OpenFst: A General and Efficient Weighted Finite-State Transducer Library

By Cyril Allauzen, Michael D. Riley, Johan Schalkwyk, Wojciech Skut, Mehryar Mohri. International Conference on Implementation and Application of Automata, 2007

672 citations recorded by the source corpus.

Subfields: Artificial Intelligence

View Paper: https://www.semanticscholar.org/paper/31e105cac80aa8f1646dfb22c95d035564ea4998

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Yoshua Bengio
2. Dong Yu
3. Jinyu Li
4. Quoc V. Le
5. Kyunghyun Cho
6. Li Deng
7. James R. Glass
8. Graham Neubig

Citations in the corpus

• Automatic Speech Recognition (2005, methodology): “The toolkit depends on two external libraries that are also freely available: one is OpenFst [9] for the finite-state framework and the other is numerical algebra libraries.”
• An Analysis of Incorporating an External Language Model into a Sequence-to-Sequence Model (2017, methodology): “Following [2], an alternative is to train the LM at the word level, and then, using the Weighted Finite State Transducer framework [17, 18], compose it with a “speller” which breaks each word into its component units (graphemes or wordpieces).”
• Learning to Transduce with Unbounded Memory (2015, background): “The most common approach leverages symbolic finite state transducers [6, 7], with approaches based on context free representations also being popular [8].”
• The subspace Gaussian mixture model - A structured model for speech recognition (2011, methodology): “We used the OpenFST [1] Weighted Finite State Transducer [24] library and command line tools to simplify the tasks of training and decoding.”
• Improved recognition of contact names in voice commands (2015, methodology): “An implementation of the on-demand replacement operation is available in the OpenFst library [5].”
• Syntactically Guided Neural Machine Translation (2016, methodology): “Our SGNMT system1 is built with the Pyfst interface 2 to OpenFst (Allauzen et al., 2007).”
• Variable length word encodings for neural translation models (2016, methodology): “Lattice manipulations are done through OpenFST through PyFst (Allauzen et al. (2007)).”
• EESEN: End-to-end speech recognition using deep RNN models and WFST-based decoding (2015, methodology): “Using highly-optimized FST libraries such as OpenFST [29], we can fuse the WFSTs efficiently into a single search graph.”
• The Kaldi Speech Recognition Toolkit (2011, methodology): “Integration with Finite State Transducers: We compile against the OpenFst toolkit [5] (using it as a library).”
• Recurrent Neural Networks as Weighted Language Recognizers (2017, methodology): “Moreover, toolkits like OpenFST (Allauzen et al., 2007) and Carmel1 implement efficient algorithms on automata like minimization, intersection, finding the highestweighted path and the highest-weighted string.”
• Outside the Closed World: On Using Machine Learning for Network Intrusion Detection (2010, methodology): “To quote from Witten et al. [21]: The idea of specifying only positive examples and adopting a standing assumption that the rest are negative is called the closed world assumption .”
• Towards Better Decoding and Language Model Integration in Sequence to Sequence Models (2016, methodology): “We have use the FST framework to compose the language model with a "spelling lexicon" [6, 12, 19].”
• Snips Voice Platform: an embedded Spoken Language Understanding system for private-by-design voice interfaces (2018, methodology): “This is achieved by replacing the compositions of equation (1) by delayed (or lazy) ones [4].”
• End-to-end attention-based large vocabulary speech recognition (2015, methodology): “We use the Weighted Finite State Transducer (WFST) framework [18, 19] to build a character-level language model from a word-level one.”