This work is an attempt to enhance query classification in call routing applications. A new method has been introduced to learn weights from training data by means of a regression model. This work has investigated applying the tf-idf weighting method, but the approach is not limited to a specific method and can be used for any weighting scheme. Empirical evaluations with several classifiers including Support Vector Machines (SVM), Maximum Entropy, Naive Bayes, and k-Nearest Neighbor (k-NN) show substantial improvement in both macro and micro F1 measures.

This work is an attempt to enhance query classification in call routing applications. A new method has been introduced to learn weights from training data by means of a regression model. This work has investigated applying the tf-idf weighting method, but the approach is not limited to a specific method and can be used for any weighting scheme. Empirical evaluations with several classifiers including Support Vector Machines (SVM), Maximum Entropy, Naive Bayes, and k-Nearest Neighbor (k-NN) show substantial improvement in both macro and micro F1 measures.

Semantic Web and Linked Data community is now the reality of the future of the Web. The standards and technologies defined in this field have opened a strong pathway towards a new era of knowledge management and representation for the computing world. The data structures and the semantic formats introduced by the Semantic Web standards offer a platform for all the data and knowledge providers in the world to present their information in a free, publicly available, semantically tagged, inter-linked, and machine-readable structure. As a result, the adaptation of the Semantic Web standards by data providers creates numerous opportunities for development of new applications which were not possible or, at best, hardly achievable using the current state of Web which is mostly consisted of unstructured or semi-structured data with minimal semantic metadata attached tailored mainly for human-readability. This dissertation tries to introduce a framework for effective analysis of the Semantic Web data towards the development of solutions for a series of related applications. In order to achieve such framework, Wikipedia is chosen as the main knowledge resource largely due to the fact that it is the main and central dataset in Linked Data community. In this work, Wikipedia and its Semantic Web version DBpedia are used to create a semantic graph which constitutes the knowledgebase and the back-end foundation of the framework. The semantic graph introduced in this research consists of two main concepts: entities and topics. The entities act as the knowledge items while topics create the class hierarchy of the knowledge items. Therefore, by assigning entities to various topics, the semantic graph presents all the knowledge items in a categorized hierarchy ready for further processing. Furthermore, this dissertation introduces various analysis algorithms over entity and topic graphs which can be used in a variety of applications, especially in natural language understanding and knowledge management fields. After explaining the details of the analysis algorithms, a number of possible applications are presented and potential solutions to these applications are provided. The main themes of these applications are entity detection, topic identification, and context acquisition. To demonstrate the efficiency of the framework algorithms, some of the applications are developed and comprehensively studied by providing detailed experimental results which are compared with appropriate benchmarks. These results show how the framework can be used in different configurations and how different parameters affect the performance of the algorithms.

Semantic Web and Linked Data community is now the reality of the future of the Web. The standards and technologies defined in this field have opened a strong pathway towards a new era of knowledge management and representation for the computing world. The data structures and the semantic formats introduced by the Semantic Web standards offer a platform for all the data and knowledge providers in the world to present their information in a free, publicly available, semantically tagged, inter-linked, and machine-readable structure. As a result, the adaptation of the Semantic Web standards by data providers creates numerous opportunities for development of new applications which were not possible or, at best, hardly achievable using the current state of Web which is mostly consisted of unstructured or semi-structured data with minimal semantic metadata attached tailored mainly for human-readability. This dissertation tries to introduce a framework for effective analysis of the Semantic Web data towards the development of solutions for a series of related applications. In order to achieve such framework, Wikipedia is chosen as the main knowledge resource largely due to the fact that it is the main and central dataset in Linked Data community. In this work, Wikipedia and its Semantic Web version DBpedia are used to create a semantic graph which constitutes the knowledgebase and the back-end foundation of the framework. The semantic graph introduced in this research consists of two main concepts: entities and topics. The entities act as the knowledge items while topics create the class hierarchy of the knowledge items. Therefore, by assigning entities to various topics, the semantic graph presents all the knowledge items in a categorized hierarchy ready for further processing. Furthermore, this dissertation introduces various analysis algorithms over entity and topic graphs which can be used in a variety of applications, especially in natural language understanding and knowledge management fields. After explaining the details of the analysis algorithms, a number of possible applications are presented and potential solutions to these applications are provided. The main themes of these applications are entity detection, topic identification, and context acquisition. To demonstrate the efficiency of the framework algorithms, some of the applications are developed and comprehensively studied by providing detailed experimental results which are compared with appropriate benchmarks. These results show how the framework can be used in different configurations and how different parameters affect the performance of the algorithms.