Hello friend

Greetings.

For sentiment analysis and text mining, you can start from machine learning algorithms. You will find plenty of literature in this direction. I have worked on the various types of machine learning algorithms and their development. You may see some papers to know more about it.

Also you will find so many types of text mining tools available freely online. You can better start from some tools.

I hope this helps you. 

Best regards

Dr. Indrajit Mandal

I am also working on similar type of problem.

As Prof Indrajit mentioned Machine learning is good.

You might want to look at semantic similarity for feature extraction or clustering:

This paper describes the first algorithm specifically created to measure semantic similarity between short texts (sentence length):

Self-archived version:

https://semanticsimilarity.files.wordpress.com/2011/09/10-1109tkde-2006-130-authors-accepted-version.pdf

Li, Y. McLean, D. Bandar, Z. O’Shea, J. Crockett, K.

Sentence Similarity Based on Semantic Nets and Corpus Statistics.

IEEE Transactions on Knowledge and Data Engineering. vol. 18, no. 8, pp. 1138-1150, 2006. (ISSN: 1041-4347).

The final publication is available at:

http://ieeexplore.ieee.org/xpl/freeabs_all.jsp?arnumber=1644735

If you want some ideas on evaluating your algorithms towards the end of your work, you may find some useful ideas in :

http://dl.acm.org/citation.cfm?id=2537046&CFID=476939562&CFTOKEN=19412262

(Free Gold Open Acces download straight from the ACM journal)

Hi,

      there is alot of approaches to  solve text mining problem. some problem have corresponding concrete algorithm. i am actively working these days via different tool like weka, rapidminer and R. Please describe concrete objective and i will suggest you best algorithm which is avaliable according to your scenario!

Regards

Muddsair

Hi, 

You should first get the data set and preprocess it according to your need. Decide about a good machine learning algorithm, construct your architecture and then you can proceed. This is what most of the algorithms do in text mining. 

You should look at these tools mentioned by Mr.Muddsair Sharif and also read about different NLP techniques you can incorporate, and for NLP too good APIs and tools are there like NLTK, Stanford OpenNLP, POS Tagger etc...

good luck .

Hi,

You have firstly to exract from texts of your corpora the representative terms, words or concepts. Secondly, you have to lookfor the correlation between them. You have statistical technics or other algorithms based on formal concept analysis to cluster documents according to their same representants (index).

Please read these papers :

http://link.springer.com/chapter/10.1007%2F11528784_9#page-1

or this one

http://search.fub.it/claudio/pdf/ICFCA03.pdf

I hope that these links help you.

Hi,

    As you have mentioned that you want to detect sentiments from financial text. I think you have to look into Dr. khurshid Ahmad's profile who is professor in University of Dublin, Ireland and has a very good research work on detecting sentiments in financial market.

Basically there are three main problems in sentiment analysis (1) Aspect and opinion word extraction and grouping synonyms (2) Sentiment Lexicon Analysis and (3) Summarizing the aspects. If you are interested in products, hotel, movies, etc. reviews, there is extensive works available better to see Dr. Bing Liu's profile from University of Illinois, USA and Dr. Erik Cambria's from Nanyang Technological University, Singapore. Also visit the site sentic.net.   

Hope that helps.

You can check these paper.

https://www.researchgate.net/publication/270579965_A_Framework_of_preparing_corpora_from_Social_Network_sites_for_Sentiment_Analysis

https://www.researchgate.net/publication/266560877_Corpora_Preparation_and_Stopword_List_Generation_for_Arabic_data_in_Social_Network

https://www.researchgate.net/publication/266560877_Corpora_Preparation_and_Stopword_List_Generation_for_Arabic_data_in_Social_Network

https://www.researchgate.net/publication/261875740_Sentiment_Analysis_Algorithms_and_Applications_A_Survey

Conference Paper A Framework of preparing corpora from Social Network sites f...

Article Corpora Preparation and Stopword List Generation for Arabic ...

Article Sentiment Analysis Algorithms and Applications: A Survey

in order to create a mining system, you need first to work on text analysis process start of 1. tokenization of sentiments reviews into sentences . 2. spliting with delimeters into several words. 3. delete check list (is, are, the,..) 4. cure the problem of segmentation with (ignore or correct)==> example: n't==> not, but (goodwork)==> good work.

with respect to the interpertation of information into score called "sentiment analysis", you will face a problem in evaluation algorithms: you have several algorithm with advantges and disadvantge

1. bag of word,

2. part of speech

3. semnatic neural network

there are one or more but not common algorithms

also you need to check with supervised or unsupervised algorithms to determine the score

I recommend bag-of-words model or it's enhacement or partofspeech algorithms

https://www.researchgate.net/publication/292943899_Enhancement_Bag-of-Words_Model_for_Solving_the_Challenges_of_Sentiment_Analysis

I recommend you read, this survey to help you in determine how to improve your work

http://www.sciencedirect.com/science/article/pii/S1018363916300071  

also prof Bing liu has several professional books and articles in this field and datasets

http://www.cs.uic.edu/~liub/FBS/SentimentAnalysis-and-OpinionMining.html

Article Enhancement Bag-of-Words Model for Solving the Challenges of...