ACL RD-TEC 1.0 Summarization of J01-1002
Paper Title:
INTEGRATING PROSODIC AND LEXICAL CUES FOR AUTOMATIC TOPIC SEGMENTATION
INTEGRATING PROSODIC AND LEXICAL CUES FOR AUTOMATIC TOPIC SEGMENTATION
Authors: Gokhan Tur and Andreas Stolcke and Dilek Hakkani-Tur and Elizabeth Shriberg
Primarily assigned technology terms:
- algorithm
- audio browsing
- automatic extraction
- automatic recognition
- automatic segmentation
- automatic speech recognizer
- automatic topic segmentation
- baum-welch algorithm
- beam search
- boundary classification
- broadcast news recognizer
- capitalization
- classification
- classifier
- classifiers
- computational linguistics
- computational modeling
- cross-validation
- decision tree
- decision tree learning
- decision trees
- detection and tracking
- disfluency detection
- error reduction
- evaluation framework
- feature extraction
- feature selection
- feature subset selection
- forward-backward algorithm
- hidden markov
- hidden markov models
- hmm segmentation
- hmm topic segmentation
- hmm-based approach
- hmm-based combination
- hmms
- hypothesizing
- information retrieval
- information retrieval technique
- k-means
- knowledge source combination
- language modeling
- learning
- learning algorithm
- learning approach
- learning techniques
- lexical modeling
- local context analysis
- machine learning
- machine learning approach
- machine learning techniques
- maximum entropy
- maximum entropy model
- model training
- modeling
- news recognizer
- normalization
- phonetic alignment
- processing
- prosodic modeling
- pruning
- recognition
- recognition systems
- recognizer
- retrieval technique
- search
- segmentation
- segmentation work
- segmenter
- signal processing
- single classifier
- speaker segmentation
- speech recognition
- speech recognizer
- statistical language modeling
- statistical modeling
- task-oriented dialogue
- tdt evaluation
- tdt segmentation
- topic detection
- topic detection and tracking
- topic segmentation
- tree construction
- tree learning
- tuning
- viterbi
- viterbi algorithm
- word recognition
- word-based segmentation
Other assigned terms:
- alignment information
- anchors
- approach
- baseline model
- beam
- boundary information
- broadcast news
- broadcast news corpus
- broadcast news data
- broadcast news speech
- case
- class distribution
- classification accuracy
- classification problem
- cluster
- clusters
- complete graph
- computational framework
- continuous speech
- contour
- corpora
- cosine similarity
- cue phrases
- cue words
- data consortium
- data set
- debugging
- discourse
- discourse markers
- discourse segment
- distribution
- duration
- duration information
- entropy
- error rate
- evaluation metric
- evaluation metrics
- evaluation paradigm
- evaluation test
- events
- exponential model
- extraction process
- fact
- feature
- feature set
- feature sets
- feature type
- feature types
- frame
- heuristics
- interpretation
- intonational contour
- knowledge
- labeling
- language model
- language models
- large training
- lexical cues model
- lexical discourse
- lexical features
- lexical information
- lexical knowledge
- lexical model
- likelihood
- likelihood ratio
- linguistic
- linguistic data
- linguistic data consortium
- linguistics
- local context
- markov models
- markup
- measure
- measures
- method
- model combination
- model parameters
- model performance
- model structure
- monologue
- names
- news corpus
- nist
- noun phrases
- paragraph
- particles
- pause
- pause duration
- pauses
- pitch
- posterior
- posterior distribution
- posterior probability
- priori
- probabilistic approach
- probabilistic model
- probabilistic models
- probabilities
- probability
- procedure
- process
- prosodic feature
- prosodic features
- prosodic information
- prosodic model
- prosody
- punctuation
- queries
- recognition errors
- relative error reduction
- relative frequency
- segmentation accuracy
- segments
- sentence
- sentence boundaries
- sentence boundary
- sentence level
- sentence punctuation
- sentences
- signal
- skewed class distribution
- speaker identity
- speaker identity and gender
- speech input
- speech prosody
- statistical model
- statistics
- style
- syntactic constructions
- syntactic structures
- technique
- term
- test corpus
- test data
- test set
- text
- text structure
- textual input
- tokens
- tone
- topic choice
- topics
- topology
- training
- training corpus
- training data
- training set
- transcriptions
- transcripts
- transition probabilities
- tree
- trees
- understanding
- unigram
- unigram language model
- utterance
- vector space
- word
- word boundary
- word error rate
- word sequence
- word usage
- word vector
- word-based evaluation
- words
- wrapper
