A context-sensitive-chunk BPTT approach to training deep LSTM/BLSTM recurrent neural networks for offline handwriting recognition.
Kai ChenZhi-Jie YanQiang HuoPublished in: ICDAR (2015)
Keyphrases
- recurrent neural networks
- context sensitive
- handwriting recognition
- recurrent networks
- echo state networks
- long short term memory
- feed forward
- historical documents
- character recognition
- neural network
- word recognition
- speech recognition
- artificial neural networks
- word segmentation
- handwritten text
- language model
- natural language
- training set
- information retrieval
- real time
- machine vision
- training examples
- data mining