An application of least squares fit mapping to text information retrieval
SIGIR '93 Proceedings of the 16th annual international ACM SIGIR conference on Research and development in information retrieval
An example-based mapping method for text categorization and retrieval
ACM Transactions on Information Systems (TOIS)
Learning to classify text from labeled and unlabeled documents
AAAI '98/IAAI '98 Proceedings of the fifteenth national/tenth conference on Artificial intelligence/Innovative applications of artificial intelligence
Machine Learning
Proceedings of the Eleventh International Florida Artificial Intelligence Research Society Conference
A system for induction of oblique decision trees
Journal of Artificial Intelligence Research
| Hi-index | 0.00 |
Electronic commerce websites often have trouble keeping up with the large amount of customer-service related email they receive. One way to alleviate the problem is to automate responding to that email as much as possible. Many customer messages are in essence frequently asked questions, for which it is easy to provide a reply. This paper explores a staged approach to message understanding: an incoming message is first classified in a specific category. If the category of the message corresponds to a specific frequently asked question, the answer is provided to the customer. If the category corresponds to a more complex question, a finer understanding of the message is attempted. Messages are categorized by a combination of Bayes classifier and regular expressions, that significantly improves performance compared to a simple Bayes classifier. A first version of the system is installed on the FTD website (Florist Transworld Delivery). It can classify more than half of the customer messages, with 2.3% error; three quarters of the categorized messages are frequently asked questions, and receive an automatic response.