瞬间灰色-to new different domains since portability is ensured following the proposed principles of design.
In the future, we hope to have available training data from
we can automatically learn the BN thus avoiding hand-assigned probabilities. We have also incorporated a dialog strategy in order to handle partially observed goals. This strategy is aimed for recovering the omitted information from
the dialog context. Suitability has been demonstrated in several cases considering a different omitted element from the dialog frame each time.
We plan to distinguish between long term and short term memory. The aim of this distinction is to give more relevance
to the short term observations respect to the long term ones since human mind behaves this way in some sense. Exactly,
we would expect the system to automatically reject the erroneous and spurious observations. Thus, we would not need any explicit negation for those observations since they are not referenced by the user anymore.
5.Acknowledgements
This work has been partially funded by the Spanish Ministry
of Science and Technology under contracts: DPI2001-3652-
C02-02 (URBANO-IVANHOE), TIC2003-09192-C11-07 (MIDAS-INAUDITO), and DPI2004-07908-C02-02 (ROBINT).
6.References
[1]J.Ferreiros, J.Colás, J.Macías, A.Ruiz and J.M.Pardo,
“Controlling a HIFI with a continuous speech
understanding system”. ICSLP 1998, Sydney, Australia,
ISBN 1-876346-17-5
[2]J.Ferreiros, J.Colás, J.Macías, R.Córdoba and J.M.Pardo,
“Control de un equipo de alta fidelidad usando frases
habladas de manera natural”. Iberdiscap 2000, Madrid,
Spain.
[3]H.M.Meng, C.Wai and R.Pieraccini. “The Use of Belief
Networks for Mixed-Initiative Dialog Modeling”, IEE
Transactions on Speech and Audio Processing, Vol.11,
NO.6, pps. 757-773, 2003.
[4] C.Huang and A.Darwiche, “Inference in belief networks:
a procedural guide”. International Journal of
Approximate Reasoning 1994 11:1-158. USA.
[5]H.M.Meng, 166ee1e8172ded630b1cb6a8m and K.F.Low. “Learning Belief
Networks for Language Understanding”. ASRU 1999.
