Parakweet Lab's Email Intent Data Set

This resource contains training and test data for detecting "intent" sentences in email messages. This data comes from the Enron email corpus. Each labeled example is one sentence from an email. We define "intent" here to correspond primarily to the categories "request" and "propose" in the paper:

Cohen, William W., Vitor R. Carvalho, and Tom M. Mitchell. "Learning to Classify Email into``Speech Acts''." EMNLP. 2004.

In some cases, we also apply the positive label to some sentences from the "commit" category if they contain datetime, which makes them useful. Detecting the presence of intent in email is useful in many applications, e.g., machine mediation between human and email.

Training data: 4213 cases with 1631 positives Test data: 991 cases with 277 positives

Also see https://github.com/vseledkin/enron_intent_dataset_verified