The main goal of this project is to build a mean and to populate a database with correct information regarding some decided pairings. This database will then be compared to the results of algorithms that process the same pairings, which provides a way to assess algorithms’ performance. The main focus is set on the following pairing, tweet messages and company. For a given set of tweet messages, we should be able to decide if either the word refers to the given company or not. In order to populate the database with such information, a web game application is used, which will allows any kind of users who wishes to play to give its feedback. We record those feedbacks and when enough feedbacks are collected, we may be able to build a subset of the correct information we want based on the feedbacks.