Sentiment analysis is the process of using computational power to assess the subjective nature of text and to decide if it expresses a positive, negative or neutral sentiment in relation to a specific subject.Summary by The World of Work Project
Sentiment analysis basically boils down to getting computers to read blocks of text and decide if the blocks say positive, negative or neural things about their subjects. There are several different processes that can be used to do this assessment, but they all aim to produce the same thing: a clear understanding of the sentiment of blocks of text.
There are many different uses for sentiment analysis including trying to understand what voters think about political candidates and trying to understand what consumers think about products. From a world of work perspective though, the most interesting application of sentiment analysis is trying to understand how employees feel about their work and their organization by analyzing their written text either in emails, feedback portals or other locations. This, of course, requires permission to assess this data.
The World of Work Project View
We think that sentiment analysis is fascinating as a subject and that it’s a powerful tool. It’s another example of the added insight that computers can bring us as their learning capabilities increase and as we become able to provide them with more and more data points.
That said, we’re not sure that it’s really a good thing to read the text of employees in organizations. We’re undecided regarding the ethical acceptability of this. We know that sentiment analysis doesn’t need to attribute text to individuals, but we still worry about abuses of power when it comes to reading and analyzing large blocks of text.
Sources and further reading
Where possible we always recommend that people read up on the original sources of information and ideas.
The concepts behind this post is based on lectures we have attended, conversations with various podcast guests and general reading around the subject. There are no specific references for it.
If you see any errors on this page or have any feedback, please contact us.