Customer Support Conversations on Twitter

Over 100,000 text-based customer support conversations sourced from Twitter. The dataset was created by TNE-AI and last updated in May 2024. It contains interactions between customers and support agents on the platform.

Use Cases

• Train a sequence-to-sequence model to generate agent responses based on customer tweet text and conversation history.
• Analyze sentiment polarity and emotion in customer tweet text to classify complaint urgency.
• Develop a classifier using conversation text to automatically route support requests to appropriate departments.
• Study conversation flow and turn-taking patterns in the text data to improve chatbot dialogue design.

Strengths

• Contains over 100,000 text conversations, providing a substantial corpus for training.
• Data is sourced from a real-world platform (Twitter), reflecting authentic customer service interactions.
• Available in Parquet format, optimized for efficient data processing with libraries like Polars and Pandas.

Limitations

• Specific column names and data schema are unknown, complicating initial analysis.
• The dataset's geographic focus is indicated as 'us', which may limit generalizability to other regions.
• Lack of information on label quality, potential noise, or class distribution for supervised tasks.

Provenance

Source

Twitter (X)

Collection Method

Scraped or collected from the Twitter platform.

Freshness

Last updated on 2024-05-14.

Geography

United States (inferred from platform tag 'Regionus')