2-2. Previous Research

　　　Likewise, many researchers confirm that the 2003 thesis submitted by Nasukawa and Yi (Nasukawa & Jeonghee) used the term “Sentimental analysis” first.[1] In the report, they show how to extract the sentiments and divide them into “Positive” and “Negative”. The survey, unlike the previous two examples, treats many sentences of various domains from news article to camera reviews on the web pages in order to find out the way to pave the new linguistic approach. In the last part of the paper, they concluded that the more difficult the target documents, which are news articles and descriptions in some official organizational Web pages, becomes, the lower the precision rate goes. The main method which this paper used for sentimental analysis on texts is A Part-Of-Speech Tagger (POS Tagger). POS Tagger “is a piece of software that reads text in some language and assigns parts of speech to each word (and other token), such as noun, verb, adjective, etc.”[2] It seems that this technology has widely adapted by several cognitive services or applications.

It would also be noteworthy that both authors of this report are IBM’s staff. Intriguingly, one of the previous thesis’s author “Shivakumar Vaithyanathan” is also a staff from IBM research center. Those people’s untiring efforts would be one of the reasons that IBM has recently been regarded as a leading company in the cognitive system due to the creation of the revolutionary computer system “Watson” which has offered many APIs including machine learning techniques to the companies and concerned individuals. Among the features, what IBM Watson provides is “AlchemyLanguage”, which enables us to analyze text and help us to understand its sentiment, keywords, entities, high-level concepts and more. Probably, the work by Nasukawa and Yi and other pioneers’ feats greatly contributed to the progress of the “AlchemyLanguage”.

Title is a crucial part for news articles and books which need to attract potential readers’ interest. This short sentence has also been a hot field which many researchers have devoted themselves to. According to Charles Grivel, title functions can be divided into three categories; “to identify the work”, “to designate its topic” and “to make the book stand out.” There seem to be general acceptance of the sorting for title functions (Genette, 1997, p. 69). On the other hand, some people raised an objection his theory. For instance, Gerard Genette, who is a French literary theorist, insisted that the Grivel’s classifications for title are “not necessarily all fulfilled at the same time”. Furthermore, he added that even though the first one is a must, the other two are “optional or supplementary (Genette, 1997, p. 69).” In addition, Genette asserted that the role of title should be parted into “thematic” and “formal”. These days, the combination of Text analysis and Sentimental analysis has led to new discoveries. For one thing, it is found that negative news headlines get much more reader’s attention, which is the third title function Grivel invented, than positive ones (“Psychology: Why bad”, 2014). This surprising fact hidden under our feet was revealed thanks to the technical collaboration. Also, this partnership is applied to assess the media trend over the certain social issues. The paper led by Erik Bleich (Bleich, Nisar, and Abdelhamid, 2015) successfully uncovered the unique media tendency of the news coverages on Islam and Muslims by conducting Sentimental investigation on New York times’ news headlines related to the topics. This survey greatly contributes to this research process. In fact, some of the method this paper use are referred from Bleich’s study.  

The rapid increase Internet diffusion rate enabled researchers to collect massive data required for calculate social phenomenon. The technics has now been used not only by scholars who hope to verify their theories, but also by companies which want to monitor the customers’ feelings or complains about their products more accurately. In recent years, the heated demands for analysis on people’s latent minds and the accumulated legacies incredibly accelerated technological innovation. As a result, it has hugely broaden the usage of Sentimental analysis from finance to politics.

Financial themes have always been a popular subject in “Sentimental analysis”. So far, a numerous number of economic treatises connected to the emotional evaluation have been proposed. It seems that the monetary scholarly field has quickly accepted this new style because the markets have a long history to cherish the miscellaneous data for the prediction of the stock movement. The data is called “Fundamentals”[3] and “Sentimental analysis” has smoothly mingled with them. In fact, as this thesis mentioned earlier, the leading paper for “Sentimental analysis” was about stock prediction using this method (Yahoo! for Amazon, 2004). Since then, there are a lot of theories proposed concerning the evaluation for the market trend.

Johan Bollen and his team’s thesis in 2010 (Bollen & Huina & Xiao-Jun, 2010) is well cited and it achieves more than 2000 times quotations from other studies according to “Google Scholar”. In Bollen’s research, they proved that sentiments of the sentences on Twitter clearly have something to do with the value of the Dow Jones Industrial Average (DJIA). They used two mood tracking tools for their survey; one is “OpinionFinder” which measures positive or negative and the other is “Google-Profile of Mood States (GPOMS) that measures mood in terms of 6 dimensions (Calm, Alert, Sure, Vital, Kind, and Happy)”. Eventually, they reported that the overall accuracy to predict the value of DJIA reached as high as 86.7%.

The utilizations of “Sentimental Analysis” encompass tangled political matters such as “poll survey” that is one of the topics for this thesis. Nowadays, SNS tools are the primary places for the estimation of political sentiments among public mind. Out of many communication services, Twitter is presumably more chosen for the analysis due to its outstanding feature that set the 140-character limitation for each posting on every user. This function enabled researchers to collect and inspect the target materials easier and faster because those texts were much shorter and clearer than the descriptions of conventional blogs or homepages. (However, Twitter announced the abolishment of the system in May, 2016) What is more, according to the research conducted by Tom Jackson and Martin Sykora who are professors and lecturer of Loughborough University, analyzing emotions on Twitter suggested Donald Trump’s victory over Hillary Clinton for “The United States presidential election of 2016” (Jackson & Sykora, 2016). This amazing discovery implies not only the premonition for the new dawn of the outmoded classic polling system but also the possibility for the innovation of other long-established customs. Traditional survey theories for voters’ phycology status like “Bando wagon” and “Underdog effect” might be able to be more sophisticated by cooperating with “Sentimental Analysis”