A novel segment-based event detection system for tweets.

• install MongoDB on gpu server
• install the tweets data in to mongodb

• get the alpha result just use the microsoft n-gram
• install the tweets data into mongodb
• get the microsoft n-gram user token
• add the length normalizaion factor
• multi-thread to do the computation and http request

• the tokenizer just use str.split is not flexible, try to find a good tokenizer for tweet