A Bengali Social Media Politeness and Formality Dataset with Conversational Context Data

This dataset presents a carefully curated collection of Bengali comments annotated for tone, collected from the official Facebook pages of two prominent Bangladeshi news portals, Kaler Kantho and Bangladesh Times, spanning the period from February 1st, 2026, to March 17th, 2026. All comments were sourced from publicly available content, ensuring that no private or personally identifiable information was included. The dataset comprises 1,100 comments, covering diverse topics such as politics, current events, social issues, and general discussions.

Each comment is labeled according to its tone as Polite, Rude, or Neutral, with the following distribution: Polite – 358, Rude – 380, Neutral – 362. Beyond basic text, the dataset provides contextual information, including the source article title and article context, conversational context in the form of previous and next comments. This allows for richer analysis of tone within dialogues and discussion threads.

Comments labeled Polite are generally respectful, constructive, and display elements of formal courtesy, often encouraging healthy discussion or offering supportive feedback. For example:“সৎ ও সাহসীরা কখনো হারে না। কিন্তু পথটা সাময়িক কঠিন হয়। আর ফিরে আসে আরো বলিষ্ঠ হয়ে।” translated as “The honest and brave never lose. But the path is temporarily difficult. And they come back stronger.” Rude comments, on the other hand, may contain sarcasm, offensive language, or personal attacks, reflecting informal or aggressive expressions. For example: “এমন মন্তব্য করা অল্পবুদ্ধির কাজ।“ translated as “Making such a comment is foolish.” Neutral comments are factual, informative, or general observations, without explicit politeness or rudeness. For example:“আজকের খবরটি খুব গুরুত্বপূর্ণ।” translated as “Today’s news is very important.”

This dataset is well-suited for Bengali sentiment analysis, tone detection, conversational AI research, and politeness studies, offering a balanced representation of polite, rude, and neutral tones. Its inclusion of both informal and formal expressions makes it particularly valuable for exploring nuances of politeness and social dynamics in Bengali online discourse.