Adotas

Where media buyers start online

Featured Author

Jennifer Major currently works as a Technical Account Director, looking after major Media and Telecommunications customers in the UK. This opportunity to work closely with both content producers and content distributors, has given her a detailed understanding of the challenges faced by organizations as they work to improve the way in which they use textual data (e.g., descriptions of content). Jennifer’s experience of utilizing Natural Language Processing techniques has provided her with the opportunity to work with a range of consumer focused industries to help them exploit the new and exciting area of Social Media Analysis.

More articles by Jennifer Major





Features

Meaning Within Measurement: Sentiment Analysis and Social Media

Written on
Dec 22, 2009 
Author
Jennifer Major  |
Share
Meaning Within Measurement: Sentiment Analysis and Social Media

microscope_small.jpgADOTAS – We’ve seen the blunders big companies have made by not monitoring and, more important, not reacting intelligently and expeditiously to social upheaval in the digital space -– how can we forget United Airlines’ fate following musician Dave Carroll’s infamous YouTube video of “United Breaks Guitars” and its sequels?

In today’s current market, businesses and their reputations cannot
afford to take that kind of hit.
Through blogs, message boards, fan pages and the like, the Internet is fast rebalancing the relationship between customers and companies, while social media networks, such as YouTube and Twitter, are giving consumers instant and, occasionally, very powerful ways to “strike back” and make their voices heard.
Companies from all industries have quickly realized the necessity to go where the conversation is going to remain relevant and pertinent to customers. Big companies, such as Ford, Procter & Gamble and Coca-Cola to name but a few, have recognized the need to use the Internet and, more important, the rich vein of market intelligence that social media sites provide, to “listen,” monitor and, if need be, counteract any bad publicity these virtual -– and viral -– conversations might
be generating in order to avoid a fate similar to that of United Airlines.

Social Media Analysis — Tools Abound
From The Financial Times’ “Newsswift” program to the funkier-named “Tweetfeel,” “Twendz” and “Twitrratr,” a plethora of social media analysis tools have hit the market. These include solutions, such as text mining, natural language processing and other sentiment analysis technologies that have been developed to help organisations gain intelligence from social media sites and build a more complete view of their brands’ reputation from a consumer’s perspective.

Businesses are clearly spoiled for choice. Social media technology and techniques include web crawling APIs that collect keywords and free format text relating to specific criteria, text mining applications that analyse key concepts, features and even segments of common terms, as well as pattern matching, probabilistic modeling and sentiment analysis technologies that evaluate the information as positive, neutral or negative.

Such tools and techniques have been added to the product portfolios of an array of companies including Internet/publishing/data mining/research/marketing organisations including the likes of Factiva, Motive Quest and Omniture, to keep on top of the social media game.

Is this seemingly “Jack of all trades, master of none” approach to social media analysis a reliable one or should businesses wanting to monitor, measure and understand social media conversations opt for organisations with an expert analytic heritage at its core?

Which One’s for Me?
A combination of the above techniques can undoubtedly provide a richer and more contextual set of data than traditional keyword spotting tools -– yet which is the best one to adopt? It is clear that there are companies out there that generally take a couple of different stances to analyzing social media information –- so it is important to note the distinctions, which will ultimately impact the relevance,
reliability and validity a business might be looking for.

The simplest algorithms work by scanning keywords to categorize a statement as positive or negative based on simple binary analysis (“love” is good; “hate” is bad). However, such an approach fails to capture the subtleties that bring human language to life: irony, sarcasm, slang and other idiomatic expressions.

Social media, which are by nature dynamic and based on unstructured forms of information, do not fit neatly into traditional database-driven analytics systems. You need reliable sentiment analysis capabilities that require the ability to understand many linguistic shades of gray.

Sentiment Analysis: The True Differentiator

Sentiment analysis is an important but very hard to master science still in its infancy. While it can be quite accurate –- reliability 80% or higher -– it does not
necessarily make the data valid or useful for making strategic decisions grounded in effective brand monitoring.

Also when it comes to languages, things get more complex than simple tweet or text
analysis, making success an even more elusive concept for sentiment analysis, where cultural differences –- an American “quite” would mean “very” whereas an English “quite” would refer to “not at all” -– and linguistics come into play. Sinful isn’t always sinfully good chocolate.

Understanding social media is much more sophisticated and demands building in an analysis of sentiment. It is fair to assume that the added value promised from a company whose heritage isn’t a sophisticated analytical one, would not be equipped to analyse and provide as relevant results when translating finer linguistic nuances, cultural factors and the vagaries of human emotion, and might not help avoid what could be very damning social media comments.
The choice of a social media analysis platform to protect brand reputation requires companies to include the important, but very hard to master, science of sentiment analysis into its analytical reporting. In this fast-paced market, it is important to be able to review information in near real time.

Bringing all this data together –- research, monitoring, sentiment analysis and other analytical capabilities –- can start to provide the grand unified vision that overlays all relevant data sets for correlative analysis. Only this way will we start to determine an ROI for social media campaigns –- finding the meaning within the measurement is critical.





Reader Comments.

Interesting perspective, but as my father said back in the 60s when he was publisher of Fortune: “If we can put a man into orbit, why can’t we, with any degree of certainty, measure the effectiveness of communications. The reason is simple, and perhaps a bit old fashioned. People. Human beings with a wide range of choice, cantankerous, capricous and consumed by innumerable conflicting interests and conflicting desires. ”
The problem with your post is that it ignores the importance of the human element. At best, automated sentiment analysis is correct about 65% of the time. If the finance department was only accurate 65% of the time, the CEO would be in jail, and the company would be bankrupt. You cannot rely entirely on computers. You need humans, trained in the science of research and analtyics to check the accuracy, as well as to interpret the results.

Posted by Katie Delahaye Paine | 10:19 pm on December 22, 2009.

Good article which states concisely opportunities vs problems in social media analytics. Being a text miner myself i could add my 2 cents by saying that accuracy in sentiment analysis is very application – specific, typical being around 72 – 75% in my experience. This technology provides very good insights in what people want and how they think, why they don’t like specific products / services or politicians. Great caution is needed not to make wrong conclusions though.

Posted by themos kalafatis | 10:57 am on December 23, 2009.

Here’s a perfect formula for sentiment analysis: http://www.conversition.com/a-formula-for-perfect-sentiment-analysis/

Posted by Jeremy | 6:47 pm on December 23, 2009.

This is an excellent post. Perhaps the best free semantic analysis tool out there is TipTop http://FeelTipTop.com

Posted by Shyam Kapur | 2:17 am on January 3, 2010.

Leave a Comment

Add a comment

No Tags
Article Sponsor

More Features


  • Right now, at the beginning of 2012, what are you watching the most closely for its ad and marketing opportunities?

    View Results

    Loading ... Loading ...

Latest News

News Archive

Spotlight

Sponsormob Leads the Way Into RTB for MobileADOTAS – For more than half a decade, Berlin-based tech firm Sponsormob has remained relevant in an industry characterized by [...] more...


Adotas Partnership