L	M	X	J	V	S	D
1	2	3	4	5	6	7
8	9	10	11	12	13	14
15	16	17	18	19	20	21
22	23	24	25	26	27	28
29	30

University on the Phone (UAH)

The spoken dialogue system University on the Phone (Universidad Al Habla, UAH) provides oral access on the telephone to academic information related to our University. The system is implemented using dynamic generation of VoiceXML documents with PHP, so that it can adapt to the users’ needs in real time during the dialogue [1].

The main novelty of the system was a module designed to automatically create JSGF (Java Speech Grammar Format) grammar that are used during the automatic speech recognition process. The vocabulary to be included in these grammars is unknown a priori, and is extracted from a database [2].

From the recordings gathered by the UAH system during its use in our university, a dialogue corpus was obtained which has been employed in several of our investigations. Firstly, this corpus, along with the results of voluntary subjective assessments was used to evaluate the system and detect which are the peculiarities of evaluating using non-recruited users [3].

Secondly, the UAH corpus was annotated with emotional categories and has been employed to carry out several studies about the appropriateness of standard measures for inter-annotator agreement when they categorize non-acted emotions [4]. Additionally, the emotional corpus has been employed as a basis for our studies on emotion recognition [5].

Currently, we are incorporating to UAH the possibility to manage the dialogue with the user according with his emotional state, which is retrieved using our emotion recognition techniques over his utterances.

REFERENCES:

[1] Callejas, Z., López-Cózar, R. 2005, Implementing modular dialogue systems. A case study, Proc. COST278 and ISCA Tutorial and Research Workshop (ITRW) on Applied Spoken Language Interaction in Distributed Environments, Aalborg (Denmark). ISSN 0908-1224

[2] Callejas, Z., López-Cózar, R. 2007, Automatic creation of ASR grammar rules for unknown vocabulary applications. Proc. 8th International Workshop on Electronics, Control, Modelling, Measurement and Signals (ECMS), Liberec (Czech Republic). ISBN 978-80-7372-202-9

[3] Callejas, Z., López-Cózar, R., 2008, Relations between de-facto criteria in the evaluation of a spoken dialogue system, Speech Communication. Speech Communication, vol. 50(8-9), pp. 646-665 ISSN 0167-6393

[4] Callejas, Z., López-Cózar, R., 2009, Improving acceptability assessment for the labelling of affective speech corpora, In Proc. of Interspeech 2009, pp. 1863-2866.

[5] Callejas, Z., López-Cózar, R., 2008, Influence of contextual information in emotion annotation for spoken dialogue systems, Speech Communication, vol. 50(5), pp. 416-433 ISSN 0167-6393