ACL RD-TEC 1.0 Summarization of H05-1026
Paper Title:
TRAINING NEURAL NETWORK LANGUAGE MODELS ON VERY LARGE CORPORA
TRAINING NEURAL NETWORK LANGUAGE MODELS ON VERY LARGE CORPORA
Authors: Holger Schwenk and Jean-Luc Gauvain
Primarily assigned technology terms:
- acoustic model adaptation
- acoustic modeling
- active learning
- algorithm
- back-propagation
- back-propagation algorithm
- boosting
- classifiers
- clustering
- clustering algorithm
- coding
- computational linguistics
- continuous speech recognition
- continuous speech recognizer
- decision tree
- decoder
- decoding
- error reduction
- estimator
- forward pass
- gaussian computation
- human language
- human language technology
- information retrieval
- kernel
- kneser-ney smoothing
- language model training
- language modeling
- language processing
- language technology
- large vocabulary continuous speech recognition
- lattice rescoring
- learning
- learning algorithm
- machine translation
- maximum entropy
- model adaptation
- model training
- modeling
- multi-layer perceptron
- natural language processing
- neural network
- neural network approach
- neural networks
- normalization
- optimization
- perceptron
- probability estimation
- processing
- pruning
- random sampling
- real-time continuous speech recognition
- recognition
- recognition system
- recognition systems
- recognizer
- regularization
- resampling
- rescoring
- sampling
- smoothing
- speech recognition
- speech recognition system
- speech recognition systems
- speech recognizer
- statistical machine translation
- statistical techniques
- support vector machines
- training algorithm
- training procedure
Other assigned terms:
- 4-gram back-off lm
- acoustic model
- acoustic models
- acoustic signal
- approach
- association for computational linguistics
- broadcast news
- broadcast news data
- brown corpus
- cluster
- coefficient
- continuous speech
- convergence
- corpora
- data sparseness
- data sparseness problem
- data structures
- development set
- entropy
- error rate
- estimation
- experimental results
- fact
- feature
- feature vectors
- french
- french broadcast news
- grammar
- hypothesis
- interpolation
- interpolation coefficients
- knowledge
- language model
- language models
- large corpora
- large corpus
- large text corpora
- large training
- large training corpora
- lattice
- lattices
- learning rate
- linguistics
- n-gram
- n-gram model
- n-grams
- natural language
- network architecture
- neural network architecture
- noise
- perplexity
- posterior
- probabilities
- probability
- probability distributions
- procedure
- processing time
- projection
- pronunciation
- sentence
- signal
- sparseness problem
- speaking style
- speech signal
- statistics
- style
- support vector
- technique
- technology
- term
- text
- text corpora
- theory
- toolkit
- training
- training corpora
- training data
- training example
- training examples
- training set
- training time
- transcriptions
- transcripts
- tree
- vocabulary
- word
- word error rate
- word error rates
- word lattice
- word sequence
- words
