• New products appearing from time to time, while 50% of them die within 5 years.

• Traditional questionnaires and surveys are very time and human efforts consuming.

• Efficiently gather and utilize customers’ immediate feedback, assist better decision-makings and product roadmap.

• Building sentiment analysis model based on real time feedback gathered from social media, also including their major complaint aspect

• Source Data: (https://archive.org/details/twitter_cikm_2010)

1. Data Preprocessing - eliminate noises
   • retweet entities
   • @people
   • punctuations
   • numbers
   • html links
2. Sentiment Analysis Model
   • positive vs negative polarity scores
   • emotions: joy, anger, disgust, fear, sadness, suprise, unknown

A much fancier interactive visualization is here

• Replace downloaded Tweets historic dataset with Twitter Stream API;

• Create some labeled training dataset to develop more accurate sentiment analysis algorithm;

• Develop text analysis algorithm to extract key information regarding product and service problems from Tweets;

• Improve the visualization tool to embed streaming data and more analytical features.