Title | ||
---|---|---|
Topic Identification Based On Bayesian Belief Networks In The Context Of An Air Traffic Control Task |
Abstract | ||
---|---|---|
In this paper we present a topic identification task based on a Bayesian Belief Network approach. These networks are trained with a number of semantic concepts which have been tagged for each utterance and defined by an expert in the application domain. The target topics are the five control positions available at the airport. In order to evaluate the performance of our approach we apply a block based evaluation scheme. The lower error rate that we obtained was 3.5% using a winner takes all evaluation scheme and using five utterances per block. Finally, we compare these results with those obtained by a Bayesian classifier considering a parameter vector constituted by the resultant perplexities, at phrase level, applying a trigram language model for each topic of the task; the obtained results allow us to know intuitively the importance of including temporal information into the BN in future works. |
Year | Venue | Keywords |
---|---|---|
2005 | PROCESAMIENTO DEL LENGUAJE NATURAL | Topic Identification, Bayesian Belief Networks, N-gram, Air Traffic Control |
DocType | Volume | Issue |
Journal | 35 | 35 |
ISSN | Citations | PageRank |
1135-5948 | 0 | 0.34 |
References | Authors | |
4 | 5 |
Name | Order | Citations | PageRank |
---|---|---|---|
fernando fernandez martinez | 1 | 0 | 0.34 |
L. F. D'HARO | 2 | 33 | 2.83 |
J. FERREIROS | 3 | 112 | 14.84 |
J. M. Montero | 4 | 37 | 2.74 |
R. SAN-SEGUNDO | 5 | 139 | 14.28 |