sjsufall2019-cmpe272 / sarcasmdetect Goto Github PK

1. If visualisation will be done using tableau, then what will be done using React js?
2. Which technology are you guys planning to use for developing API's for fetching data?
3. How you guys are planning to deploy tableau on Aws EC2?

I understand from the abstract that you feel social media analysis rely on past data. But you can get the current feed from Twitter, Facebook etc. And isn't it the same with news headlines as well. It contains past articles as well as current happenings. So what is the differentiating factor that will help out a business?

Also, news is a generic report about a current happening and usually does not contain much information about public sentiments about a product. How will you bring the correlation between a current happening to a product for it to be useful for the business ?

Human sarcasm is a product of human brain and we must not forget that the possibilities are numerous in which a human brain could frame words to express its opinion. There is a complex relationship between sarcasm and sentiment as a feature, which could be hard to uncover using models and algorithms.
An example of a business use case of Sarcasm Detection for identifying the popularity of a product is mentioned below where it is difficult to comprehend the actual tone of the review texts for a product.

For IPhone handset, the social media post texts could be as below:
“Wow..Apple has launched such an awesome phone..IPhone 10.. I wish I had not donated my kidney earlier for IPhone6!”

For a certain kind of crafting paper, review text could be as below:
“I love this paper so much that I made a garbage bag out of it.”

The tone of these kind of texts is hard to be figured out, since it contains quite a lot of positive words, yet there is a tone of sarcasm in it as per human interpretation.

Accuracy could be greater if the author’s information was incorporated into training data. Without contextual information (for example, author’s background info, responses, audience) consideration in the prediction algorithm and merely analyzing text content may not be appropriate to train the model.