Mark Pack

There are two nearly ever present comments about the ability of online search tools to provide sentiment analysis: first, the sellers of such tools saying how enormously useful such analysis is and, second, theoreticians saying how enormously flawed and inaccurate such analysis is.

So it’s perhaps no surprise that it’s take up is still fairly limited, prompting questions such as those from Emma Mulqueeny about its use and usefulness.

Having dabbled with such analysis on and off for a few years, my overall experience is:

1. Yes, the individual scoring may often be inaccurate, but as long as you are looking at large enough piles of data with a decent tool the errors average out and so overall trends are meaningful.

2. Be very wary, though, of any area in which high levels of sarcasm are deployed.

3. The very broad-brush pictures can be very useful: overall is online chatter about a brand of computers positive or negative? Has it changed much since the new range of computers was launched? Etc.

4. Frequently there is other, better, data available – e.g. public opinion poll data in politics – but even then sentiment analysis can be helpful to help you get into why people are being negative or positive.

In summary, take with a pinch of salt, apply a dose of realism – but don’t ignore completely.