| << Prev | - Up - |
Automated tagging and recognition of dialogue act is based on training probabilistic models for recognising dialogue acts.
Automated tagging and recognition of dialogue act is based on training probabilistic models for recognising dialogue acts. Training is done by use of annotated corpora. Recognition can be based on cues (e.g., lexical, prosodic). Word and dialogue act n-grams as well as decision trees and neural networks are used in modeling dialogue acts and their function. Discourse structure information is modeled as a hidden Markov model (HMM). A hidden Markov model in this case is a kind of an automaton where dialogue acts correspond to states and utterances are the observable units. The discourse structure information is used for the recognition/prediction of dialogue acts. This process is further enhanced by word n-grams for every dialogue act.
| << Prev | - Up - |