For information retrieval,

You find in this thesis, many experiments comparing the effect of some Weighting Schemes and natural language processing according to short and long queries.

Hope it will help.

Good luck

Thesis Information Retrieval: Applications to English and Arabic Documents

Dear Mosab,

For information retrieval: Bi-gram and Statistical co-occurrence and for sentiment analysis lexicon based opinion classification and summarization. I think the following link might helps to have related papers:

https://www.researchgate.net/profile/Tulu_Tilahun

Regards

Tulu Tilahun

Lecturer at Arba Minch University

Ethiopia

Hi Mosab,

for short text you can first clean you text remove the stop words does the stemming and then build the term to document matrix, this process could be implement in few lines of code (R or other) , from the term to document you can build your metrics such as TF-IDF or other based on your context.  

Then As Tulu mentioned you can build the n-gram and later the chains (HM) if you need to do some prediction. 

One point to keep in mind will be the number of documents, if you have one enormous amount you will have to work with Nosql Db in the back end, such as Hadoop for example to do the preprocess.

Hope this help a little.

Alain

Dear Fadoua,

I think that the best model for short text  is the vector space model

Read : http://www.csee.umbc.edu/~ian/irF02/lectures/07Models-VSM.pdf

Best regards

Dear Mohamed,

Please, can you specify according to the file attached why you consider that VSM is the best?

Don’t forget that there are some common advantages of LSA and VSM.

Regards,