title: shouldinvent description: “” category: tags: 
My new side project - ShouldInvent is online, which is basically an automatic idea aggregation web application based on Twitter Search. The primary motivation is to provide inspirations for engineers in various domains.
After thought about it for a minutes, I asked myself - how about create an app that search Twitter periodically for “should invented” things? Can great ideas be mined from such informally mentioned things?
The algorithm behind the application is illustrated in the following figure.
The application works like this:
- periodically, a bot is called to search Twitter using keywords “should invent” (also “should develop”);
- naive natural language processing techniques like tokenizing and chunking is then performed on the content of tweets to identify the target objects;
- finally, each of the target objects are automatic classified into hierarchical categories.
Based on tweets collected in the last month, the “most desired inventions” on Twitter are:
- mirror [4899 tweets] a mirror that takes pictures
- waterproof ipod [414 tweets] a waterproof ipod
- snooze button [366 tweets] a snooze button that hits back
- food machine [223 tweets] a food machine where when you ask for something
- version of twitter [89 tweets] a version of twitter for people who spell things like rappers
The current named entry recognition and classification algorithm are by no means perfect and still need to be improved over time. On the other hand, the UI need improvement as I am not a competent designer. But at least it appears to work.
Go to the website and happy “steal” ideas from Twitter!
Any feedback are appreciated and welcomed.