When a product is getting launched on the market, obviously the manufacture would like to understand the reception of the product among the end user. The tried and proven method of analyzing the pulse of the end user is through end user review surveys / feedback survey over a phone call yesterday. It is a time consuming process and sometimes it is not reliable as the calling agent / survey agent would fill their own responses to meet the targets.
To avoid this ambiguous situation, currently all e-market portals like Amazon, Flipkart etc enabled the customer to share their reviews / feedback on their web application portal wherein the reviews are genuine. To understand the mood / sentiments of the customer about the product, there will be a dedicated team which will be going through each of the reviews and segregate the product & service reviews. This may be easier if you have listed bundle of products but it is complex and almost not possible if you have listed thousands of products on the website.
RPA is all about quick and accuracy. so to analyse the sentiment of the end customer, we have built SAB (Sentiment Analysis BOT) which would be a USP of this solution. It would analyse the entire list of customer reviews which are associated with the particular product and will give the polarity analysis using Python NLTK package. It will also give you percentage split on positive & negative reviews against the total reviews.
How we built it
Python Package - TextBlob supports a function called 'Sentiment' that produces two outputs(Polarity and Subjectivity) for each input given. Polarity defines if the customer has good/bad opinion about the product. It has its value ranging from -1 to +1. '-1 to 0' indicates negative opinion about the product and '0'indicates neutral opinion and '0 to 1' indicates positive opinion about the product. Subjectivity indicates if the input is factual or opinion. In our case we have made use of only polarity which is sufficient to analyse the sentiment of the customer with the feedback they have given. For the demonstration purpose, we have used reviews from an online market and arrived at the polarity of the review using sentiment function.
Also the bot has the ability to pick up the negative reviews and further bucketize them into Service issue customer feed backs & Product issue customer feed backs basis the configurable keywords which are added to the dictionary.
This makes the manufacturer to directly analyse the Product specific feed backs and help them understand the mass market opinion. At the same time, service provider can also get the pulse of the customer on their delivery methods.
Challenges we ran into
- Python code integration with UIPath
- Routing negative reviews and bucketize them as Product & Service related customer reviews.
Accomplishments that we're proud of
- Was able to complete the task before the timeline
- As an amateur professional, Hackathon gave me the opportunity to explore the different functionalities of UIPath tool.
What we learned
- Learnt Python.
- New functionalities of UIPath which I have not used before.
- Python & UIPath Integration.
What's next for SAB - Sentiment Analysis Bot
- Considering this is more of a POC, we have chosen one specific product and analysed the polarity and segregated the Product & Service review. Down the line, any products can be configured without any major coding effort considering online market websites list thousands of products.