Tone Analyzer - insight into algorythms
With a few fellow IBMers, I am participating in an IBM internal Cognitive Learning Lab. In our ' graduation project', we would like to combine the Twitter API, Tone Analyzer and a Visualization tool to show (consumer reactions to) outages at our customers, e.g. cable companies, banks, mobile service providers.
We are running some experiments and we have trouble explaining the output of Tone Analyzer. There seems to be always a lot of 'fear' and 'anger', even when entering tweets related to for example ' Mother Theresa'.
Is there some documentation about the inner workings/heuristics/algorithms/AI of Tone Analyzer?
Thanks,
Pascal ter Horst
The "Science behind the service" section of the Tone Analyzer documentation (http://www.ibm.com/smarterplanet/us/en/ibmwatson/developercloud/doc/tone-analyzer/science.shtml) discusses how the service works and provides links to research papers and other publications that provide insights into how the service works. If that information does not answer your questions, please identify the other types of information that you were hoping to find.