“I think that’s what she meant by that…” is a common question of sentiment. Such misinterpretation invades everyday lives. And innovative organizations are supposed to trust a machine to solve these dilemmas? It can be tough to do. The following excerpt from the white paper Text Analytics for Social Media, by Jim Sterne, gives insight on how systems can help the misinterpretation problem. Stern is an internationally known writer and speaker on electronic marketing and customer interaction.
Sentiment analysis is the most difficult task of all. Humans often have trouble understanding each other, even when speaking face to face. Without facial expressions or vocal clues, textual misunderstanding is common. That is why legal writing, for example, is excessively dense. It is an attempt to avoid misinterpretation. The problem of misinterpretation has challenged technology for years. In the past, enthusiastic experts were convinced of the power and sophistication that high-level math brought to bear on the problem.
The answer actually lies in humans working closely with machines. Computers can calculate that a given phrase is positive or negative with a certain degree of confidence. Humans with specific domain expertise can review low-confidence results and advise the machine how to grade them. Over time, the computer absorbs more and more of the expert’s perspective and becomes more accurate and useful.
But even before the domain expert offers the first course correction to a text analytics system, the system has some distinct advantages that make it a necessity.
Explore those advantages in the white paper, Text Analytics for Social Media.