By Chris Albon

Since launching in 2008, Ushahidi's tools have helped individuals and organizations receive millions of crowdsourced reports on topics ranging from flooding in Pakistan to co-working spaces in Africa. We love that people use our products in diverse ways and locations. However, because our users are so geographically distributed, it is often difficult to speak to them face to face about how they use our crowdsourcing products. When we do, the results are amazing, but since that is often not an option, we're always looking for other ways to gain insights about our users.

A few months ago, a man came up to me after I had given a presentation about Ushahidi and asked if "all this crowdsourcing is really just a system for people to complain about their lives?" I said that I didn't believe so — I could certainly think of anecdotal cases of individuals using Ushahidi tools to express both positive and negative sentiments — but frankly I didn't have a solid answer. His question bothered me, were we just a tool for complaints? I decided to find out.

To answer his question, I needed data. More specifically, I needed crowdsourced reports from a variety of Ushahidi deployments. Aggregating multiple data sources manually is frustrating and time-consuming. However, thanks for CrisisNET, two minutes and a few lines of code gave me 4040 reports from Ushahidi and Crowdmap deployments worldwide. Fast and simple.

With data in hand, it was time to start the fun part. I ran a script using a pre-existing list of English words with positive and negative connotations to assign a sentiment score to each English-language report in the data (2239 report). In the end, the analysis found 746 positive and 838 negative reports. More importantly, since the score assigned a degree of sentiment (that is, the number of occurrences of positive or negative words in a single report), I could easily visualize the distribution of the reports' sentiment.

What was the answer to the man's question? Thanks to CrisisNET, I can say with some confidence that Ushahidi deployments receive roughly an equal number of positive and negative reports. That is, no, Ushahidi and Crowdmap are not primarily used to complain. Instead, the visualization of reports sent into the Ushahidi platform and Crowdmap reveals a wonderful spectrum of sentiment, from very negative to very positive. Most importantly, it reveals something we didn't know about how people use Ushahidi's crowdsourcing product before now, giving us real insights into our users and their work.

Want to conduct your own sentiment analysis? Here is a two part step-by-step tutorial (with the full code) explaining how we created this visual. We hope you'll use it to replicate our analysis and run your own sentiment analyses on other data available through CrisisNET.

Note: Since it wasn't part of my original question, I removed all reports that received a neutral score (meaning that no positive or negative words were matched, or the positive and negative scores canceled each other out). I will dig deeper into this and other parts of the analysis in the tutorial