amirieb/churn-dataset

Churn Dataset

The dataset contains labeled tweets about three telco brands: Verizon, AT&T, and T-Mobile. Tweet are labeled as churny or not-churny, where churny tweets indicate a high risk of canceling the brand's service. Labels are obtained through crowdsourcing and each tweet is labeled by at least three annotators. Fleiss’ kappa is 0.62, which indicates substantial agreement among annotators.

Data Description

• uid: user id
• tid: tweet id
• brand: name of target brand
• choose_one: label of the tweet wrt to the target brand, either "Churny" or "Non-churny"
• choose_one:confidence: a value c=<1.0 showing the aggregated confidence of the judgements

Twitter developer agreement & policy doesn't allow shareing tweet content. Reach out to Hadi Amiri at [email protected] if you'd like to obtain access to the full dataset.

Publication

Hadi Amiri and Hal Daumé III, Short Text Representation for Detecting Churn in Microblogs. AAAI 2016. PDF Hadi Amiri and Hal Daumé III, Target-dependent Churn Classification in Microblogs. AAAI 2015. PDF