Differentiable Weighted Finite-State Transducers
Hannun, Awni, Pratap, Vineel, Kahn, Jacob, Hsu, Wei-Ning
E B. (2) The primary difference between ASG and CTC is the inclusion of a blank token, b, represented by the graph in figure 3a. Constructing CTC amounts to including the blank token graph when constructing the full token graph T. The intersection T Y then results in the CTC alignment graph (Figure 1b). Note, this version of CTC does not force transitions on b between repeats tokens. This requires remembering the previous state and hence is more involved (see Appendix A.1 for details). A benefit of constructing sequence-level criteria by composing operations on simpler graphs is the access to a large design space of loss functions with which we can encode useful priors. For example we could construct a "spike" CTC, a "duration-limited" CTC, or an "equally spaced" CTC by substituting the appropriate token graphs into equation 2 (see Appendix A.2 for details).
Oct-2-2020
- Country:
- Asia > India (0.04)
- South America > Chile
- North America
- Puerto Rico (0.04)
- United States
- New York (0.04)
- New Jersey (0.04)
- Europe
- Genre:
- Instructional Material (0.46)
- Research Report (0.40)
- Technology:
- Information Technology > Artificial Intelligence
- Representation & Reasoning (1.00)
- Speech > Speech Recognition (0.71)
- Vision (0.69)
- Natural Language > Text Processing (0.68)
- Machine Learning > Neural Networks
- Deep Learning (1.00)
- Information Technology > Artificial Intelligence