Analysis for Social Media: A Survey
Radhe Shyam Yadav
In the recent
past years, the world wide web (www) has become the large source of data and opinionative
information. People uses social media such as twitter, Facebook, hi5 etc. to
express their view, sentiment and opinion on daily basis. As this source is
going to be one of the huge source of information for any kind of decision
making decision. To automate the examination of such information, the territory
of Sentiment Analysis has risen. The aim is to segregate the opinionative data
based on information it carries whether negative or positive. Sentiment
analysis is a problem of text-based analysis, yet there are a few difficulties
that make it troublesome when contrasted with conventional text-based analysis.
It has opened up a few open doors for future research for taking care of negation,
shrouded conclusions recognizable proof, slangs, polysemy.so, this growing data
requires automatic data analysis technique.
is also called as opinion mining or emotion AI, which measure the slant of
individuals’ sentiments through the natural language processing (NLP), computational
linguistics and text analysis. As social has become the big platform for
finding people opinion and their reaction on particular topic. This paper is
about the way to tackle the different types of approaches and their related
challenges. Some of the approaches are as follow:
Approach 1: –
The first approach is user topic opinion prediction. The technique used for
this approach is Social
context and Topical context incorporated Matrix Factorization (ScTcMF) to
predict the opinion of people on unknown topics using the data scope as twitter
and data sources as tweet.
Approach 2: -The
second approach is polarity shift in sentiment classification using dual
sentiment analysis (DSA) to check the polarity classification task using
multiple domain sentiments English and Chinese dataset and amazon.com as data
source. Later on, DSA framework was extended from polarity to 3 class(positive-negative-neutral)
classification, considering the neutral feeling as well.
Approach 3: –
The third approach is Hashtag-level sentimental classification using the
Support vector machine classifier. This classifier will help to generating the polarity
of given Hashtag in certain time interval using twitter as data source and
tweets as data.
Approach 4:- The
fourth approach is Sentiment polarity classification and sentiment strength
detection using hybrid technique i.e. Machine learning and lexicon based to
classify polarity and detect sentiment strength using software or movie review
as data score and IDMB,CNET as data sources.
There are so
many challenges in sentimental analysis such as Incremental approach due to
increase in data in real time, parallel computing for massive data for good
efficiency, Behaviour in social media, sarcasm, grammatically incorrect words, Review
author Segmentation will be helpful during decision making etc. There are
massive data share by users on social media to share theirs feeling and opinion
on particular topic. It also helped in getting the review about the product, sales,
celebrities etc. This also observed that most of the work can be done using
Machine learning rather than the lexicon-based method.