7th International Conference on Natural Language Processing. Published by Springer.
Eriks Sneiders. Automated Email Answering by Text Pattern Matching.
Abstract. Answering email by standard answers is a common practice at contact centers. Our research assists this process by creating reply messages that contain one or several standard answers. Our standard answers are linked to representative text patterns that match incoming messages. The system works in three languages. The performance was evaluated on two email sets; the main advantage of our email answering technique is good correctness of the delivered replies.
Hercules Dalianis, Magnus Rosell, Eriks Sneiders. Clustering E-Mails for the Swedish Social Insurance Agency? What Part of the E-Mail Thread Gives the Best Quality?
Abstract. We need to analyze a large amount of emails sent to a customer service of a governmental organization based in Sweden from citizens. To carry out this analysis we need to cluster a large number of emails with the aim of automatic email answering. The research problem is whether we should use the whole email including the thread or just the original query for the clustering. We have in this paper described the experiments to find out this. Our results shows that only the query and the answering part should be used, but not necessarily the whole email thread. Our normalized mutual information (NMI) for the query and answering part is 0.12 higher than for only the query.