic Suggestion
End-to-end speech recognition with deep Recurrent Neural Networks(RNNs)
Pre-submission advice
By following all our pre-submission advice and reviewing our Resources folder, you will maximise your chances of your topic being approved.
We ask that you please be patient as our team works through approving and publishing all articles/tutorials in a timely manner.
Allow 1-3 days for a topic to be reviewed and/or approved - allow 3-7 days for an articles to be reviewed and/or published.
Be sure to visit our Resources Page for tools, resources, and example articles that will help you propose and write a successful article.
Please ensure that you have only one open issue + linked pull request at a time. This will ensure that we complete the article in a timely manner from inception to publishing.)
We tend to stray away or tend not to publish reviews/comparisons of commercial product offerings.
Proposal Submission
[Machine Learning]End-to-end speech recognition with deep Recurrent Neural Networks(RNNs)
Proposed article introduction
Neural networks have a long history in speech recognition, usually in combination with hidden Markov models . They have gained attention in recent years with the dramatic improvements in acoustic modelling yielded by deep feedforward networks . Given that speech is an inherently dynamic process, it seems natural to consider recurrent neural networks (RNNs) as an alternative model. HMM-RNN systems have also seen a recent revival , but do not currently perform as well as deep networks.
Key takeaways
Article quality
This article will present a speech recognition system that directly transcribes audio data with text, without requiring an intermediate phonetic representation. The system is based on a combination of the deep bidirectional LSTM recurrent neural network architecture and the Connectionist Temporal Classification objective function. My article will also cover in-depth Deep learning frameworks that would be alternative for RNN.
References
N/A
Conclusion
The reader would be shown that the combination of deep, bidirectional Long Short-term Memory RNNs with end-to-end training and weight noise gives state-of-the-art results in phoneme recognition on the TIMIT database. An obvious next step is to extend the system to large vocabulary speech recognition. Another interesting direction would be to combine frequency-domain convolutional neural networks
Templates to use as guides