ACL RD-TEC 1.0 Summarization of C04-1168
Paper Title:
A UNIFIED APPROACH IN SPEECH-TO-SPEECH TRANSLATION: INTEGRATING FEATURES OF SPEECH RECOGNITION AND MACHINE TRANSLATION
A UNIFIED APPROACH IN SPEECH-TO-SPEECH TRANSLATION: INTEGRATING FEATURES OF SPEECH RECOGNITION AND MACHINE TRANSLATION
Authors: Ruiqiang Zhang and Genichiro Kikui and Hirofumi Yamamoto and Frank Soong and Taro Watnabe and Wai Kit Lo
Primarily assigned technology terms:
- algorithm
- automatic speech recognition
- automatic translation
- automatic translation evaluation
- beam search
- computing
- continuous speech recognizer
- coupling
- database
- decoder
- decoding
- decoding procedure
- dialogue translation
- dialogue translation technology
- giza
- hmms
- hypothesis translation
- japanese-to-english translation
- language translation
- lexicon translation
- machine translation
- machine translation system
- matching
- model training
- normalization
- optimization
- optimization algorithm
- parameter optimization
- phrase translation
- processing
- pruning
- recognition
- recognition engine
- recognition improvement
- recognizer
- rescoring
- search
- search method
- searching
- speech dialogue translation
- speech recognition
- speech recognition and translation
- speech recognition engine
- speech recognition module
- speech recognizer
- speech translation
- speech translation system
- speech-to-speech translation
- statistical machine translation
- statistical machine translation system
- transducer
- translation evaluation
- translation process
- translation system
- translation systems
- translation technology
- translators
- viterbi
- word bigram
Other assigned terms:
- acoustic information
- acoustic model
- acoustic models
- approach
- arithmetic mean
- baseline model
- beam
- bigram
- bleu
- bleu score
- brevity penalty
- case
- compact representation
- compound words
- continuous speech
- data set
- edit distance
- english language
- english language model
- english part-of-speech
- english sentence
- error rate
- evaluation metrics
- experimental results
- feature
- formalism
- frame
- geometric mean
- hypotheses
- hypothesis
- ibm model
- information sources
- japanese language
- language model
- language model probability
- language models
- large corpus
- lexicon
- lexicon model
- likelihood
- log-linear model
- log-linear models
- machine translation module
- measure
- method
- model parameters
- model probability
- n-gram
- nist
- objective translation
- part-of-speech
- part-of-speech tags
- phrase
- probabilities
- probability
- procedure
- process
- recognition accuracy
- recognition module
- reference translations
- sentence
- sentences
- single-best case
- source language
- speech data
- speech recognition accuracy
- speech recognition performance
- speech translation model
- statistics
- syntactic structures
- tags
- target languages
- technique
- technology
- terms
- test corpus
- test data
- test data set
- text
- text database
- training
- training corpus
- translation candidate
- translation candidates
- translation model
- translation models
- translation module
- translation probabilities
- translation quality
- translations
- travel expression corpus
- trigram
- trigram language model
- triphone
- utterance
- vector space
- vocabulary
- word
- word error rate
- word order
- word pair
- word trigram
- words
