Sentiment analysis is an excellent tool for evaluating communication, especially information directed at the public. In automated sentiment analysis, text – spoken in this case – is examined particularly for terms or words that convey a certain mood. This type of speech analysis provides valuable insight into the style of communication and helps to identify changes of sentiment at an early stage.
We took advantage of the automatic mood analysis of our BI Data Cubes and examined the communication via TV and radio in connection with Covid-19. We focused on the areas of public information and government communication with the core topics of government as well as cure and travel.
In the past few weeks, our broadcast data cubes have recorded a total of more than 2 million contents on the subject of corona, with more than 222,000 of them dealing with our core topics. Governments and leaders around the world have tried to provide the population with the most important information as effectively as possible and to draw their attention to the seriousness of the current situation.
How successful were they at that? Looking at the sentiment analyses, we note that results were consistently positive. We focused on the areas USA, Europe and China, in order to be able to draw comparisons between regions. Of course, the number of contributions also plays a certain role, as logically, the highest amount of English-language content was produced in the USA.
Nevertheless, the development of the numbers is interesting, as there are significant differences between the three areas. While figures in China reached two peaks in the first half of May and were otherwise rather unstable, there was a clear downward trend in Europe. In the United States, the numbers were relatively stable before falling sharply.
The global numbers of the sentiment analysis speak for themselves: While only 3% of the content was identified as negative and around a fifth as neutral, the sentiment of the large majority of the contributions is positive.
Here is a short extract from our dashboard:
The full analysis can be found here: