Topic Classifier for Customer Service Dialog Systems
Authors: Manex Serras Saenz Naiara Perez Miguel María Inés Torres Raquel Justo
Date: 11.12.2015
Abstract
Using dialog systems to automatize customer services is becoming a common practice in many business fields. These dialog systems are often required to relate the users’ issues with a department of the company, which is especially hard when each department covers a wide range of topics. This paper proposes an entropy-based classifier to support the dialog system’s dialog manager in the decision-making. As the classifier is implemented in a feedback-available scenario, an extra class has been inserted with which uncertain cases are labeled, in order to retrieve more information from the user. This way, robust decisions are always ensured. The classifier’s input is the sequence of semantic units decoded from the user turn, extracted from technical records in this paper. This allows the designers to introduce domain specific knowledge and reduce the classifier’s workload. Experiments show that the classifier achieves a high precision, slightly improving some SVM and Bayes classifiers.
BIB_text
title = {Topic Classifier for Customer Service Dialog Systems},
pages = {140-148},
keywds = {
Topic classification, Dialogue systems, Semantic grammars, Entropy-based classification
}
abstract = {
Using dialog systems to automatize customer services is becoming a common practice in many business fields. These dialog systems are often required to relate the users’ issues with a department of the company, which is especially hard when each department covers a wide range of topics. This paper proposes an entropy-based classifier to support the dialog system’s dialog manager in the decision-making. As the classifier is implemented in a feedback-available scenario, an extra class has been inserted with which uncertain cases are labeled, in order to retrieve more information from the user. This way, robust decisions are always ensured. The classifier’s input is the sequence of semantic units decoded from the user turn, extracted from technical records in this paper. This allows the designers to introduce domain specific knowledge and reduce the classifier’s workload. Experiments show that the classifier achieves a high precision, slightly improving some SVM and Bayes classifiers.
}
isbn = {978-3-319-24032-9},
isi = {1},
date = {2015-12-11},
year = {2015},
}
