Flexible speech understanding based on combined key-phrase detection and verification

Abstract

We propose a novel speech understanding strategy based on combined detection and verification of semantically tagged key-phrases in spontaneous spoken utterances. Key-phrases are defined in a top-down manner so as to constitute semantic slots. Their detection directly leads to robust understanding. A phrase network realizes both a wide coverage and a reasonable constraint for detection. A subword-based verifier is then incorporated to reduce false alarms in detection and attach confidence measures of the detected phrases. This set of phrase confidence measures, when incorporated in a spoken dialogue system, forms a basis for designing intelligent speech interfaces that accept only verified key-phrases and reprompt users to clarify unspecified or unrecognized portions. Several forms of confidence measures based on subword-level tests are investigated. The proposed approach was tested on field data collected from realworld trial applications. The combined detection and verification strategy drastically improves the accuracy in handling out-ofgrammar utterances over the conventional decoding approaches while maintaining the performance for in-grammar utterances.

Speech act modeling and verification of spontaneous speech with disfluency in a spoken dialogue system

“…Section III elucidates a hybrid verification model, LSA-based BBM, for speech act verification. Section IV compare experimental results obtained using the proposed method to those obtained using the keyword-based system [9], which is applied to an ATIS. Finally, Section V summarizes the findings and draws a brief conclusion.…”

“…Such methods will generate semantic errors in accepted in-domain input. Some previous studies on verification strategies [9] have been based on the combined detection and verification of semantically tagged key-phrases in spontaneous speech. They detect key-phrases in speech directly and perform optimization jointly with applying the semantic constraints in a sentence parsing step.…”

“…Second, the identified semantic slots are verified to improve the accuracy of their identification. For speech act modeling, the keyword-based approach [9] is frequently used. This approach cannot extract the semantic information and syntactic structure of a sentence.…”

“…During recent years, it has become essential to equip spoken dialogue systems with the ability to accommodate spontaneous speech input [9]. Some approaches [15], [16] verify word-level speech recognition outputs and out-of-domain requests using language model characteristics.…”

Miscommunication handling in spoken dialog systems based on error-aware dialog state detection

“…In spoken language understanding (SLU), dialog states (DSs) are the basic functional units [16] that describe the dialog behaviors in humancomputer or human-human communication [17]. The features used to represent an utterance for DS detection include parts-of-speech (POSs) [18], semantic roles [19,20], prosody [21,22], and keywords [23]. With semantic analysis, statistical dialog management models, such as weighted finite-state transducers (WFSTs) [24], Markov decision processes (MDPs) [25,26], and partially observable MDPs (POMDPs) [8,13], were proposed for stable dialog flow control, especially in goal-oriented SDSs.…”