Speech and Natural Language Processing

A curated list of speech and natural language processing resources. Other lists can be found in this list. If you want to contribute to this list (please do), send me a pull request. All Sub-caterogires are listed in alphabetical order

Finite State Toolkits and Regular Expressions

• AT&T FSM Library The AT&T FSM libraryTM is a set of general-purpose software tools available for Unix, for building, combining, optimizing, and searching weighted finite-state acceptors and transducers.
• Carmel Finite-state toolkit, EM and Bayesian (Gibbs sampling) training for FST and context-free derivation forests/
• Categorial semiring Categorial semiring as described in Sproat et al. 2014
• dk.brics.automaton Java toolkit for FSAs and regular expression.
• Fare Fare is a finite state and regular expression libary for the .NET framework written in C#.
  am is a JavaScript library for working with automata and formal grammars for
  regular and context-free languages
• Foma Finite-state compiler and C library
• fsa Toolkit used in RWTH ASR engine
• fsm2.0 Thomas Hanneforths fsm 2.0 library written C++ has a few nice operations
  such as three-way composition
• fstrain A toolkit for training finite-state models
• jopenfst Java port of the C++ OpenFst library; originally forked from the CMU Sphinx project
• Kleene programming language High level finite state programming language built on top of OpenFst.
• MIT FST Toolkit WFST toolkit no
  maintained anymore but feature a few commands not found in other toolkits
• MoMs-for-StochasticLanguages Spectral and other training algorithms for WFSAs.
• n Shortest Path for PDT n Shortest Path for PDT
• Noam
  "Noam is a JavaScript library for working with automata and formal grammars for regular and context-free
  languages". Also has pretty cool examples using viz.js
• OpenFst OpenFst is a library for constructing, combining, optimizing, and searching weighted finite-state transducers (FSTs).
• openfst-utils Nice set of
  utilities for OpenFst includes implementation of Categorial
  semirings.openfst-utils.
• openlat Toolkit for manipulating word
  lattice built on top of OpenFst. Includes support for reading and writing HTK
  compatible lattices.
• PyFst Python interface to OpenFst
• SFST - Stuttgart Finite State Transducer Tools
  "SFST is a toolbox for the implementation of morphological analysers and other
  tools which are based on finite state transducer technology."
• Treba "Treba is a basic command-line tool for training, decoding, and calculating with weighted (probabilistic) finite state automata (PFSA) and Hidden Markov Models (HMMs)."

Many of the toools in the machine translation section also implement interesting graph and semiring operations.

Language Modelling Toolkits

• Bayesian Recurrent Neural Network for Language Modeling This is a C/C++ implementation for Bayesian recurrent neural network for language modeling (BRNNLM)
• Berkeley LM
• Bigfatlm
  Provides Hadoop training of Kneser-ney language models, written in Java.
• CSLM "Continuous Space Language
  Model toolkit. CSLM toolkit is open-source software which implements the so-called continuous
  space language model.
• DALM Double array language model.
• KenLM
  Kenneth Heafield's language model toolkit, uses a very fast and low memory
  representation.
• lwlm lwlm is an exact, full Bayesian implementation of the Latent Words Language Model (Deschacht and Moens, 2009).
• Maximum Entropy Modeling
  Le Zhang has a comprehensive set of links related MaxEnt models.
• Maximum entropy language models: SRILM extension
  "This patch adds the functionality to train and apply maximum entropy (MaxEnt)
  language models to the SRILM toolkit. Currently, only N-gram features are
  supported"
• mitlm
  My personal favourite LM toolkit, super fast and seems to get slightly higher
  accuracy.
• MSRLM
  "This scalable language-model tool is used to build language models from large
  amounts of data. It supports modified absolute discounting and Kneser-Ney
  smoothing."
• OpenGrm
  Language modelling toolkit for use with OpenFst.
• cpyp C++ library for modeling with Pitman-Yor processes
• RandLM Bloom filter based random language models
• RNNLM
  Recurrent neural network language model toolkit.
• Refr
  Re-ranking framework from the Johns-Hopkins workshop on confusion language
  modelling.
• rwthlm A toolkit for training neural network language models (feedforward, recurrent, and long short-term memory neural networks). The software was written by Martin Sundermeyer.
• SRILM Very popular toolkit,
  source code avaliable but only non-free for commerical use.

Speech Recognition

• AaltoASR Aalto Automatic Speech Recognition tools
• Barista Barista is an open-source framework for concurrent speech processing.
• Bavieca New open source toolkit
  featuring static and dynamic decoders.
• kaldi-nnet-dur-model Neural network phone duration model on top of the Kaldi speech recognition framework, (Interspeech paper)
• CMU Sphinx Open Source Toolkit For Speech
  Recognition Project by Carnegie Mellon University
• HTK "The Hidden Markov Model Toolkit (HTK) is a
  portable toolkit for building and manipulating hidden Markov models."
• Juicer Juicer is a Weighted Finite State Transducer (WFST) based decoder for Automatic Speech Recognition (ASR).
• Julius "Julius is a high-performance, two-pass large vocabulary continuous speech recognition (LVCSR) decoder software for speech-related researchers and developers."
• Kaldi Modern open source toolkit lead by
  Dan Povey featuring many state-of-the-art techniques.
• OpenDcd An Open Source WFST based Speech Recognition Decoder.
• Phonetisaurus
  Josef Novak's super fast WFST based Phoneticizer, site also
  has some really nice tutorials slides.
• Sail Align SailAlign is an open-source software toolkit for robust long speech-text alignment implementing an adaptive, iterative speech recognition and text alignment scheme that allows for the processing of very long (and possibly noisy) audio and is robust to transcription errors. It is mainly written as a perl library but its functionality also depends…
• SCARF: A Segmental CRF Toolkit for Speech Recognition
  "SCARF is a toolkit for doing speech recognition with segmental conditional
  random fields."
• trainc
  David Rybach and Michael Riley's tool for direct construction of
  context-dependency transducers (Interspeech best paper).
• RASR RWTH ASR - The
  RWTH Aachen University Speech Recognition System

