Dear Noah,

Here are some papers with a short summary of their best feature sets for opinion mining + their references.

Pang et al. (2002) reported that unigrams outperform bigrams when performing the sentiment classification of movie reviews,

Dave et al. (2003) show that bigrams and trigrams worked better from unigrams for the product-review polarity classification.

Pak and Paroubek (2010) showed that their classifier is able to determine positive, negative and neutral sentiments of documents. Their classifier is based on the multinomial Na¨ıve Bayes classifier that uses N-gram and POS-tags as features.

Full references:

Kushal Dave, Steve Lawrence, and David M. Pennock.

2003. Mining the peanut gallery: opinion extraction

and semantic classification of product reviews. In WWW

’03: Proceedings of the 12th international conference on

World Wide Web, pages 519–528, New York, NY, USA.

ACM.

Bo Pang, Lillian Lee, and Shivakumar Vaithyanathan.

2002. Thumbs up? sentiment classification using machine

learning techniques. In Proceedings of the Conference

on Empirical Methods in Natural Language Processing

(EMNLP), pages 79–86.

Pak, A., & Paroubek, P. (2010, May). Twitter as a Corpus for Sentiment Analysis and Opinion Mining. In LREC (Vol. 10, pp. 1320-1326).‏

Best regards,

Yaakov

Hi Noha,

                are you interested in extraction minimization or abbstructive technique for text extraction! After selection it, i will further guide you.

You are welcome if you will have any more concern!

Regards

Muddsair

Dear Noah,

Here are some papers with a short summary of their best feature sets for opinion mining + their references.

Pang et al. (2002) reported that unigrams outperform bigrams when performing the sentiment classification of movie reviews,

Dave et al. (2003) show that bigrams and trigrams worked better from unigrams for the product-review polarity classification.

Pak and Paroubek (2010) showed that their classifier is able to determine positive, negative and neutral sentiments of documents. Their classifier is based on the multinomial Na¨ıve Bayes classifier that uses N-gram and POS-tags as features.

Full references:

Kushal Dave, Steve Lawrence, and David M. Pennock.

2003. Mining the peanut gallery: opinion extraction

and semantic classification of product reviews. In WWW

’03: Proceedings of the 12th international conference on

World Wide Web, pages 519–528, New York, NY, USA.

ACM.

Bo Pang, Lillian Lee, and Shivakumar Vaithyanathan.

2002. Thumbs up? sentiment classification using machine

learning techniques. In Proceedings of the Conference

on Empirical Methods in Natural Language Processing

(EMNLP), pages 79–86.

Pak, A., & Paroubek, P. (2010, May). Twitter as a Corpus for Sentiment Analysis and Opinion Mining. In LREC (Vol. 10, pp. 1320-1326).‏

Best regards,

Yaakov

Thanks a lot Yaakov!

I will check the papers you listed.

Best Wishes,

Noha

Hi Muddsair Sharif,

I am interested in exploring techniques of extracting aspects (features) from online reviews. My research interest is in mining online reviews (opinion mining).

If you come across any good paper in this area, please forward it to me.

Thank in advance for your help,

--Noha 

Dear Noha,

An easy but not very sophisticated method is to use for example GATE for POS annotation and use only certain types of words for example nouns could be a good indicators of aspects a little bit more complicated way how to extract hidden aspects is for example LDA where you can get the hidden topics of the review.

Best

Peter

This a little further down the line from where you are now, but when you are interested in complex sentiments expressed in short text fragments of up to 20 words, rather than single words or very short terms, you may want to use semantic similarity:                                                                                       https://semanticsimilarity.files.wordpress.com/2011/09/10-1109tkde-2006-130-authors-accepted-version.pdf (typeset version on IEEE journal website). Also, a surprisingly useful amount of information may be available from the function words in the text, see                                                                                                                                                                                                      https://semanticsimilarity.files.wordpress.com/2011/09/using-a-slim-function-word-classifier-to-recognise-instruction-dialogue-acts-10-1007978-3-642-22000-5_4.pdf                                                                                                                                                              for a highly computationally efficient technique for using function word features (computational efficiency for high volumes of data is often overlooked).

For opinion mining I would be interested in positive / negative aspect of a text and modification of strength using adjectives and adverbs. Unfortunately adverbs are the least studied and understood parts of speech.

Hi,

   As you have mentioned, you are interested in feature/aspect extraction from online reviews. Basically, there are two type of aspects, explicit and implicit. There is extensive research work on explicit aspect extraction but very few work on implicit aspect extraction as it is a very hard task to do. For explicit aspect extraction, topic modelling has shown promising results see Dr. Bing Liu's profile from University of Illinois, USA for latest research. For implicit aspect extraction there is no single technique but combination of different techniques. You can visit sentic.net site, there are number of papers on it. Also look into the following paper which not only extracts explicit aspects but also implicit aspects.

 S. Poria, E. Cambria, L.-W. Ku, C. Gui, and A. Gelbukh. A rule-based approach to aspect extraction from product reviews. In: COLING, Dublin (2014)

Hope that will work. If need any other help you can ask as I am currently doing research in the same domain. 

Thanks Toqir! The paper you attached proposed pretty much different approach than others. It seems so interesting due to it's simplicity and tackling both problems (explicit & implicit aspects)! Also, it has a link to a dataset of opinion reviews that I was seeking for. Thanks again for your help.

Hi,

I believe each method has its pros and its cons.Moreover, there is a major difference between summarization and abstraction.Bing Lu has a good book that reviews most of the work.

So, the method will depend on the task, the data that you have and your objective.

For example, if the dataset has grammatical errors, then rule-based methods might be questionable.

Generally, unsupervised methods are usually robust, but does not perform as well as supervised methods. The Semi-supervised methods usually perform well.

The aspect-based sentiment is still a hot research area. Semantic Evaluation (SemEval) is an ongoing series of evaluations of computational semantic analysis systems that has a special task for feature extraction every year (Pontiki et al., 2014).

some examples are below.

Qiu et al. (2009) proposed a double propagation approach based on dependency grammar of (Tesnière & Fourquet, 1959).  Recently, an association between features and opinions using Latent Semantic Analysis (LSA) and Likelihood Ratio Test (LRT) has been employed to find frequent features (Hai et al., 2012).

Pavlopoulos and Androutsopoulos (2014) extended Hu and Liu method with an additional pruning step. They used continuous vector space representation (Mikolov et al., 2013) to prune words for domain and common language spaces.

Join us in Granada at SAMSN 2019

http://emergingtechnet.org/SAMSN2019/