While sentiment analysis may thrive well in identifying, extracting and reporting on personal information from the written materials, there are some challenges to encounter. Since we have most sentiment analysis systems automated, there is the problem of trying to decode contexts in the public face. Some of the reasons automated sentiment analysis fail to produce good results in the public face include:

Rigid programming               

Developers program most of the automated sentiment analyses to identify a particular group of words as either negative or positive. These automated machines lack the ability to identify sarcasm. For example, in a sentence like “oh she’s not here yet? Excellent I love waiting so much” the machine will recognize words like “love” great” and awesome and flag the sentiment as positive while it is, in fact, negative. Humans are capable of understanding sarcasm, but machines cannot.

How to solve the machine problem?    

In such a case, the best move to resolve the machine problem is to halt reporting the review results from the computer. You can present the faults found to a board panel and convince them that the automated sentiment analysis is not a reliable when sorely judged by machines in the spheres of public face.

The remedy    

Doing a human driven sampled sentiment analysis using statistical tools that filter certain specific words; like in a sample of 5 million tweets, you can filter the name of your company and sample only tweets that are talking about your business.

Lastly, you can tally the positives, negatives and neutrals having put into considerations the sarcastic comments. This method could work to efficiently solve the automated sentimental analysis problem in the public face.


Comments are closed.