The Baahubali fever gripped the nation almost two years back, and ever since then, the magnum opus has managed to capture the imagination of people like no other movie ever has.
While the first movie received praise for its magnanimity and story, the sequel, heavily bogged by expectations, has been lapped up by audiences across regions and geographies. Social media channels are full of opinions and the general view seems to be positive.
However, what these reviews don’t tell us is what worked or didn’t work in favour of the movie. Was it the story, or the answer to the pertinent question – “Katappa ne Baahubali ko kyu maara?”.
DataWeave, a company that provides competitive intelligence as a service for retailers and brands, has unveiled the truth. The company analyzed over 75,000 tweets and identified the sentiments expressed on several facets of the movie – visuals, acting, actors, and more.
While the actors and the director met with universal approval and praise, several viewers perceived it as a movie that has redefined cinema. Then there were the not-so-positive sentiments regarding visuals and the second-half of the movie.
While these insights seem simple enough to understand, DataWeave went a step ahead with a ‘Sentiment Analysis’ to filter all online content and deliver meaningful insights about the movie.
MarkUp spoke to Team DataWeave to unearth the insights behind this analysis and came up with interesting observations about brands, big data and more.
What is the Sentiment Analysis and how did you come up with it?
Sentiment Analysis helps brands study customer preferences at a product attribute level by analyzing customer reviews. We used the same technology to analyze the reaction of audiences globally to Baahubali 2. We divide our sentiment analysis in three parts – features extraction, features-opinion pairs, and sentiment calculation. Features extraction identifies ‘features’ that customers are talking about. In this case, we first understood the syntactical structure of tweets and separated words into nouns, verbs, adjectives, etc. This also accounted for complexities like synonyms, spelling errors, paraphrases, and noise. Our AI-based technology platform then used various advanced techniques to generate a list of ‘uni-features’ and ‘compound features’ (more than one word for a feature).
Next, we identify the relationship between the feature and the opinion. For instance, “I saw the movie visuals awesome bad climax felt director unnecessarily dragged the second half”. In this case, the feature-opinion pairs are visuals: awesome, climax: bad, second half: unnecessarily dragged. Clearly, something as simple as attributing the nearest opinion-word to the feature is not good enough. Here, we use advanced AI-based techniques to accurately classify feature-opinion pairs.
Lastly, we calculate the sentiment score, which is determined by the strength of the opinion-word, number of retweets and the time of tweet. A weighted average is normalized and we generate a score on a scale of 0% to 100%.
The biggest use case that we see for this is summarizing sentiment across multiple reviews, irrespective of the purpose of the review. When you look at a consumer brand, there are thousands of reviews about the product. We run the intelligence on these brands, and summarize the top consumer feelings for a product. So, you give us a data set of any consumer opinion and we will be able to summarize it and give a score for each of the attribute.
What are the benefits of this analysis for marketers and brands?
Along with marketers, the benefit we see is for product and R&D teams. When a consumer brand is trying to get deeper into its product strategy – it could be anything from a bicycle to a washing powder – a review would cover multiple attributes of that product. This analysis is capable of rating each of the attributes on a scale and providing insights to the consumer brand. And they can use those insights to build a product strategy. The other thing this will help with is brands’ comparison. We could help brands compare consumer sentiments for their brand versus their competitor brands right down to specific attributes of their products.
Big data has evolved from being a hot topic of discussion at science symposiums to playing a most pivotal role across industries. Do you think brands are leveraging it right?
Big data is becoming pervasive across a lot of functions within an organization. We are specific to competitive intelligence as a firm. Our target customers are retailers and brands. Now coming to our work, what we see is differences across geography. For instance, North America is very evolved as a market. There is a strong competitive environment there and very structured data. However, when it comes to India, this space is still evolving. And we’re seeing some traction in organizations using this, but there is still a learning curve. This is specific to the big data service we offer – competitive intelligence.
What kind of sentiment analytics trends do you think we should watch out for?
Over time, the accuracy of insights is sure to improve. When the AI behind this analysis processes more data, it will get better over a certain period. Sentiment analysis is just one small part of the intelligence that we can offer. The overarching trend that we can see is that the usage of these kinds of insights – leveraging massive amounts of publicly available data to drive business strategies – will go up. A part of that is the sentiment analysis which can be used to objectively analyze consumer sentiments behind a product.