Sentiment Analysis 2015

From New Media Business Blog

(Redirected from Sentiment Analysis New)
Jump to: navigation, search

Contents

Big Data

What is Big Data?

Big data, as defined by Tom Davenport, American academic and author specializing in analytics, business process innovation and knowledge management, is “the ability of society to harness information in novel ways to produce useful insights or goods and services of significant value”. Big data also allows “…things… [to be done] at a large scale that cannot be done at a smaller one, to extract new insights or create new forms of value.”[1]

There is so much data available today. More data is produced from our phones, computers, cars, tablets, smartwatches, and other Internet of Things devices, in a minute, than data produced from the beginning of time to the year 2000.[2] With all this data comes the challenges of how to manage and interpret all the information to make decisions. Fortunately, with advancements in sentiment analysis, researchers and other professionals have another tool to interpret the vast amounts of data, specifically textual data.

Sentiment Analysis Overview

Sentiment analysis (SA), also known as opinion mining, uses natural processing language, text analysis and computational linguistics, to extract, identify and categorize a writer’s attitude towards a certain topic, product or service as positive, negative or neutral.[3] In layman’s terms, it allows for algorithms to extract opinions, emotions and attitudes from text from sources such as Twitter, Facebook, news outlets, journals, blogs, comments, reviews and many more.

The importance of sentiment analysis began in the early 2000 and coincides with the rapid growth of social media content on the web.[4] It has continued to grow over the years as the digital global economy has expanded and the voice of consumers has grown stronger. More than ever, corporate and public institutions are paying attention to what their customers are saying about their products, services, organizations, employees, issues, events, topics and their attributes.

There are several reasons why sentiment analysis has become a very important area for research and development. Firstly, SA has a wide range of applications, including business, politics and social sciences. The commercial applications have been a strong force in its development. Secondly, SA offers many challenges that have never been studied before. Thirdly, the rapid growth of social media and opinions derived from these services had allowed for huge amounts of opinionated data. Without these data sets to be practiced on, much of the advancements in SA would not have been possible.

Understanding sentiment can be a challenging task. At the moment, textual information has mainly been used in sentiment analysis; however, with advancements in this technology new forms of information, such as audio and visual content, are being used in sentiment analysis. However, to get audio and visual content to provide the same level of insight as textual sentiment analysis, further advancements in this technology are needed.

Natural Processing Language, Text Analysis & Computational Linguistics

Sentiment analysis would not be possible without the use of computational linguistics, natural processing language and text analysis:

Computational Linguistics

Computational linguistics is a branch of linguistics from a computational perspectives that takes into account and applies statistical or rule-based modeling to analyze and synthesize natural language and speech.[5][6]

Natural Processing Language (NPL)

Natural processing language is an interdisciplinary field comprised of computational linguistics, computer science and artificial intelligence concerned with computer-human interaction.[7][8] The goal of NPL is to find understanding and meaning in natural human language and speech.

Text Analysis

Text analysis, or more recently known as text mining, is a systematic method of analyzing textual content to find patterns and trends to determine the meaning of the natural language.[9] One of the most common applications of text analysis is in sentiment analysis.

Levels of Sentiment Analysis

Sentiment analysis can be investigated in three ways:

Document Level

In document level analysis, SA is used to determine whether a whole document expresses positive, neutral or negative sentiment. This is called document level sentiment classification.[10][11] This level of analysis is only capable on a single entity.

Example: In a movie review, the system would determine whether the reviews expresses a positive, neutral or negative opinion about the movie.

Sentence Level

This level of analysis looks for opinion at the sentence level to determine positive, neutral or negative sentiment.[12][13] This level of analysis is capable of differentiating between two types of sentences, subjective sentences, which express subjective views and opinions, and objective sentences, which express factual information. It should be noted that objective sentences can imply opinions.

Example:

Subjective Sentence - The iPhone is a horrible phone.

Objective Sentence - The hockey game will start at 7:00 pm.