Signal Processing

• An Interactive Source Separation Editor "ISSE is an open-source, freely available, cross-platform audio editing tool that allows a user to perform source separation by painting on time-frequency visualisations of sound."
• Bob Bob is a free signal-processing and machine learning toolbox originally developed by the Biometrics group at Idiap Research Institute, in Switzerland.
• Matlab Audio Processing Examples
• SAcC - Subband Autocorrelation Classification Pitch Tracker "SAcC is a (compiled) Matlab script that performs noise- robust pitch tracking by classifying the autocorrelations of a set of subbands using an MLP neural network."

Text-to-Speech

• HTS HMM-based speech synthesis
• RusPhonetizer Grammar rules and dictionaries for the phonetic transcription of Russian sentences

Speech Data

• cmudict CMUdict (the Carnegie Mellon Pronouncing Dictionary) is a free pronouncing dictionary of English.
• LibriSpeech ASR corpus LibriSpeech is a corpus of approximately 1000 hours of 16kHz read English speech, prepared by Vassil Panayotov with the assistance of Daniel Povey. The data is derived from read audiobooks from the LibriVox project, and has been carefully segmented and aligned.
• TED-LIUM Corpus The TED-LIUM corpus was made from audio talks and their transcriptions available on the TED website.

Machine Translation

• Berkeley Aligner
  "...a word alignment software package that implements recent innovations in
  unsupervised word alignment."
• cdec
  "Decoder, aligner, and model optimizer for statistical machine translation and
  other structured prediction models based on (mostly) context-free formalisms"
• Jane
  "Jane is RWTH's open source statistical machine translation toolkit. Jane
  supports state-of-the-art techniques for phrase-based and hierarchical
  phrase-based machine translation."
• Joshua
  Hierarchical and syntax based machine translation decoder written in Java.
• Moses
  Standard open source machine translation toolkit.
• alignment-with-openfst
• zmert
  Nice Java Mert implementation by Omar F. Zaidan

Machine Learning

• BIDData BIDMat is a matrix library intended to support large-scale exploratory data analysis. Its sister library BIDMach implements the machine learning layer.
• libFM: Factorization Machine Library
• sofia-ml Fast incremental learning
  algorithms for classification, regression, ranking from Google.
• Spearmint
  Spearmint is a package to perform Bayesian optimization according to the
  algorithms outlined in the paper: Practical Bayesian Optimization of Machine
  Learning Algorithms Jasper Snoek, Hugo Larochelle and Ryan P. Adams Advances
  in Neural Information Processing Systems, 2012

Deep Learning

• Benchmarks - Comparison of different convolution network implementations.
• Cafee - Really active deep learning toolkit with support for cuDNN and lots of other backends.
• cuDNN - Deep neural network from Nvidia with paper here. Torch 7 has support for cuDnn and here are some Python wrappers.
• CURRENNT - Munich Open-Source CUDA RecurREnt Neural Network Toolkit described in this paper
• gensim - Python topic modeling toolkit with word2vec implementation. Extremly easy to use and to install.
• Glove Global vectors for word representation.
• GroundHog Neural network based machine translation toolkit.
• KALDI LSTM C++ implementation of LSTM (Long Short Term Memory), in Kaldi's nnet1 framework. Used for automatic speech recognition, possibly language modeling etc.
• OxLM: Oxford Neural Language Modelling Toolkit Neural network toolkit for machine translation described in the paper here
• Neural Probabilistic Language Model Toolkit "NPLM is a toolkit for training and using feedforward neural language models (Bengio, 2003). It is fast even for large vocabularies (100k or more): a model can be trained on a billion words of data in about a week, and can be queried in about 40 μs, which is usable inside a decoder for machine translation."
• RNNLM2WFST Tool to convert RNNLMs to WFSTs
• ViennaCL <http://viennacl.sourceforge.net/> - ViennaCL is a free open-source linear algebra library for computations on many-core architectures (GPUs, MIC) and multi-core CPUs.

Natural Language Processing

• BLLIP reranking parser "BLLIP Parser is a statistical natural language parser including a generative constituent parser (first-stage) and discriminative maximum entropy reranker (second-stage)."
• OpenNLP The Apache OpenNLP library is a machine learning based toolkit for the processing of natural language text.
• SEAL Set expander for any language described in this paper
• Stanford CoreNLP "Stanford CoreNLP provides a set of natural language analysis tools written in Java"

Applications

• Cloud ASR using PyKaldi "CloudASR is a software platform and a public ASR webservice."

Other Tools

• GraphViz.sty
  Really handy tool adding dot languge directly to a LaTex document, useful for
  tweaking the small colorized WFST figure in papers and presentations.

Blogs

• Between One and Zero by William Hartmann
• cmusphinx CMU Sphinx related blog
• Language Log
• LingPipe Blog Natural Language Processing and Text Analytics
• Natural Language Processing Blog by Hal Daumé III
• Spoken Language Processing "Some thoughts on Spoken Language Processing, with tangents on Natural Language Processing, Machine Learning, and Signal Processing thrown in for good measure."

Books

• DEEP LEARNING: Methods and Applications By Li Deng and Dong Yu
• Foundations of Data Science Draft by John Hopcroft and Ravindran Kannan
• Introduction to Matrix Methods and Applications (Working Title) S. Boyd and L. Vandenberghe