In the previous video, our CEO Riccardo Osti explained the pros of human text analysis, which can be summarised into ease of getting started, the ease to reach good quality and our ability to identify irony and other anomalies.
Now let’s talk about the limitations that humans have when analysing texts:
This is one of the biggest problems that we have. In fact, as humans we have emotions, and we evaluate things differently, following our mood. It’s clear that we will hardly derive any scientific result from manual analysis.
Dictionaries can become very specific, especially when we need to analyze data about different products in various industries. I have seen dictionaries with more than one hundred topics, and you can imagine how difficult would be to remember all of them, and their description while we go through the records. This task, that is clearly difficult for us, is instead simple for a machine, that can potentially remember infinite ontologies.
Human analysis can be very slow. Well…that’s pretty straightforward, in fact, humans can analyze 10 to 12 records per hour on average. Machines, once configured can process millions of records in the same amount of time.
So what is the conclusion? Having humans is a good exercise to understand what your customers say about you and your products, however, as soon as we grow the complexity or the volume of the analysis, all kinds of limitations arise.
The lack of consistency in our behavior, but also our limited memory, and our low speed make the automated analysis more attractive, especially for brands that intend to do text analysis seriously, at scale.
We look forward to hearing your point of view about human text analysis.