Here is a Google Ad that appears on my Blog on Barack Obama. The Ad is exactly opposite and demeaning to what I write about him. All of my blogs on Obama has mostly praises and appreciations with some constructive criticism on him. So, instead of similar sentiments, what does the automated Google Ad show? "Obama exposed!". Is their Grammar analysis tool all screwed up? I think their Researchers are missing the point with counting words and index driven popularity rankings. They are efficient in analyzing single Sentences but seems failing miserably on how Sentences join to create one Sentiment. How difficult is to implement the following?
Read the text (Like the blog(s))
Find the Subject (Like a Person)
Break the text in individual sentences
Create Two maps - Positive and Negative
Analyze each Sentence, Rank the words and fill the maps.
Look for words like "All I said", "Aforementioned", "Contrary to what I said", etc
and add a weight to the Map entries.
Count the entries in the map.
If one is higher than the other by 80%, then find a Positive Ad or vice-versa.