A Product manufacturing company
Case Study : Mobile Sensor Data Analytics
Executive Summary
An intelligent analyzer which extracts and aggregates large volume of customer reviews in unstructured English text from social media and finds key aspects and topics which are driving their sentiments and ratings about a popular smartphone. Publicly available review data has been used as review corpus.
Review comments are in unstructured informal English – difficult for machine to understand
Critical to find out the key aspects and topics driving customer ratings from large review corpus
Large volume – increasing every day with ever increasing online presence of customers
Almost impossible to extract and analyze manually
Automatically scrape reviews from websites
Automatically extract aspects and related topics in order of impact on customer reviews and sentiments
Visualize in phrase cloud
SkyBits Solution Approach
Key features developed are:
Automatic Web scraping to extract data from websites
Aspects are extracted from corpus which have maximum impact on customer sentiments
Subsequently hidden topics are extracted which are related to a particular aspect
Phrases are extracted which drive the sentiments (aspect – topic interactions)
Aspects are dimensions of text corpus
Visualizations in phrase cloud
Machine Learning Algorithms used to address the key challenges
LDA (Latent Dirichlet Allocation) for extracting topics, aspects
NMF (Non-negative matrix factorization) for ranking the aspects
NLP for extracting the n-grams from review corpus
‘R’ programming language
