Analytics Backend

[BasicallyOk]

Sentiment, word cloud, semantic network graph

[BasicallyOk]

Implementation

• Word cloud. Return k most common words (with pre-defined weights based on some language distribution). Uses the english word frequency database to assign weights for common words. Should also theoretically filter out words like "the", "a" or "and" since those would have bad weights. This also makes integrating any language a matter of finding a word frequency database for that language.
• Sentiment. Unsure, we can either assign each submission a pre-defined label, or we simply show how many other people shows similar response based on its sentiment encoding.
• Semantic Network Graph. Or just any network graph, needs a little more cooking.

[BasicallyOk]

Word Cloud is handled by #32. The algorithm finds important word by using a language distribution. For every word, we compare its language frequency to the most common word in the language and use that to weight the sampled frequency in the chat. By doing this, only less common "situational descriptors" will end up with the highest weight frequency. Moreover, this allows the usage of any language provided a language distribution.

[BasicallyOk]

Closing this, splitting into more detailed issues with #34 and #35 for sentiment analysis and semantic network