ACL RD-TEC 1.0 Summarization of N06-1029
Paper Title:
UNSUPERVISED AND SEMI-SUPERVISED LEARNING OF TONE AND PITCH ACCENT
UNSUPERVISED AND SEMI-SUPERVISED LEARNING OF TONE AND PITCH ACCENT
Primarily assigned technology terms:
- accent classification
- accent recognition
- algorithm
- approximation
- asymmetric clustering
- bagging
- binary classification
- boosting
- class assignment
- classification
- classification algorithm
- classification techniques
- classifier
- classifier learning
- classifiers
- clustering
- clustering algorithm
- clustering technique
- contextual representation
- data representation
- decision trees
- decomposition
- detection and tracking
- discriminative classifier
- feature extraction
- feature selection
- feature selection process
- hidden markov
- hidden markov model
- identification
- k-means
- k-means clustering
- kernel
- language acquisition
- language understanding
- learner
- learning
- learning approach
- learning framework
- machine learning
- machine learning and classification
- manifold regularization
- markov model
- markov model speech recognition
- modeling
- natural language interaction
- nearest neighbors
- parameter setting
- pitch accent classification
- pitch accent recognition
- radial basis function
- recognition
- recognition framework
- regularization
- regularization framework
- reporting
- sampling
- segmentation
- selection process
- semi-supervised classification
- semi-supervised learning
- single clustering
- spectral clustering
- speech recognition
- speech recognition framework
- supervised classification
- supervised learning
- support vector machine
- support vector machines
- surface realization
- svm classifier
- syllabification
- syllable segmentation
- tone classification
- tone learning
- tone learning and recognition
- tone realization
- topic detection
- topic detection and tracking
- transcription
- unsupervised acquisition
- unsupervised clustering
- unsupervised learning
Other assigned terms:
- accent
- anchor
- annotation
- approach
- approximation model
- broadcast news
- broadcast news audio
- broadcast news corpus
- broadcast news data
- broadcast news speech
- case
- classification accuracy
- classification task
- cluster
- clusters
- contextual features
- contextual information
- corpora
- cosine distance
- cosine distance measure
- data set
- data sets
- dictionary
- distance measure
- distribution
- english corpus
- english pitch accent
- euclidean distance
- experimental results
- extraction process
- feature
- gold standard
- implementation
- intonation
- labeled training data
- language model
- linguistic
- linguistic data
- mandarin chinese
- mandarin chinese tone
- manual annotation
- mapping
- maps
- meaning
- measure
- measures
- modeling problem
- natural language
- news audio
- news corpus
- nucleus
- phonetic research
- phrase
- phrase structure
- pinyin
- pitch
- pitch accent
- pitch height and range
- pitch rise
- process
- pronunciation
- pronunciation dictionary
- radio news corpus
- recognition accuracy
- recognition task
- semi-supervised approach
- sentences
- speech data
- support vector
- svms
- syllables
- technique
- terms
- text
- theory
- tone
- training
- training data
- training examples
- training set
- transcriptions
- transcripts
- trees
- understanding
- unlabeled examples
- vowel
- word
