Customer Sentiment

Real-time sentiment analysis brings higher engagement for USA-based company

» Real-time sentiment analysis brings higher engagement for USA-based company

Las-Vegas based company is a multi-linear data aggregation portal for verified business information for small companies across industries seeking capital from equity, banks or other businesses.

Business Need

The data aggregator company wanted to create a proof of concept for a tool to be used by a sports organization. The need was to mine real-time user sentiment from various social media streams about NBA, NFL, baseball and school basketball and college football league matches. Positive, negative or neutral sentiment for each team would help from coaching decisions to fantasy sports.

Millions of viewers posted about the on-going matches on social media channels like Facebook, Twitter and Instagram. However, due to the heavy influx of unstructured text data generated every minute, actual analysis and mapping actual sentiment was difficult.

Inability to comprehend real-time and voluminous social media data and need to analyze customer sentiment led to a partnership between the data aggregator and Hitech Analytics.

Challenges

Tapping in multiple social media platforms for comments, tweets, retweets and captions.
Managing a massive flow of sentiment data from viewers.
Identifying the underlying context and segregating the data under positive, negative and neutral sections.

Want to transform unstructured text data into structured form to harness valuable insights?

Get in Touch with us »

Solution

We built a sentiment analysis tool to capture and stream text data from multiple social media platforms daily to perform sentiment analysis. Using advanced natural language processing and AI/ML techniques, the data was visualized with intuitive dashboards that helped the client to make strategic decisions.

Approach

Hitech Analytics team identified the list and schedule of sports matches to be tracked regularly.
After understanding business objectives, our data specialists defined the polarity score ranging from -1 to +1 that also factored in subjectivity and sarcasm.
We built a data extraction tool to fetch data from various social media platforms on daily basis. Data was collected, cleansed, transformed and stored in a centralized database for further analysis.
The team ensured that all the data was annotated using ML models based on the given KPIs and prepared it for training.
Multiple ML and recurrent neural network language modelling models were built and trained. The team selected a relevant model based on predefined accuracy parameters.
They created a data visualization to help the client analyze the results every day and draw actionable insights.

Results

100% real-time visibility of public sentiment on customer preferences
85% reduction in manual survey spend for strategic decisions
97% accuracy achieved in analyzing social media sentiment to drive better marketing activities