TensorFlow tutorial using character-level LSTM for sentiment analysis on Twitter data. Charles Ashby goes in depth into how to implement a char-CNN network and how to use long-short term memory networks in Tensorflow (he also explains how to expend the model to bi-directional LSTMs). While the subject is slightly advanced, the approach the author take is quite accessible to anyone with a very limited background in Deep Learning and TensorFlow. The author also provides a pre-trained LSTM model ready to be used. He also briefly explain how to use the Stanford CoreNLP Server to do anaphora/coreference resolution as well as named-entity clustering for analysing the sentiment towards different entities in a sentence or even in a document such as in news articles.