LSTM is good for sequence prediction, because it can remember the previous context. What is the rationale behind using it in classification tasks ? In particular, they have used it for the following task:
The Large Movie Review Dataset (often referred to as the IMDB dataset) contains 25,000 highly-polar movie reviews (good or bad) for training and the same amount again for testing. The problem is to determine whether a given movie review has a positive or negative sentiment.
In other words, can the classification problem be reduced to sequence prediction problem in some way?