• 

  Market research question inside the game

• 

  Automatically generated trivia question inside the game

• 

  Scorecard at the end of a game

• 

  Market research survey creation

Questor

Inspiration

• Problem. A big challenge that market researchers face is inspiring target users to take their survey. Survey responses are valuable, but variable completion rates may leave market researchers with biased results and uncertain conclusions.
• Idea. Tens of millions of people play trivia games around the world. Apps like QuizUp and Trivia Crack boast large bases of recurring users. We asked ourselves whether there's a way to generate something valuable out of the time people spend playing trivia games.
• Project summary. We use market research questions to generate trivia games that users can play. We utilize the time that users spend playing trivia games in order to increase survey competition rates. It's mutually beneficial for both groups; users receive fun trivia games, and market researchers receive useful feedback. The best part is that we're able to generate trivia questions automatically without any human involvement, making our solution scalable to a large set of market research questions.

What it does

1. A market researcher visits our website and creates a new market research question "What are 2-3 words that come to mind when you think of General Motors, Microsoft, VMware AirWatch, and Google?"

2. Our system extracts named entities from the question, and finds a few other related entities ['General Motors', 'Electro-Motive Diesel', 'Microsoft', 'Xbox', "Macy's", "Wanamaker's", "Google", "Android (operating system)"]

3. For each named entity, our system visits its corresponding Wikipedia article, and automatically generates several trivia questions from the article's content. "Did Popular Science name Google Now the 'Innovation of the Year' for 2012?"

4. The market research question and the trivia questions are packaged as a game that users can play.

How we built it

Languages

• Python
• HTML/CSS/JavaScript
• SQL

Frameworks

• Flask
• Polymer

APIs

• Microsoft Entity Linking Intelligence Service

Deployment

• Heroku
• Google App Engine

Other

• Natural language processing for turning Wikipedia sentences into trivia questions

Challenges we ran into

• Generating questions from Wikipedia. This was the core functionality of the app, and we benefitted from earlier work done on this task.
• Integration. Once we had finished our individual components, one hurdle was bringing them all together so that the overall system worked as expected.

Accomplishments that we're proud of

• Question generation and NLP. We learned how to apply Stanford Core NLP and NLTK to the task of question generation.
• Polymer. We figured out how to use Polymer without any prior experience.

What we learned

• NLP: Stanford Core NLP, NLTK
• Microsoft Cognitive Services
• Polymer

What's next

• Get our trivia game in the hands of more users, and improve it based on their feedback

Built With

• flask
• javascript
• polymer
• python
• sql