Social Media Word Cloud Analysis
This post was originally published on the muddywall.com website when it was a social media based platform, however unfortunately the website has now been taken down and is to do with plants and art by the looks of it, so I have republished the article below.
Word clouds, often known as text or tag clouds, are not a very new concept. But they are a pretty neat visual representation of text data that make it easy to spot the frequency of keywords within a body of text or document. They started off as a navigation aid on websites and blogs back in the mid 90s, and more recently have become a bit of a decorative social media craze. But their real power lies in the analysis of online conversations, helping to identify key issues, trends and opinions and are a staple part of generating social media intelligence.
How do they work?
Any word mentioned in a list or body of text is represented in a word cloud and it’s the font size and weight represents the frequency of which that word appears in the body of text. Now the technology behind such things can be quite complex, however fortunately there lots of tools online already there that make them easy to construct and download for analysis.
Let’s take a look at a sample word cloud generated from the RSS feed on the Muddywall website.
Muddywall is heavily focused on social media topics, in particular Facebook and Twitter. We can tell this by the size and weight of the words social, media, Twitter, Facebook, tweet, page (likely to be in the context of Facebook Pages).
But what else can we tell? Let’s look at some of the less prominent keywords… infographic, engagement, market, business, company, consumer… these aren’t just coming from a single article these are topics that are being discussed again and again on the Muddywall blog. We can say then that Muddywall isn’t just social media focused but is heavily weighted towards social media use for business.
Now let’s compare this with some sample data from social media conversation across the web about and surrounding Muddywall. This data has been collected through the online reputation management and social media monitoring platform Trackur.com, exported into a .CSV file and then the body of the tweets, statuses, blog posts, news articles etc… fed through the word cloud generator.
From this we can see that a lot of the conversation happens on Twitter and as expected is heavily focused on social media. The high frequency of the term ‘RT’ shows that the Muddywall content is frequently shared by others and we can see many of the individuals that are interacting with Muddywall via their @usernames.
But there is something more interesting in the analysis of this. The first cloud taken from the blog RSS feed showed that Facebook is commonly discussed on the site, yet barely makes an appearance in the social media conversation — it is there, however very small and just above the T of RT.
This tells me that although Facebook is at the core of the Muddywall blog, the articles published aren’t generating as much conversation and buzz as some of the other topics covered. Now this could be for a huge number of reasons well outside the scope of this article, and it would also be silly to make sweeping changes or decisions based purely on this, however we have identified something to investigate and analyse further.
Some words of caution
They identify key topics and trends, but word clouds don’t analyse the data. You still need to organise and interpret the information yourself.
Word clouds emphasise the frequency of words, but not necessarily importance.
There is no context, so the meaning of individual words can be lost and open to interpretation.
One of the biggest problem with word clouds though, is they are too often used when textual analysis isn’t appropriate. Word clouds make sense when analysing word usage but they simply don’t work for complex topics such as election campaigns… Why? See the previous three points above.
Conclusion
Word clouds can be an effective method for analysing text data. They are great for getting an overall feeling of the conversations about a chosen topic or brand online and even better for comparative analysis between you and your competition or perhaps two different products.
There are many online tools that you can use to create word clouds, but initially I recommend having a look at wordle.net and tagcrowd.com.
