Sentiment analysis for social media is a rapidly growing market, and one for which we wanted to demonstrate our expertise. This interactive Microsoft Power BI report captures data from the official Robot Wars Facebook page.
Please note that this report features content from an external web source. Change this Limited will not be held responsible for any content contained therein.
As a long-standing fan of the UK TV show “Robot Wars”, I was very excited when it came back on the air on BBC Two in the summer of 2016.
Since I am unlikely to ever build a robot myself, I thought the best way I could contribute to this momentous event was by using my expertise in data visualisation. I thought it would be interesting to analyse public reaction to the show. To do this, I decided to track activity on the show’s official Facebook page. This involved capturing the community interactions (Likes, Comments and Shares) that stemmed from official posts made on this page.
Check out the fully interactive dashboard below. Please note that due to the high volume of content being processed in this report, certain elements may take a while to load on some browsers.
You can view this dashboard in full screen. To do this, click this icon in the bottom right-hand corner of the report.
Sentiment Analysis for Social Media – Robot Wars Community Interactions
The first page shows an in-depth analysis of community interactions. Users can filter this by Episode, allowing you to track how the level of interactions changed over time. Deselecting all Episodes from the menu will display outcomes for the full period from 1st June 2016.
The right-hand section of this page analyses the sentiment of the comments, using the Text Analytics API available from Microsoft Cognitive Services. This tool analyses text and returns a % score based on the sentiment expressed (i.e. whether it is a positive, negative or neutral reaction).
Just a couple of things to note about this API.
Firstly, it needs to be able to identify sentiment-applicable content in order to derive a rating. Therefore, any comments that have no applicable content (e.g. those which are tagged to a Facebook user with no additional text) are ignored. This is why the number of comments in the chart on the right doesn’t always align with the number shown in the ‘Comments’ metric.
Secondly, the API can only process a maximum of 1,000 comments in a single run. Since there are more than 1,000 comments included in this report, we’ve directed it to process the most recent comments. This ensures that the data reflects all comments made during and after the air dates of the series. Some comments in the pre-season are therefore not analysed for sentiment.
It’s this API that really makes sentiment analysis for social media possible. The ability to analyse natural language text and derive a sentiment rating will be game-changing for organisations looking to track what customers are saying about them.
The second page shows all official posts made on the site. You’ll notice a chart that tracks Likes and Comments against these posts over time.
There are three noticeable peaks on the top left chart showing the number of Comments. The first peak, on 23rd June, coincides with the official revealing of the new House Robots. The second, on 13th July, coincides with the official announcement about the air date of the first episode. And the third peak, on 24th July, coincides with the airing of Episode 1.
The number of Likes peaks on the 27th – 28th August, which was the weekend of the Grand Final.
This report is a simple but effective demonstration of how Microsoft Power BI can be used to derive intelligence from social media data.