Objective Sentence implying an Opinion - My Samsung S6 phone is new and shuts off after 60 minutes of use.

Entity and Aspect Level

This level of analysis is different than document or sentence level analysis. It is able to perform a finer-grained analysis and determine what people do and do not like about a specific attribute of an entity.[14] Aspect level analysis is based on the idea that an opinion has two parts: sentiment and target. The objective of this type of analysis is to determine sentiments on entities and their attributes. This ability to perform such analysis, turns unstructured text to structured data, and can be used for various types of qualitative and quantitative analyses.[15][16]

Example: Volkswagen’s drive quality is good, but releases too many emissions.

The sentiment on Volkswagen's drive quality is positive, but the sentiment on its release of emissions is negative. The drive quality and release of emissions of the Volkswagen are the opinion targets.

With the current technologies, analysis at a document or sentence level analysis is difficult, however analysis at an aspect level is even more difficult as it requires a deeper level of analysis and recognition of context.

Types of Opinions

Two types of opinions exist to make matters even more difficult, regular opinions and comparative opinions.

Regular Opinions

Regular opinions express opinion on only one specific entity or an attribute of an entity.[17]

Example: The sunset is beautiful.

Comparative Opinions

Comparative opinions compare multiple entities based on a common attribute.[18]

Example: Sony Xperia Z5’s picture quality is better than iPhone 6’s.

History

The early 2000’s mark the explosion for research activity into the problems, solutions and opportunities sentiment analysis provided.[19] The reasons behind this increase in research are:

    • The rise of machine learning methods in natural language processing and information retrieval.[20]
    • The availability of datasets for machine learning algorithms to be trained on, due to the development of the World Wide Web and review-aggregation web-sites.[21]
    • Realization of the fascinating intellectual challenges and commercial and intelligence applications that the area offers, such as business, politics, social sciences and more.


    Current sentiment analysis originated from content analysis which is “analysis to determine the meaning, purpose, or effect of any type of communication, as literature, newspapers, or broadcasts, by studying and evaluating the details, innuendoes, and implications of the content, recurrent themes, etc.”[22] Although the exact date of origin is not known, there are some records that go as far back as the 1890s when content analysis was performed manually in qualitative newspaper analysis.[23] In the 1940’s, content analysis was an often used research method despite lack of technology and human errors that limited its effectiveness.[24] Further, at the time, studies were limited to examining the frequency of the occurrence of specific text. Fortunately, in the 1950’s, technology advanced and researchers were able to apply more sophisticated methods of analysis to focus on concepts and semantic relationships.[25]

    Applications

    Sentiment analysis can be utilized in a wide variety of applications ranging from marketing, politics and finance. With the use of various computer programs and human expertise in linguistics and text analysis, corporations can sift through large amounts of data on social media sites, such as Facebook, Twitter, YouTube, Urbanspoon, blogs, forums, chat rooms, as well as online news outlets, and make sense of all this information. Companies can now more effectively manage their brands and understand consumer opinions to better respond to complaints and measure the success of marketing campaigns.

    Marketing/Marketing Research

    Sentiment analysis and its advancements has provided marketers and market researchers with another tool to analyze textual data without sacrificing on quality, reliability and generalizability. Sentiment analysis has provided solutions to problems that marketers have been facing. There is an incredible amount of user-generated content on the web due to Web 2.0. There are blogs, Facebook, Twitter, Instagram, YouTube, etc. and this is just a fraction of what’s being generated on the web. With all this information, there was not an effective way of analyzing such large sets of textual data in a digestible manner than could provide insightful findings. But with advancements in SA, it’s now easier, and will continuously become easier to make use of all the information on the web. Here are the two biggest problems SA has provided solutions for:

    • The recent explosion of user generated content, mainly through social media platforms, has provided researchers with an overwhelming amount of textual data. Advancements in sentiment analysis has allowed researchers to make use of this data to make insightful findings.
    • Advancements in programs and algorithms have led to “classification accuracy and user friendliness”. No longer are focus groups or in-person surveys essential to gather attitudes and opinions towards a brand or product. As a result, market research activities costs and sampling errors have been reduced, while validity and reliability of research finds have been enhanced.[26]


    The following firms focus on sentiment analysis in the marketing field:

    Bitext

    Bitext

    A multilingual provider of text analytics technology of major European languages located in San Francisco and Spain. Bitext provides services in deep linguistic and sentiment analysis, and text, entity and concept extraction. They provide solutions for market research, big data, social CRM, reputation management and user generated content analysis, and more. They service clients such an Intel, Salesforce, Ernest & Young, Movistar and Repsol. Bitext’s sentiment analysis program is customizable to allows organizations to add their own words and rate its sentiment based on jargon of their specific organization or industry.[1]

    IBM Social Sentiment Index

    IBM Social Sentiment

    IBM’s social sentiment index gauges and aggregates public opinion from a variety of sources, such as social media and news outlets. It provides real-time snapshots, identifies trends and how consumers feel about everything from products, sports to traffic. One aspect that IBM’s social sentiment index is leading in is its ability to detect the difference between sincerity and sarcasm in text. This is one aspect of sentiment analysis that has been difficult to do and will require further advancement to become accurate.[1]

    Finance

    The price of a company’s stock is the result of not only financial metrics, such as revenues, profit, earnings per share, price to equity ratio, etc., but is also determined by investor sentiment towards a company’s stock, or the economy. As a result, sentiment analysis can quantify qualitative information and play a role in predicting the future price movement of an equity.

    Some of the most useful sources of financial information comes from business new outlets and Twitter accounts of investors, hedge fund managers, traders, business news outlets, etc. But with all these sources of information comes the issue of ‘noise’ from the investment community that can provide conflicting information and making trading decisions more difficult. Utilizing complex algorithms, large amounts of data can be sifted through to find the most relevant information. The biggest advantage of sentiment analysis in trading, is the ability to adjust trading strategies instantaneously based on real-time information and analytics, in combination with algorithms. This has shown to be effective in high-frequency trading. Sentiment analysis on stocks can also be used for medium to long-term time horizons.[2]

    Very simply put, Tweets can be assessed by giving a -1 or +1 rating. For example, if the following Tweet was made, “Shorted $RBC”, it is considered negative, thus would be given a -1 sentiment rating. Tweets with more than one subject and attribute can be rated and are appointed a sentiment rating for each subject attribute pair rating. For example, if an investor said, “Shorted $AAPL. $TSLA seems to be positioned for strong growth after Volkswagen fiasco”. In such an scenario, Apple could be given a sentiment score of -0.35 and Tesla a score of +0.55. Thus, the overall sentiment rating would be +0.20 for the Tweet.

    Although sentiment analysis can aid in the decision-making process for stock selection, it should not be relied upon on its own. There have been many academic studies and real-life examples showing the benefits of integrating sentiment analysis into trading strategies; however, there have also been examples of where such strategies have failed. Although there are many skeptics of the role of sentiment analysis in trading, it continues to grow and show its effectiveness in aiding the prediction of stock price movements.

    A variety of firms are involved in text analytics and sentiment analysis for financial markets:

    iSENTIUM

    How One Company Uses Tweets to Gauge Stock Sentiment?

    iSENTIUM is a market leader in real-time sentiment search with their proprietary natural processing language technology, currently supporting over 8,500 US stock indices and exchange traded funds (ETFs). iSENTIUM offers its Daily Directional Indicators on stocks and ETFs where sentiment has proven to be a leading indicator of price movement. Their proprietary system “breaks down sentences into key components, and analyzes adjectives and actions associated with subjects mentioned in Tweets, as well, as their location in sentences, among other factors” and then provides a sentiment score for the stock. There are a few interesting things to note about iSENTIUM. They only use Twitter data, supplied by StockTwits, and apply their proprietary filters to find Tweets about stocks and the stock market, but do not filter out Tweets between novice investors and professionals. Despite including Tweets from novice investors, iSENTIUM still manages to be correct about a stock’s price movement 60% of the time.[1][2]

    StockTwits

    What is StockTwits?

    StockTwits, as mentioned, is a social media platform with more than 250,000 active members. It has a similar interface to Twitter, where instead of hashtags, dollar signs are used in front of a stock ticker. Here, investors and traders share their investment ideas. These comments are then gathered, analyzed and given a sentiment score to give investors a better idea of how people feel about certain stocks.[1]

    Politics

    The dramatic rise of social media since 2005 has seen an important new medium introduced that has allowed politicians to reach wider audiences, and for the general public to be more engaged in political activities. In 2008, Barack Obama made history not only because he became the first African-American President, but he became the first presidential candidate to use social media as an integral part of his campaign strategy to engage voters.[2]

    The microblogging service Twitter, has especially become a popular platform for the general public to use for political deliberation. With analysis showing that sentiment profiles of politicians and parties reflect those of the election campaigns, Twitter can be considered a valid indicator of political opinion. The sentiment analysis of political tweets is also closely reflected in the poll results, however, it should be noted that this is currently being looked at as a supplement to traditional election forecasting, since it is not a definitive indication of future election results.[3] Analysis of Twitter has also become an important tool for political parties to use when gauging sentiment of the general public. It has become an excellent method to gather sentiment on specific political news events and the sentiment generally, accurately represents consumer confidence.[4][5]. In addition, real-time analysis is providing parties with changes in sentiment of the general public as political events and news unfold, allowing them to modify, and adjust their strategy on the go.[6]

    Understanding Employees

    An organization’s understanding of its employees and what motivates and demotivates them can have a huge impact on the overall performance of the organization. Sentiment analysis is now being used to gauge employee morale by monitoring employee emails and intranets. The one downfall of this is that employees may feel like they are being watched by ‘Big Brother’. In recent past, companies have been using text and sentiment analytics software to gather information about employees and their sentiment. The overall goal of using such technologies is to assess “where employees are dissatisfied and design strategies for enhancing engagement and in turn, improving productivity, employee retention and customer service. To be able to conduct this is an easy process, saves time and removes the need for HR to prepare surveys. Rather, through these analytic programs, information is automatically gathered and reports are prepared without taking up staff time. Jeff Catlin, CEO of Lexalytics stated, “The ultimate purpose is not to keep tabs on individuals, but rather to assess attitudes reflected by multiple employees, and in turn use the data in promoting positive activities”. Elizabeth Charnick, CEO of Cataphora, a firm specializing in analytics software, said, “If the majority of employees really disapprove of something, it is to the advantage of both the company and the employees for management to be aware of that… No one is singled out”. Implementing text and sentiment analytics has the potential to improve employee motivation and company performance as long as it respects the privacy of employees.[7]

    Sentiment Analysis Software Vendors

    The following two sentiment and text analytics vendors are some of the leading firms in the industry.

    Lexalytics - Semantria

    Lexalytics has over 10 years of experience in sentiment analysis of people, companies and products. Semantria, founded in 2011, also focused on sentiment analysis and was one of the fastest growing sentiment analysis firms in the world focusing on retail, market research, consumer reviews and social media marketing applications. In 2014, Lexalytics bought Semantria to create Lexalytics Semantria (Semantria). Semantria is capable of detecting sentiment on several levels, from emojis, hashtags, sentences, documents, etc. Simply put, Sematria works by first identifying emotive phrases within a document. Then, the program scores the various phrases it finds and, finally, combines all the scores to determine the overall sentiment of the piece of writing.[8][9]

    Semantria is a very flexible and customizable tool to meet a company’s need. Words can be added and removed from the system, and ratings can be adjusted for each word to best score the jargon used in the company, or industry they operate in. Semantria can be used via API (application programming interface) or Excel for further analysis of results. In addition, there is a web demo available to users for free that can calculate sentiment for any piece of writing.

    Sentdex

    Sentdex is another well known company in the text and sentiment analytics field. Sentdex primarily focuses on the analysis of financial, political and geographical sentiment. Sentdex utilizes two methods of analysis. The first method, ‘Bag of Words’ model, focuses on words, or a string or words, and does not pay attention to the context, which is a very basic level of analysis. Each word, or string of words, is given its own value, the values are added up and a total sentiment score is given. The second method uses natural language processing to understand the context of a piece of writing. Rather than just focusing on a word, or string of words, it focuses on the succession of a string of words. This method allows the algorithm to “understand… grammar principles… tag parts of speech, name entities, and more, in order to actually understand the ‘language’ of the text”.[10]

    Challenges & Limitations

    Social Media Selection Bias

    While analysis of Twitter has provided fairly accurate sentiment, one must consider that this is not fully representative of all public sentiment. In fact, a recent report by the Pew Research Center indicates that, while 65% of the entire adult population of the US use some form of social media, only 23 % of those users use Twitter.[11][12] Breaking this down even further, it shows the greatest number of users are between 18-29, while the next age group using twitter is the 30-49 demographic. This is not taking into account other contributing factors such as race, education level, gender, income level or where in the country they reside. As such, this is an important consideration when considering the results of any sort of sentiment analysis performed on social media data and that the results might not necessarily truly reflect the proper sentiment of the entire population.

    Subjective Classification of Words and Context

    Sentiment Analysis relies heavily on how it is trained and the techniques used to implement it. This is especially important when it comes to classifying words, text and the context it is being used in. For example, “sick” can have different sentiment depending on which context it is being used in. If spoken between skateboarding friends and someone just performed an exciting new trick, it could be considered a “sick trick”, indicating it was a trick that was incredible and exciting. If however, “sick” was used in a medical environment, there is little doubt it will be considered something negative.[13] This subjective context can be extended to phrases as well, and leads to the same problems. Classification is an important aspect of sentiment analysis that is much more complex than simply determining if a word or phrase is positive, negative or neutral.[14]

    Language Nuances

    English is a complicated language that has many nuances related to it. Sarcasm, multiple meanings, order of words, idioms, homophones, words that have no direct English translation are all characteristics of the English language.[15] Further complications arise when emphasis placed on a different part of a sentence, can subtly change its meaning and even its sentiment, creating sentiment ambiguity. This makes it even more difficult to tell the original, intended sentiment of the writer, as opposed to the interpreted sentiment of the reader.[16]

    Analysts and Accuracy

    While sentiment analysis has been around since early 2000, due to limitations and challenges mentioned above, it has never been performed with 100% accuracy. In fact, humans themselves can only agree on the correct sentiment of a sentence 80% of the time, so reaching 100% correct sentiment for any automated system is still some ways off. While the results returned from sentiment analysis are beginning to reach human level accuracy, it is how the results are interpreted and used that is the next great challenge. While the heavy lifting of sentiment analysis is performed by computers and algorithms, it is left to skilled analysts to interpret those results and act upon the final analysis. There is currently a shortage of these skilled analysts to properly act upon these opportunities, and this has led to slower growth and expansion in this field.[17]

    Implications for Businesses

    Brand and Reputation Management

    In this connected world, where bad publicity can reach the masses quicker than ever before, it has become more important than ever for businesses to perform brand and reputation management. Positive social brand conversations are a key driver that can lead to increased sales and brand loyalty, but the reverse is also true. Negative sentiment can seriously hurt a company not only in terms of how they are viewed in the public eye, but ultimately their stock price, and bottom line.[18] While some companies can recover from this, the best thing any organization can do, is prevent it from happening altogether.

    This is where sentiment analysis becomes an excellent tool to monitor in real-time, how a particular product, brand or service is being perceived, and adjust their strategy as required. Understanding how their products are thought of in the marketplace, gives insight into the consumer and ways for an organization to improve upon them. [19] Equally important is that the perception of a brand or product can vary from audience to audience, so knowing the sentiment of a particular group can help focus efforts to make sure the company is viewed in the most positive light. It also serves as an excellent source of root-cause analysis, where if the sentiment of a particular group is negative, it gives organizations a real-time indicator that there is something in the product or service that requires re-evaluation. Social media has given the general public a very loud and important voice, one that businesses need to listen and pay attention to. Sentiment analysis is one tool that can be used to have that conversation with the public, listening to concerns, understanding customers’ attitudes and keeping their brand and reputation in a positive light.[20]

    Supply Chain Management

    Sentiment analysis is now being used to help reduce the well-known phenomenon in supply chain management known as the bullwhip effect. In supply chain management, the goal is to produce and deliver products and services from their origins, to the customers, as efficiently as possible. With previous forecasting models relying on outdated, and over or under-estimated data, any change in demand for a service or produce can lead to significant problems down the supply chain. Using sentiment analysis however, suppliers are now able to use demand sensing, that is real-time, and is able to provide insight about consumer attitudes and more accurately predict future demand.[21] With it, suppliers can forecast if demand is where they expected it or if changes are needed in their supply chain, that will evidently lead to a reduced bullwhip effect and a leaner, more efficient supply chain process.[22]

    Future Use

    Accuracy

    Due to technical limitations, sentiment analysis is approximately as accurate as human beings, 80%. This leaves much room for improvement in the future. The key to improving accuracy with current technology is to focus on the specific topic area; the more specific the topic type and the type of texts being analyzed, the better the sentiment analysis can be tuned to perform. That said, 100% accuracy is not at all that necessary for useful qualitative conclusions to still be drawn.

    Sentiment Analysis in a Context Beyond Text

    The science behind Facebook's new emoji

    With the increasing popularity of Facebook, Instagram and Vine videos, people are expressing themselves in their videos and text. Sentiment analysis programs and softwares need to keep pace with this change.[1] How will Facebook classify the use of a ‘like’? How does Twitter classify a retweet or favourite? Are they considered to be three positive interactions? Is an “unfollow” on Twitter classified as a negative interaction? Recently Facebook introduced its new emoji feature allowing users to express their reaction to posts using seven emojis (like, love, haha, yay, wow, sad, angry).[2]. This new feature offers greater business insight and usability beyond the positive, negative or neutral scoring systems.

    Similarly the detection and exploitation of emotion in speech and images (such as facial and body language) implied by video-captured behaviours will increasingly come into play, including actively in meeting commerce and security needs.[3]

    Online Dating

    In recent years, online dating sites and apps, such as Plenty of Fish, Tinder, and many others, have exploded in popularity. More and more people are turning to such platforms to find dates and partners, but with an increase in online communication comes the issue of being able to truly sense the emotions of the person on the other side of the screen. Communicating in-person with someone, whom they like, can be difficult, but not being able to pick-up on emotional or physical cues can make the situation even that much more difficult. However, there is an emerging solution to this, the use of sentiment analysis. Such an application of sentiment analysis is in its infant stages, but with advancements in this technology, SA could be seen in more communication based apps. There are a couple of apps on the market that are using sentiment analysis.

    Lulu

    Lulu is a mobile dating app for women that allows women to make positive or negative ratings on male users. The app only allows women users to access these ratings. Lulu’s goal is “to provide a private network for women to share their experiences and get information to make smarter decisions”. Lulu works by having women perform two functions. Firstly, women answers a set of multiple choice questions about a male’s characteristics. Secondly, Lulu let’s “female [users]... select hashtags that describe male [users]”. Then, the used analytics system scores these hashtags and gives a score. The system then combines the multiple choice answers and scores from the hashtags, to give an overall score for the male users.[4][5][6]

    TextAt App

    TextAt is a Korean dating app that “uses sentiment analysis to report on a texter’s level of affection”. By detecting key words and phrases, in relation to context, the app can, to a certain degree of successfulness, determine the level of affection that one, or both, texters have for each other.[7]

    Although such analysis of the online dating world is in its very infant stages, these two apps illustrate the potential that sentiment analysis, combined with other technologies, can have on not only the online dating world, but in general online communication.

