Sample #askgove word cloud created from around 2,500 tweets
Education asked last Tuesday if we could create a word cloud on Friday from the questions asked on Twitter using the #askgove hashtag. One of those jobs that seems simple on the surface but isn’t!
- Problem one – by Tuesday there were already thousands of tweets, and Twitter will only allow you to search so far back on a keyword.
- Problem two – they wanted the cloud generated on Friday (when they go to print) so they could include as many #askgove questions as possible, which meant checking for new tweets every couple of hours during the week to compile an immense list.
- Problem three – because there were so many tweets, it was impossible to go through and weed out all the extraneous words like reply, retweet, favorite, open, askgove before generating the cloud, to say nothing of all the stop words (and, a, the…). They wanted a cloud that highlighted the key questions being asked, so no words relating to usernames, no why/will/what/when… and sadly no swearing!
- Problem four – I don’t work on Fridays.
I got as far as I could with it – I searched for #askgove on Twitter and pasted the available list of tweets so far into a program called word counter, to generate a list of words ranked by frequency. That weeded out some of the basic stop words. But how to turn that into a Wordle? I could see the most popular terms, but they only occur once in the text generated by the counter so the word cloud would be meaningless.
Step forward production, specifically a systems editor, who showed me a nifty bit of code which takes the word counter list and returns each word, repeated as many times as the frequency number next to it. Weed out the words we don’t want (check the ones we’re not sure about – ebacc, ict, hei – on Twitter), paste this into Wordle and voila! a word cloud.
I showed the process to the art director who works on Education, and mocked up a word cloud using the layout and colours she chose, to see whether it worked on the page.
I wrote detailed instructions for colleagues, and at their request I talked them through the process at my screen, so they could create the cloud without too many difficulties. They started to add to the list of tweets at the end of Wednesday (while I was still in, to check they’d got the process right).
…the word cloud was dropped from the supplement. This happens fairly often in journalism – a story is superceded by breaking news, the space is needed for advertising or a better alternative presents itself. The reason in this case was space – the word cloud simply didn’t work in the space available on the page. And they let us know early on Thursday, so my colleagues didn’t spend too long on it (sometimes we don’t get told at all).
So was it a waste of time? No. I learnt some valuable lessons, about how to generate word clouds but also about working with different departments (and colleagues) to create something for the paper.
- If something seems impossible at first glance don’t just dismiss it, there’s usually a solution and sometimes you have to put a bit of work in.
- Ask for help if you don’t know how to do something – in such a big organisation there will usually be someone in the building who has the knowhow.
- Collaboration is key – education came to us at the beginning with a clear idea of what they wanted but little knowledge of how it could be done; I took it as far as possible then consulted someone with the technical knowledge; and collaborated on the design so the editors could make a final decision. Sharing knowledge led to a better end result, even though it wasn’t used.
- Now I know how to create a word cloud from any volume of text, so if it comes up again it’ll be easy (she says…).
- Walking colleagues through a complicated process is better than just emailing a list of instructions, which can be confusing (some people learn better with visual aids) and can seem a little superior (not everyone responds well to being told what to do remotely).
I think that last one is the lesson I should really take to heart!