Hi, 

I am not sure I understand your problem, are you trying to analyse a movie dataset and infer the sentimental response it will evoke on a particular population? If that is the case you don't need to have all subjects reviewing all the movies on the dataset. You can establish some relationships based on the reviews you do have, for instance, say subject A watched and reviewed a comedy film, then you can use his review to assess other comedies (with a certain error) but not dramas. I hope it helps

Hi, I wrote my master's thesis on sentiment analysis, and here are some things that might be of use to you, and some of which apply in general to testing computational linguistics/natural language processing systems:

1) If you are going to use statistical techniques, you need a large training set. Otherwise, some methods such as Maximum Entropy will give you trouble.

2)You need to set aside part of your corpus for testing purposes. That is, you should not touch that in development, otherwise your results will be inaccurate.

3)Ideally, you would select another portion of your corpus to help you improve the sytem on repeated runs, moreso if you are going to combine statistical techniques with rules to catch low-frequency cases.

That said: having many reviews all of the same movies is not necessary. It is probably more important to have reviews that vary more in terms of positive/negative.

Best,

Fiorella

Hi, I am pursing my research in Natural language processing and in the field of Sentiment analysis. Mostly when sentiment analysis is carried out, we have to do it for a data set and finally we have to compare our obtained result with any paper previously published to show that the result obtained in  proposed approach is better in compare to the previous paper.  So, I suggest you to go for movie review data set instead of a specific movie review.

@Abinash Tripathy, @Angel Mario Castro Martínez , @Fiorella Dotti..

Thank you so much for your reply. It really helped me.

Hi,

    As I understand you have the dataset which contains reviews for more than one movies (is it true). If so then producing generalized summary will leads to wrong results. What you can do is to identify reviews for movie A and generate a summary for that and then for movie B and so on. By doing this there will be different summary for each movie. There could be another way i.e. you can find the sentiments on some common aspect of different movies and generate a generalized summary of all the reviews which contains movie A, B and so on. By doing this you can summarized that how many users have liked the picture quality and how many disliked the sound quality but I think the first one is the best approach.

Hope it will help. 

@Toqir Rana,

Thank you so much