Unanswered but Asked

Inspiration

Online question-and-answer platforms contain millions of questions, yet many receive no responses at all.
Instead of studying popularity or engagement, this project focuses on silence — questions that are asked but never answered.

The idea was inspired by the belief that absence itself can be a signal, and that unanswered questions can reveal hidden patterns about clarity, structure, and participation in online knowledge-sharing communities.

What it does

Unanswered but Asked analyzes questions from a public StackExchange dataset and uses machine learning to predict whether a question is likely to remain unanswered.

The project studies structural characteristics such as:

• Question length
• Level of detail in the description
• Number of tags used

It then models how these factors influence the probability of silence.

How we built it

• Used a public StackExchange data dump from the 3D Printing community
• Parsed raw XML data to extract only question posts
• Labeled questions as answered or unanswered
• Cleaned HTML content from titles and bodies
• Engineered simple numeric features such as word counts and tag counts
• Trained a Logistic Regression model for interpretability
• Evaluated performance using precision, recall, and F1-score

All analysis was performed in Python using Jupyter notebooks.

Challenges we ran into

• Processing large XML files efficiently without exhausting memory
• Cleaning noisy HTML content into usable plain text
• Handling class imbalance between answered and unanswered questions
• Keeping the model simple while still extracting meaningful insights

Accomplishments that we're proud of

• Successfully modeling silence as a measurable outcome
• Demonstrating that unanswered questions are not random
• Extracting interpretable insights using a simple machine learning model
• Completing the full data-to-insight pipeline within a short hackathon timeframe

What we learned

• Absence of interaction can be as informative as interaction itself
• Simple features often reveal strong signals before complex models are needed
• Interpretability is critical when telling a data story
• Real-world datasets require careful cleaning and thoughtful assumptions

What's next for Unanswered but Asked

• Extend the analysis across multiple StackExchange communities
• Incorporate text-based features such as TF-IDF
• Analyze time-based effects like posting hour and day
• Build a lightweight interface to evaluate new questions

Built With

• beautiful-soup
• jupyter
• notebook
• pandas
• python
• scikit-learn