Connectionist Temporal Classification: Labelling Unsegmented Sequence Data with Recurrent Neural Networks

6 Discussion and Future Work

1. What are the benefits and limits of CTC not segmenting its input?
   It needn’t locate inherently ambiguous boundaries (in speech or handwriting) and can group predictions of labels that occur together; segmenting is wasted effort if only labels are needed. It seems unsuited where segmentation is required, but its spikes roughly align with the input, enough for e.g. keyword spotting.
   • Graves et al., 2006, Connectionist Temporal Classification (Proceedings of the 23rd International Conference on Machine Learning (ICML 2006)), §6, pp. 375–376
2. Does CTC model dependencies between output labels?
   Not explicitly, unlike graphical models that assume a \(k\)th-order Markov chain over labels. It does so implicitly, through the RNN’s state, e.g. predicting labels that commonly occur together as a double spike.
   • Graves et al., 2006, Connectionist Temporal Classification (Proceedings of the 23rd International Conference on Machine Learning (ICML 2006)), §6, pp. 375–376