Essential New Language Technologies Finally Unlocking Full Value of Social Listening and Intelligence
Let’s face it, human language is complex. Linguists are still unsure how children acquire syntax. Clearly, though, in an era of social connection, understanding human language - ideas, opinions, wants, needs, concerns, emotions and more -- is arguably more important than ever.
Never before have we had such powerful access to the unstructured, unvarnished voice of the customer. And, yet, never before have we been challenged with having to separate as many meaningful signals from as much noise to get to those voices in a reliable way. Words like "natural language processing,” “machine learning,” “deep learning,” “cognitive technology,” “semantic analysis,” and “artificial intelligence” are indeed rapidly becoming buzzwords of the day, but they should not obscure the powerful impact these technologies are having on social listening and intelligence, and other related voice-of-customer data. Understanding these language technologies is increasingly important to everyone involved in social intelligence, consumer insights, data-driven marketing, analytics, business intelligence, customer care and more.
Today, most large businesses have installed some form of social listening, however imperfect. They recognize that they can’t afford to ignore what people are saying off the cuff, unprompted in social media—about their company, their products, their competitors, their industry, and many other topics. Leading companies have invested staff time, technology budgets, and executive attention to understanding the conversation, expecting to extract insights from the data that drive better business decisions. But many organizations report that the data remains underutilized, since nearly every part of the enterprise could benefit from these insights.
While the adoption of social media has skyrocketed, the technologies required to effectively “listen” to these massive unstructured, unprompted discussions have struggled to keep up. The challenges are many: new words (neologisms) emerge daily. Slang, sarcasm and implicit meanings dominate the discussions. Spam is pervasive. The emergence of emojis, for example, further complicates the analysis of opinions. All opinion is derived from a point of view (or target).
Effective analysis today also has to go beyond basic “sentiment” analysis to unlock a richer array of variables in the data that provide more actionable meaning and insight, such as emotion, intensity, advocacy as well as a deeper understanding of “who” is expressing the opinions on a segment and individual basis. It has to ensure the data you’re analyzing is actually relevant to what you’re looking for and purged of the noise that often pollutes these datasets and renders them ineffective for analysis. And it has to be conducted on the detailed facet or target levels so that one understands not just generalities of opinion, but the specific root cause drivers.
The good news is that new language technologies, specifically machine learning and AI approaches, are providing a much higher level of precision and reliability in the data. With the proper precision, relevancy and recall (the number of “signals” gleaned from a specific data set) provided through new active machine learning techniques, multiple third party studies have recently shown the power of social to predict sales, measure brand health as well as advocacy. One such study, conducted by the Word of Mouth Marketing Association (my company, Converseon, provided the social data), together with group of leading brands, including Weight Watchers, Pepsi, AT&T, among others, integrated social and word of mouth data into a marketing mix model. The result shows “word of mouth” including social media drove approximately 13 percent of sales and further demonstrated that attribution or marketing mix studies that do not include this data are flying partially blind. Other studies have found that social listening data, properly cleansed and annotated for high precision, can effectively predict brand health. Professors Wendy Moe and David Schweidel, in an award winning Marketing Science Institute (MSI) paper conducted a series of experiments (also using Converseon data) that showed, properly modeled, that social listening data had a very high correlation with traditional brand trackers, providing a way forward to use social data in brand health analysis for “better, faster, cheaper” results.
Understanding the history and evolution of these language technologies is the thrust of a new White Paper now available that delves into various approaches and provides guidance on how to best choose a solution that best meets your organization's needs to fully drive value and ROI from this data. It is available for free here. It's an essential read for those not only directly involved in social listening, but also consumer insights, business intelligence, customer care and more.
Rob is CEO of Converseon, a recognized leader in social intelligence, as well as its sister pure-play semantic technology firm Revealed Context. Both firms offer the proprietary ConveyAPI technology, Dataweek's Top Innovator in Social Data Mining, which uses active learning techniques for enhanced social listening capabilities. For more information, please contact [email protected]