TY  - CHAP
AB  - This article presents the speech understanding and dialog system EVAR and concentrates on the linguistic processing of the system. The task of EVAR is to lead an information retrieval dialogue about the German InterCity train system. The linguistic knowledge is uniformly represented in a semantic network structure. The knowledge base is well structured following the layered linguistic model. Interface to the speech recognition is the level of word hypotheses. The control algorithm is independent of the application and allows to alternate dynamically between the two fundamental control strategies top-down and bottom-up. The linguistic knowledge represented in the system is used to control the recognition phase and for the understanding phase. The system can handle defect input due to recognition errors. Results are presented for speaker-dependent and multi-speaker versions of the recognition module.
KW  - Sprachverarbeitung
KW  - Kongress
LA  - ger
PY  - 1992
SN  - 3-540-55959-0
TI  - EVAR: ein sprachverstehendes Dialogsystem
UR  - https://nbn-resolving.org/urn:nbn:de:0070-bipr-21632
Y2  - 2024-11-22T04:42:41
ER  -