Sentiment analysis

Sentiment analysis is a means of determining what people think of a product, service, social topic, or political initiative. Sometimes known as opinion mining, sentiment analysis is the process of studying such things as customer conversations, electronic communications, social media posts, publicly shared reviews and opinions, and other sources to gauge people's opinions and reactions. This information is then used by companies, organizations, politicians, and others to make decisions related to the subject of their study. It can be a useful tool for evaluating customer service employees, determining the public's perception of a brand or product, identifying factors that upset or annoy customers or constituents, and even predicting such diverse happenings as financial trends and terrorist activities.

Background

Any poll or survey has at its heart an effort to learn people's opinions about something. It is likely that people have been trying to discern other people's opinions from their written and oral communications for as long as communication has existed. However, it was in the first half of twentieth century that the earliest questionnaires were designed specifically for the purpose of learning what people thought about a product or other topic. Public Opinion Quarterly, the first scientific journal dedicated to opinion analysis, began publication in 1937.

Opinion mining efforts grew steadily throughout World War II (1939–1945) as governments worked to gauge people's opinion of the war in order to gain much-needed support for rationing and funding efforts. It was around this time that academic study of these efforts also began, as researchers sought to find better ways to measure opinions. However, it was the growth of the internet and the ability to gather large amounts of information from varied areas that really boosted the use of sentiment analysis in the late twentieth and early twenty-first centuries.

The development of algorithms that allowed researchers to interpret the frequency of various words as well as search engines and computer programs that could filter vast amounts of information to help with this analysis made sentiment analysis relatively quick and easy to accomplish. In contemporary times, companies and organizations can use these resources to monitor many types of communications and determine what people are saying and thinking about virtually anything. There programs have become so accessible that there are even versions that can tell individuals what words and pictures they use most frequently on social media sites.

Overview

Sentiment analysis examines the attitudes, opinions, emotions, and feelings of people about specific items or topics. This helps companies and organizations to discover and understand aspects that would otherwise be very difficult to learn. For example, it is relatively easy to get people who buy a product to share the things they like about it, but far harder to get people who do not buy something to explain why they chose not to make the purchase.

Part of the reason this is difficult is that these decisions are often not based fully on facts. People's perception of the product or service, how they were treated by a representative, how long they had to wait for service, even the background music on a company's commercial can affect people's feelings and opinions about a company or organization. This can have an impact on a person's decision to do business with a company, support a cause, or take some other action.

Companies and organizations often use sentiment analysis to determine how well a customer service team is performing. By listening to calls and determining a caller's satisfaction from not just their words but also their vocal tone, it is possible to mine information about such factors as how the roll out of a new product or service is going, what areas of a company or product are causing the most satisfaction or frustration, and how well the company's representatives are doing in turning around complaint calls to create satisfied customers.

The technique can also be used to scan vast amounts of information on the internet to look for comments about a product, company, or other area of interest. Social media is a frequent target for these types of searches. It is this type of sentiment analysis that allows companies or groups to gather information on why people do not purchase a product or service. For instance, someone who will not voluntarily comment on a company website or survey about how cheaply they think a product is made may comment when a friend or family member asks for opinions on a social media site.

Another use for sentiment analysis is assessing a population's opinion about a political or social issue. Candidates in contemporary political elections will often employ sentiment analysis to find out what areas are of the greatest interest and use this information to shape political platforms. Marketers use sentiment analysis to assess the success of advertising campaigns. Companies and candidates have been known to discontinue or alter campaigns in response to social media comments or complaints about their ads.

While many of these uses are helpful, at least one frequent use of sentiment analysis and opinion mining can be lifesaving. Experts have developed ways to assess internet and other electronic forms of communications that identify patterns indicating potential terrorist activity. By analyzing this information, government officials can sometimes identify possible threats or pending attacks.

The analysis can be accomplished using technology, including some forms of artificial intelligence. It can also be done by humans, or by a combination of humans and technology. This approach often works best because there are some inherent weaknesses in most computer technology used for sentiment analysis.

For instance, computers have trouble rendering correct interpretations when opinions are stated in the negative. If a person reviewing the quality of a new car reports it was "not bad at all," the programming may identify the word bad and turn a favorable assessment into a negative. It is also difficult to create a program that can understand sarcasm, so if an angry customer states, "Oh, that is just great!" in response to hearing it will take a month to repair an item, a computer program will likely count it as a favorable response. These programs also have a difficult time identifying "flaming," in which one person or many people launch an unsubstantiated attack against someone or something for purely personal reasons.

Sentiment analysis has applications in other areas as well. Sociologists can identify and analyze trends, psychologists can determine a population's mood after significant events such as elections or natural disasters, and law enforcement can use it to learn more about crime patterns. Experts expect its use to expand throughout the twenty-first century.

Bibliography

Bannister, Kristian. "Understanding Sentiment Analysis: What It Is and Why It's Used." Brandwatch, 26 Jan. 2015, www.brandwatch.com/blog/understanding-sentiment-analysis/. Accessed 7 Dec. 2017.

DeVault, Gigi. "What Is Sentiment Analysis?" The Balance, 13 Apr. 2017, www.thebalance.com/what-is-sentiment-analysis-2296941. Accessed 7 Dec. 2017.

Hughes, Lance. "What Is Sentiment Analysis and How Can Machine Learning Help Customers?" Concur, 30 Oct. 2017, www.concur.com/newsroom/article/machine-learning-with-heart-how-sentiment-analysis-can-help-your. Accessed 7 Dec. 2017.

"Introduction to Sentiment Analysis." Language and Communications Technologies, lct-master.org/files/MullenSentimentCourseSlides.pdf. Accessed 7 Dec. 2017.

Mäntylä, Mika V., et al. "The Evolution of Sentiment Analysis—A Review of Research Topics, Venues, and Top Cited Papers." Computer Science Review, vol. 27,Feb. 2018, pp. 16–32.

Medhat, Walaa, et al. "Sentiment Analysis Algorithms and Applications: A Survey." Ain Shams Engineering Journal, vol. 5, no. 4, Dec. 2014, pp.1093–1113.

"Sentiment Analysis." Techopedia, www.techopedia.com/definition/29695/sentiment-analysis. Accessed 7 Dec. 2017.

Stanley, Robert. "What Is Sentiment Analysis? Examples, Best Practices, and More." Callminer, 16 Mar. 2017, callminer.com/blog/sentiment-analysis-examples-best-practices/. Accessed 7 Dec. 2017.