Twitter Sentiment

I am in need of a Twitter algorithm to measure how emotionally loaded tweets are. The algorithm needs to search Twitter and create a list of users who mention certain products (=keywords). I want this search to be automated so the list of users continues to grow. The system should automatically download the tweet history of each extracted user. We use a sample of these tweets to analyze what mood swings these users encounter when they tweet. The technology must have a filter function, so we can categorize tweets and their emotional content depending on keywords used. The idea is that it measures how people feel when they mention certain products. Each tweet should get a sentiment score based on the language used. We need to discuss how many mood dimensions we should focus on (depends on analysis of sample tweets). The algorithm needs to be able to handle thousands of tweets a minute providing aggregate sentiment data for different time frames and display them in a real-time database. When certain criteria are met (unusual volume, unusual emotions etc), I want the algo to send out alerts. The criteria which trigger alerts need to be adjustable. It should be an intelligent algorithm which is able to teach itself and learn over time. Obviously the smarter and more accurate it is, the better. Once a working prototype is developed, I would like the programmer to shoot a short video (1 – 3 minutes) to demonstrate how the algorithm works. A screencast with a verbal description of the technology is sufficient. I will need someone to maintain the technology after completion, so if you are interested, please let me know how many hours a week you could be available under what conditions. Freelancer must be willing to sign NDA. Thank you. I look forward to your proposals.