    Creating a reliance on data could change the dynamics of intimate relationships. By quantifying the unknown and unquantifiable, like gut feelings and crushes, the early stages of intimate relationships go from a guessing game to a series of calculations that take into account new social practices. These new social practices could include the manipulation of data via "ballot stuffing" and "vote brigading." Allowing technology to track the behaviour of an individual disturbs the equilibrium of trust in an intimate relationship. Instead of being faithful to not harm each other, partners may remain faithful in order to save face.[8]

    Psychology

    Many people today suffer from mental illness, such as depression or symptoms of depression, but many are unwilling or unable to seek proper treatment. Other methods of therapy or psychology, such as online or over the phone, have been used to make such services more easily accessible and more comfortable for users, but many still do not use these services. Although not implemented or used, there has been discussions of using programs, such as Apple’s Siri, Google Now, or Microsoft’s Cortana, as a virtual therapist or psychologist. The idea is very interesting and would require the combination of artificial intelligence, sentiment analysis and expertise from therapists and psychologists to perform such a capability for these systems. There is no date for this to occur, but perhaps one day everyone could have a therapist or psychologist in their pocket.[9]

    Research has been conducted in detecting signs of depression through online social media sources. Depression, simply put, is the the presence of “severe negative emotions and lack of positive emotions”.[10] Sentiment and text analysis can be combined to detect signs of depression using two methods:

    • Subject-independent analysis: Detecting polarity between Tweets without considering if it is related to the subject. This may refer to Tweets that contain statements about negative things, but may not refer to the person making the Tweet.
    • Subject-dependent analysis: Detecting polarity between Tweets that refer to the target. This may refer to Tweets that contain negative things and refers to the person making the Tweets.


    This application of sentiment analysis is still in the research stage, but has the potential to potentially find and alert individuals who are suffering or are showing symptoms of depression. The difficulty with this application is differing between someone who is actually showing signs of depression and someone who just makes negative posts as they may be a negative person, but not necessarily depressed.[11]

    Terrorism & Crime

    Analyzing textual data and trying to detect patterns and sentiment has been something policing agencies have been doing for many years; however, until recently, this process has been done manually, and is very time consuming. Although the use of sentiment analysis in fighting, solving and predicting crime is not widely used, it can save a large amount of time. With sentiment analysis, policing agencies can now fight, solve or even foil crimes before they occur. Criminals are using online communication to pass on information, thus by analyzing these sources for key terminology and names, and combining this with geographic characteristics, policing agencies can better track and potentially predict the location of crimes.[12] The following is an example of how sentiment analysis could have played a role in preventing a mass murder.

    A lone gunman killed six and injured thirteen on May 23, 2014, in Isla Vista, California. A day before this incident, the killer posted a video on YouTube, called “Elliot Rodger’s Retribution” where he vented his anger and his intentions. A data scientist professor from University of California, Santa Barbara, wanted to see if sentiment analysis programs could detect the level of negativity in Rodger’s video and flag it to have alerted the authorities. To do this, the professor downloaded the video’s script and utilized various text and sentiment analysis programs. After the professor’s experiment, they concluded that there was a high level of negativity and many of the words and phrases used were common among those used by other murderers in the past. Although this is just a proof of concept, with other technologies, this concept could be applied to real-world situations, to be able to alert authorities in a timely manner and have the potential to save lives.[13]

    Medicine and Healthcare

    Another application of sentiment analysis that is still in its infant stage is in the field of medicine and health care. The use of social media and sentiment analysis has illustrated its benefits to the health care system. The following are two examples that demonstrate the power that sentiment analysis can have in medicine and health care.[14]

