Tweet Classifier
- Split the tweets into sentences
- Tag them with a PoS tagger provided
- Gather some feature information from each tweet
- Use this information in machine learning classifiers that
estimate the Twitter account (i.e., the twit) that produced a given tweet
Statistical Machine Translation
- Build bigram and unigram models of English and French, to smooth their probabilities using add-δ discounting
- Build a world-alignment model between English and French using the IBM-1 model
- Use these models to translate French sentences into English with a decoder
Speech Recognition
- Experiment with speaker identification by training mixtures of Gaussians to the acoustic characteristics of individual speakers
- Identify unknown speakers based on these models.
- Experiment with speech recognition by training hidden Markov models, and analyzing system’s performance with word-error rates