By Jaskarn Singh & Jitin Khanna
Sentiment analysis has become very popular especially in social media context where lots of use cases are there, which require to learn the sentiment of tweet. The major application of sentiment analysis is applicable to product reviews, political opinions, movie reviews, and even health related trends. The source of such reviews or data could come from varieties of applications such as E-commerce websites, news journals, and most importantly from social media applications (Twitter, Facebook, etc.). Among all these, Twitter has turned out to be the most popular source of data for almost any topic in the world. So, having an application which could tap into a stream of Twitter topics and provide sentiment of the incoming stream could be a valuable solution for many business use cases.
Current challenge with having a stream analytics application running in production environment is time consuming, expensive and difficult to maintain due to the fact that deploying a good model into production involves having right form of infrastructure and quality models. This application trying to solve both problems that are complexly outsourcing the part to Cloud services and making it easy to use to end users with easy to use application for most common use cases.
Although, Python has inbuild support for NPL and used heavily among the Python community members, the availability of cheap cloud APIs has made it very attractive and easy to use state of the art Algorithms when it comes to any kind of machine learning or sentiment analysis.
Before we could even use these libraries, the biggest challenge will be to parse the tweets and make sure they are ready to be consumed by the API. Few limitations are very common, especially when it comes to sentiment Analysis such as Language issues. Most sentiment analysis libraries only support small percentage of languages notably English, French and Chinese (Google). So, if the tweets are in some different language the library will simply not work or throw errors.
As we are considering streaming the data rather than storing and analysing it later, we have to do the data preparation on the fly. Fortunately, Python has rich set of string manipulation libraries which could help us with these issues. There are plenty of sentiment analysis APIs and libraries available; however, later Cloud that provided APIs have become norm across multiple application to get the best possible outcome.
About the authors:
Jaskarn Singh is a researcher, inventor, innovator, consultant, leader, coach, academician. Over the years he has engaged with Top Healthcare consultants and select top Fortune 500 companies. He is a leading authority and speaker on innovation, design, cloud computing, big data & analytics, future technology and user experience.
Jitin Khanna is a Data Architect and SAP BW/4HANA, S/4HANA Solution expert with 13 years of Exp. He has independently implemented SAP BW, FIORI, Analytics for Cloud, Predictive Analytics with more than 10 clients across North America, EU and Asia. His current project is to integrate S/4HANA with Qlikview and BW/4HANA and get custom data using Hana API and making a product. He has vast experience of SAP Analytics implementation using Core data services as well