    Sentiment Analysis and Heart Disease

    Researchers from the University of Pennsylvania, through the use of Twitter data, were able to predict the rate of coronary heart disease. By applying sentiment analysis to Tweets posted in various locations, the researched showed that words expressing “negative sentiment such as anger, stress and fatigue with [an area]… were associated with higher risk for heart disease”.[15]

    Sentiment Analysis and Predicting Emergency Room Visits

    Researchers from the University of Arizona teamed up with a Dallas hospital to try and predict the number of emergency room visits due to asthma-related incidents. Sources of information used were from Twitter, electronic medical records and air quality sensors. The researchers found that the number of emergency visits due to asthma-related incidents increased as air quality fell. As well, emergency room visits increased as more Tweets containing asthma related words, “such as asthma, inhaler, or wheezing”, were posted. By utilizing these sources of data, the researchers were “able to combine machine learning [and sentiment analysis] to predict the volume of asthma-related emergency room visits”.[16]

    As more time is spent on research in this area in the future, researchers may be able to identify keywords and phrasing that are related to various illnesses, diseases, and other health concerns to recognize outbreaks, best allocate resources, improve the healthcare system and understand the overall well being of the communities we live in.[17]

    Conclusion

    Sentiment analysis is becoming a more and more crucial tool for businesses, government and other organization in the world of Web 2.0. It allows business to better understand consumer perceptions of their brand, products and services, government to understand citizen’s views on policies, and allows organizations, such as hospitals and police, to understand trends and makes predictions in health care, crime and other applications.

    The technological advancements for sentiment analysis is crucial as our world becomes more digital and moves towards Web 3.0. Advancements, such as better understanding and detecting of context, sarcasm, sincerity and moving beyond text to audio and facial applications will allow sentiment analysis to be more important than ever.

    References

    1. https://www.socialmediaexplorer.com/social-media-monitoring/sentiment-analysis/ accessed on October 30, 2015
    2. https://www.youtube.com/watch?v=HgWGCN-KRto accessed on October 30, 2015
    3. http://www.informationweek.com/where-sentiment-analysis-heads-next/d/d-id/1106464?page_number=2 accessed on November 1, 2015
    4. https://onlulu.com/about_us/ accessed on October 17, 2015
    5. https://en.wikipedia.org/wiki/Lulu_(app) accessed on October 17, 2015
    6. https://iapp.org/news/a/data-driven-dating-how-data-are-shaping-our-most-intimate-personal-relation accessed on October 17, 2015
    7. https://iapp.org/news/a/data-driven-dating-how-data-are-shaping-our-most-intimate-personal-relation accessed on October 17, 2015
    8. https://iapp.org/news/a/data-driven-dating-how-data-are-shaping-our-most-intimate-personal-relation accessed on October 17, 2015
    9. http://emotion-research.net/workshops_folder/workshop.2011-04-21.8416231528 accessed on October 18, 2015
    10. http://www.ubiwot.cn/download/A%20Depression%20Detection%20Model%20Based%20on%20Sentiment%20Analysis%20in%20Micro-blog%20Social%20Network.pdf accessed on October 20, 2015
    11. http://www.ubiwot.cn/download/A%20Depression%20Detection%20Model%20Based%20on%20Sentiment%20Analysis%20in%20Micro-blog%20Social%20Network.pdf accessed on October 20, 2015
    12. http://lexalytics.com/lexablog/2015/five-ways accessed on October 23, 2015
    13. http://lucafoschini.com/2014/05/25/machine-learning-isla-vista-shooting.html accessed on October 27, 2015
    14. http://lexalytics.com/lexablog/2015/five-ways accessed on October 23, 2015
    15. http://lexalytics.com/lexablog/2015/analyzing-social-media-improve-healthcare accessed on October 27, 2015
    16. http://lexalytics.com/lexablog/2015/analyzing-social-media-improve-healthcare accessed on October 29, 2015
    17. http://lexalytics.com/lexablog/2015/analyzing-social-media-improve-healthcare accessed on October 29, 2015
  • Personal tools