Recommending Recipients in the Enron Email Corpus

Email is the most popular communication tool of the internet. In this paper we investigate how email systems can be enhanced to work as recipient recommendation systems, i.e., suggesting who recipients of a message might be, while the message is being composed, given its current contents and given its previously-specified recipients. This can be a valuable addition to email clients, particularly in large corporations. It can be used to identify people in an organization that are working in a similar topic or project, or to find people with appropriate expertise or skills. Recipient recommendation can also prevent a user from forgetting to add an important collaborator or manager as recipient, preventing costly misunderstandings and communication delays. In this paper we present the first study of recipient recommendation in a real large-scale corporate email collection, the Enron Email corpus. We begin by defining the problem as a large multi-class multi-label classification task, where each email can be addressed to multiple recipients in the user’s address book (i.e., each class is equivalent to an email address in the address book). We propose various baselines to the problem, along with a classification-based reranking scheme to combine two types of features: textual contents and network information from the email headers. Experiments indicate that the reranking scheme significantly outperforms the baselines, and that the best scheme is accurate enough to be useful in email clients. Results are encouraging also because the proposed solution can be easily implemented in any email client – with no changes in the email server side.