Báo cáo khoa học: "Spontaneous Speech Understanding for Robust " potx

... large corpus and can therefore be well trained to perform satisfactory speech recognition results. A prominent example for this is the telephone based weather forecast in- formation service JUPITER ... in these systems the speech understanding analysis is constrained to a small set of commands and not oriented towards spontaneous speech. However, deep speech understanding is nece...

Ngày tải lên: 31/03/2014, 01:20

... con- crete recommendations for modifying a question-answering system for improving robustness to spoken input. 1 Introduction This paper demonstrates a multimodal interface for asking questions and ... University Saarbr ¨ ucken, Germany zheng@coli.uni-sb.de Abstract Speech interfaces to question-answering systems offer significant potentialfor find- ing information with phones and mo- b...

Ngày tải lên: 08/03/2014, 04:22

... See text for explanation of notation. The definitions of URL-MI, LEX-MI, and BOW-MI for LOC, ORG and O are analogous to those for PER. For better readability, we write  [[x]] for  x . for compound ... like Ford is often a fully qualified, unambiguous name like Ford Mo- tor Corporation or Gerald Ford. In a short query like buy ford or ford pardon, there is much less con- text than...

Ngày tải lên: 23/03/2014, 16:20

... similarity measure. As for BC b and BC a , the results for the tuned and several other values for α are shown. Figure 1 shows the parameter tuning for BC b with MAP as the y-axis (results for BC a are ... tool for constructing a synonym dictionary using context information. In IPSJ SIG Technical Report (in Japanese), pages 87–94. 256 Proceedings of the 48th Annual Meeting of the A...

Ngày tải lên: 30/03/2014, 21:20

... with a speech repair identify- ing language model to handle speech repairs, editing terms, and two speakers. Section 2 details the allowable forms of collaboration, interruption, and speech ... take the ban- um the oranges, three constituents would be produced, one for urn, another for take the ban-, and a third for take the oranges. Another complicating factor of dialog i...

Ngày tải lên: 20/02/2014, 19:20

... acoustic-prosodic cues for repair processing. Discussion In this paper, we have presented a" ;speech- first" model, the Repair Interval Model, for studying repairs in spon- taneous speech. This ... the speech signal may facilitate repair processing. We identify several cues based on acoustic and prosodic analysis of repairs in a corpus of spontaneous speech, and prop...

Ngày tải lên: 20/02/2014, 21:20

... baseline performance achieved by always guessing the majority class is 67.09% for break index, 54.10% for pitch accent, 66.23% for phrase accent and 79.37% for boundary tone based on the speech ... compo- nent provides very rich syntactic and semantic information which has not been explored before for intonation modeling. This includes, for ex- ample, the semantic role play...

Ngày tải lên: 08/03/2014, 05:21

... 202–210, Athens, Greece, 30 March – 3 April 2009. c 2009 Association for Computational Linguistics Re-Ranking Models For Spoken Language Understanding Marco Dinarelli University of Trento Italy dinarelli@disi.unitn.it Alessandro ... project C (bestpath 1 (λ SLU )), where bestpath n (in this case n is 1 for the 1-best hypothesis) performs a Viterbi search on the FST 203 and outputs th...

Ngày tải lên: 08/03/2014, 21:20

... performance on both by sharing infor- mation between them. We present a joint model for these two subtasks: it is joint not only in that it performs both tasks simultaneously, sharing in- formation, ... used top three tag-sets and top three lemmas for each tag for training. for word w i , for j = 1 . . . k. Also, let l i (t) j de- note the top lemmas for word w i given tag t. An a...

Ngày tải lên: 17/03/2014, 01:20

... examined how to compensate for such noise using acoustic techniques, few have attempted to leverage information in the visual stream to improve speech recognition performance (for an exception see ... can also provide use- ful information for representing non-linguistic con- text. We use boosted decision trees to classify audio into segments of speech, excited _speech, cheering...

Ngày tải lên: 17/03/2014, 02:20