Mention counting and standard sentiment analysis is not enough; it cannot sustain the successful executive's need to have greater visibility and accuracy, and more importantly, control over the market's perceptions of a brand in social media.
So says Robin Meisel, business line executive at Business Intelligence Application International (BiA), adding: "The old way of doing social media analytics is obsolete; without full linguistic analysis overlaid on business intelligence, you are blind in social media."
Meisel says sentiment analysis alone cannot deliver a full picture.
"You need to include semantic processing to understand sentences that contain ambiguity in intent, where meaning is assigned to a phrase and not only a single word," he notes.
In an age of big data and massively growing social media interaction, only assigning humans to monitor all mentions and derive a complete picture of the market's sentiment is not viable.
"However, next-generation analysis allows brands to stop pulling individual needles from the haystack, and instead x-ray the entire haystack for an accurate view of all the needles," he says.
It also allows marketers to more accurately measure ROI on their investments into brand, marketing and social media strategy.
At a tactical and operational level for marketers, linguistic analysis provides the capability to accurately identify trends and to take appropriate action in a manner more aligned to its customer base and market expectations.
Next-generation linguistic analysis allows brands to analyse any written content and get a contextual understanding against a predetermined set of measures. In addition, they can identify and assess the market demographic, based on the style of writing, grammar, spelling and the use of punctuation in those comments, says Meisel.
The vast amounts of comments that are in the social space means that linguistic analysis mechanisms are primed to provide the insight C-Suite executives are looking for, says Meisel.
"The old tools only look for keywords and numbers of mentions, they also look for the sentiment behind them, but the context in which those words were used are ignored, for example, 'this ad kills me!' which is actually a positive statement."
Identifying patterns based on writing style, grammar, spelling and use of punctuation is unique to linguistic analysis because we are able to identify a specific demographic by specific measures to enable marketers to better target their messages to the most appropriate grouping of consumer demographic for their product or service.
Every current-generation tool can measure market sentiment; only next-generation linguistic analysis can drill down into psychological drivers of specific target markets' thinking behaviours, based on their language and grammar use.
Meisel illustrates the difference in language use by various demographics by pulling assorted tweets.
Linguistic analysis would identify those who come from a lower LSM: "i just want to say gudnyt to my lovely girlfriend Z i luv her very much i hope she is listng from K", as well as those from higher LSMs: "it was amazing ? you are so special! Don't forget the promise we made" and "I'm really enjoying "Married to Jonas" on E! so real!!!".
Sentiment analysis would pull out the positive, negative or neutral from these tweets but would lack the context of demographic measure.
Meisel says sentence structure, use of language, grammar and punctuation are dead giveaways on the demographic and psychological profiles of the writers. This, he says, makes linguistics analysis a necessary tool whose application extends beyond marketing, to fraud and risk mitigation as well as law enforcement, for example.
Meisel adds that earlier iterations of sentiment analysis tools may have looked to keyword searches and adjective analysis. However, slang, expletives and colloquialisms may throw the software off track.
"In some contexts, an expletive may be used in a positive way, for example: '&^%$, I love this beer', which would be viewed as negative through typical sentiment analysis. In addition, these tools may have misread a comment such as "I love Zuma, I hate Zille".
Standard sentiment analysis would pick up Zuma and Zille being subjects and picking up one negative sentiment and one positive sentiment for each, and therefore be neutral. A key element in linguistic analysis is tokenising, whereby the sentence is split into two with the correct sentiment being allocated to the correct subject and the resulting output being correct."
Meisel will address the upcoming ITWeb BI Summit on the use of linguistic analysis in social media channels. For more information about this event, please click here.
Share