What is the sentiment analysis all about?
In a nutshell:
From the image above you will find the philosophy of the sentiment analysis. What follows then in the blog, is the more detailed basis of the sentiment analysis, technology and most important – how it is utilized in social media.
Sentiment analysis is in recent year discovered way of learning more about customer behavior. Increase of social media, the companies have understood that their clientele is out there in the internet sharing their reviews. Sentiment means person’s emotion towards something, sentiment is derived from feeling rather than reason. Through the emotion under the written text, sentiment analysis can discover attitude, opinions and mood of the writer. (Heires 2015)
The sentiment analyze apps evaluate whether the text is positive, negative or neutral. First the analysis were made manually, but now it is automated with the help of technology and analysts. There have been also developments in the text analytic tools that help to receive accurate information. (Heires 2015) Identifying the text either positive or negative is called categorization. Automated sentiment categorization has been utilized for example in product or movie reviews and preview of election result. (Khoo etc. 2012)
Would it be difficult to evaluate the sentiment of this review on the above?
The sentiment categorization within computer and information scientists have been claimed to be too straightforward. Social and communication researchers have been proposing that in sentiment analysis should be taken more factors into account. Within social and communication experts the focus is more in the polarity of positive and negative sentiments. There should be more frameworks that covers also the linguistic, social and psychologic sides of the analysis. (Khoo et al. 2012)
The problem of course is, that no app or computer think like human brain, so it is challenging to integrate emotion nuances to sentiment analysis. That is why sentiment analysis results have been accused to be too “black and white”. Human brains can’t be automated so in order to create efficiency, the meter must be quite simplistic. (Khoo et al. 2012)
Where would you need sentiment analysis?
Increase of social media has been activated customers to share their opinions. With online reviews, many people evaluate whether to purchase or not by other’s reviews. With the help of those reviews in social media, companies can learn about their customers and modify their product towards more attractive. By knowing the clientele, companies will have great competitive advantage. In order to do so, companies need also to be aware of their competitors and constantly collect information about theirs’ products and reviews. (He et al 2016.)
Social media tools give companies countable opportunities to create business. With social media, customers and companies can communicate like never before. Companies have understood the power of social media – and the power of the customers. It is usual, that customers compare the servicescape and share their opinions in social media. It is important that companies develop their social analytics skill in order to quickly respond to customers’ demands. In this environment nowadays, companies need to be able to change their strategies rapidly. (He et al. 2016.)
In social media, sentiment analyse is an automated technique of tracking positive, negative and neutral emotions among customers. With the use of sentiment analysis, companies can collect widely customers or users emotions. Even though the sentiment analysis are being accused for being too simplistic, it gives you quickly information of large groups and the accuracy is good enough. (He et al. 2016.) The biggest social media sites Facebook, Twitter and Youtube have created applications for tracking down data. In recent years, also other operators have launched these sentiment analysis apps such as IBM social media capture. (Avery & Narayanan 2015.)
Especially sentiment analysis in Twitter has been interested companies. For automated applications, Twitter is easy to analyze due to the limited amount of characters. The short tweets simplify the meaning and people tend to write only the main point. The tweets are also available for everybody, unlike in Facebook. (Avery & Narayanan 2015.) There are almost boundless quantity of tweets in Twitter, every SECOND 6000 tweets are tweeted! That gives 500 million tweets per each day. After the first year of Twitter, year 2007, the quantity of tweets per day was 5000. (Internet live stats 2016) This crazy amount of online information about the customers is now being utilized.
To analyze sentiment, there are four dimensions used in general:
1) keyword spotting
2) lexical affinity
3) statistical methods
4) consept-based techniques. (Avery & Narayanan 2015.)
Let’s take the keyword spotting as an example. The analysts come up with 20 negative emotion keywords – including for example anger, confusion and depression. Then the analysts decide 20 positive emotion keywords – including for example happy, great and kind. After that, these words are filtered through large amount of tweets. A database program goes through a large number of tweets and analyze whether the tweet contains positive or negative keyword. The app classified the tweets into positive – negative – or neutral (=don’t have any of the positive or negative keywords given). The problem with this keyword spotting is, that algorithm cannot identify all the words.
“This morning was great.” or “This morning wasn’t great at all.”
That is why using different approaches aka dimensions, is the most accurate way for collecting information. (Avery & Narayanan 2015.)
Sentiment Analysis of Social Media texts lecture on Youtube:
Click the image to watch Sentiment Analysis of Social Media Text on Youtube. Link is also here. (Mohammad & Zhu 2014.)
Avery, Heather; Narayanan N.H. 2015. Sentiment Analysis Solutions for Insurance Business Decision Support. Industrial and Systems Engineering Conference. Proceedings. (2015): 527-535. Referred 18.10.2016 <http://search.proquest.com.ezproxy.utu.fi:2048/abicomplete/docview/1791989232/fulltextPDF/D7B153DEC91140A6PQ/13?accountid=14774>
Heires, Katherine. Risk Management 62.10. Sentiment Analysis: Are You Feeling Risky? (Dec 2015): 14-15. Referred 12.10.2016 <http://search.proquest.com.ezproxy.utu.fi:2048/abicomplete/docview/1747316076/fulltextPDF/F56CE9100C7844E7PQ/8?accountid=14774>
He, Wu; Tian, Xin; Chen, Yong; Chong, Dazhi. The Journal of Computer Information Systems56.2. Actionable social media competitive analytics for understanding customer experiences. (Winter 2016): 145-155. Referred 18.10.2016 <http://search.proquest.com.ezproxy.utu.fi:2048/abicomplete/docview/1795622769/fulltextPDF/53CC343DF50A4328PQ/8?accountid=14774>
Internet Live Stats. 2016. Twitter statistics. Referred 18.10.2016 <http://www.internetlivestats.com/twitter-statistics/>
Khoo, Christopher Soo-Guan; Nourbakhsh, Armineh; Jin-Cheon, Na. Online information Review 36.6. Sentiment analysis of online news text: a case study of appraisal theory. (2012): 858-878. Referred 13.10.2016 <http://search.proquest.com.ezproxy.utu.fi:2048/abicomplete/docview/1193817321/fulltextPDF/A8CEE7233CAD485FPQ/3?accountid=14774>
Mohammad, Saif, M.; Zhu, Xiaodan. 2014. Sentiment Analysis of Social Media Texts. Youtube. Referred 18.10.2016 <https://www.youtube.com/watch?v=zv16Xyph7Ss>
Pujari, Pradeep. 2013. Slideshare Simple Sentiment Analysis Using Solr. Sentiment Analysis Innovation Summit. Slideshare. Referred 17.10.2016 <http://www.slideshare.net/PradeepPujari/sais-20431863>
Pophangover. 2014. Referred 17.10.2016
Sloan, Anjelica. 2014. Courting your customer. Referred 18.10.2016 <https://blogs.oracle.com/marketingcloud/courting-customer-relationship-advice-marketers>
Zuhora, T. Fatima. 2016. 5+ Best Twitter Extension for Joomla. Themexpert blog. Referred 18.10.2016 <https://www.themexpert.com/blog/best-twitter-extension-for-joomla>