How JamiQ's Sentiment Detection Works

JamiQ's Sentiment Detection is simple but effective. It is definitely not as intelligent as a human but it gets the basics right and is infinitely tireless.

Given a set of keywords, the Natural Language Processing engine parses each unstructured English sentence into a syntactic structure that represents the parts of speech of each word. It also establishes the relationships between the words.

The Sentiment Detection engine then navigates this syntactic structure looking for matching keywords and detects for sentiments relating to each keyword. Sentiments are formed by adjectives, verbs or even nouns that have been scored by our dictionary of 150,000 English words. The sentiments are also processed for any related adverbs that modifies the sentiment scores.

This simple but effective approach reduces the rate of false positives dramatically. Each sentiment detected are bound to be related to a given keyword.

Human languages often describe objects not directly but via pronouns and abbreviations. Things get complicated when there are more than one key subject in the text - Sentiments may refer to the subject's attributes and not the subject itself. In some cases, the human language is peppered with sarcasm. While our sentiment detection is fast and effective, it is definitely not foolproof.

Our researchers are constantly pushing the boundaries of artificial intelligence in Information Extraction. You will see consistent improvements in our methods and these are best exhibited by the results you see in your account.