Social Listening – Social Analytics – Social Intelligence
Social Listening, Social Analytics, and Social Intelligence - are they the same or are they integral parts of a sequential process, a so-called continuum?
Quite a few pundits have discussed this question in their articles, blogs, and essays. The most controversial of the three is social intelligence; if you Google it you will find its Wikipedia definition on first position explaining that it is “the capacity to know oneself and to know others”.
Of course, the alternative definition, the one that was coined after social media monitoring and analytics became popular only appears on page four of Google search - which you would only know if you are me and you search for social intelligence. In this secondary context, this bigram* means: the knowledge and insights that organisations can extract from online posts (mainly on social media platforms) published by their customers and other stakeholders.
The precondition to get to actionable insights is to avoid “Garbage-in” during the so-called data harvesting process, and to use appropriate machine learning models to maximise the accuracy of brand, topic, and sentiment annotation.
But let us look at the three bigrams one at a time.
*combination of two words
Social listening is short for social media listening, and to some a synonym of social media monitoring. Most people use social listening as an all-inclusive term for all online sources from which online posts can be gathered or harvested. However, in addition to popular social media platforms such as Twitter, Facebook, Instagram, YouTube, we also can harvest data from blogs, forums, news, and reviews. There are other sources that are country specific, such as Weibo for China and VK for Russia.
It is of paramount importance to harvest for all synonyms and avoid all homonyms (garbage-in) which can be over 80% of all harvested posts. A homonym is a word that is spelled the same way as the keyword for harvesting but means something entirely different, which makes it irrelevant for the project at hand. The classic example used to explain this problem is: wanting to harvest posts about Apple the company but ending up with lots of posts about the fruit or juice; let alone Apple Martin who is Gwyneth Paltrow’s daughter and if not excluded will invite a lot of noise in the dataset rendering it not only useless but dangerous for the user!
Fig1 listening247 screenshot: Query Builder for keyword based search / data gathering from blogs, forums, news, reviews, videos, and Twitter
Social analytics is what happens after the posts are harvested from the various sources and irrelevant posts are removed as “noise”. For data analysis to take place, accurate intelligence needs to be added to the dataset using machine learning and/or rules-based NLP methods. The most common annotations added to each post or relevant snippet within the post (this is a longer story that requires its own blog post) are: brand, sentiment, emotions, and topics/subtopics/attributes.
Fig 2. listening247 screenshot: Data explorer detailed view of documents/posts
These annotations can be added for text in any language, and the accuracy sought after – measured in precision, recall and F-score – should be over 75% in all cases. It is even possible to reach F-scores that are over 95% with focussed and context related training of suitable machine learning algorithms.
Social intelligence is the wisdom discovered by exploring the intelligent dataset. You see, adding brand, sentiment, emotion, and topic annotations to a dataset makes it “intelligent” but in order to find wisdom or “actionable insights” a lot more than just accurate annotations is required.
For the time being, an intelligent data cruncher and a powerful filtering or drill down tool is still needed to explore a dataset and find the gold nuggets.
Fig 3. listening247 screenshot: filters to explore the annotated data discover insights
We like to think of listening247, the DigitalMR unstructured data analytics platform also used for social listening, as the Google Maps of big data. It enables a user to navigate in a maze of millions of online posts - or other documents for that matter - safely and accurately from A to B; B being the destination or in our case the actionable insight. The path to finding the actionable insight is oftentimes a data story worth telling.
Share this article: