/ciol/media/agency_attachments/c0E28gS06GM3VmrXNw5G.png)
Follow UsPUNE, INDIA:
Can you help us understand how novel the whole concept is and how Persistent leveraged its repertoire to build this platform?
Viewer engagement analytics is a very novel concept in India until now, most broadcasters and shows rely only on the TRPs and the hearsay about the impact of the show. Satyamev Jayate was a pioneering show not only in content but also in analytics. The concept is to gauge the impact of the show through analysis of the responses that viewers sent out. The source of the responses included social media platforms like Facebook, YouTube and Twitter, along with the Satyamev Jayate website and SMS and IVR responses. Persistent, being a leader in Big Data was aware of the technical challenges.
What were the key mandates?
The key to success was to gauge the exact sentiment of the responses, along with the ranking of the responses in order for them to be featured on the website and to be shared with the production team. Given the fact that the respondents would also use non English languages to respond adding to add the multi-structured and non-textual (audio and video) responses made us opt for a hybrid approach. In this approach we chose to use technology as well as human intervention to analyse the responses. The next challenge, given the hybrid approach, was to build a system which would allow us to analyse the data seamlessly both through the technology tools as well as human analysis cycle. In fact, the Content Filtering, Ranking and Tagging System (CFRTS) which is a multi-user, multi-media response analysis system also has a provision to analyse the analysis speeds and the data pipelines. The data analysed there would once again be analysed in the second level analysis to ensure that the tagging is correct.
How much of it is about data drilling, data visualization and how much about intelligent analytics?
It was about 50-50. Data gathering/visualization was only the first step. Value was added in terms of meaningful insights.
How close are we to Siri, AI and supercomputing levels of real-time interpreted intelligence?
Not very close, we are using text analytics techniques specifically for handling Hinglish, Hindi and English. The goal was to automatically tag or classify the content into a set of values; most of these techniques are custom built using rules.
Where exactly did the show benefit from a customised platform vis a vis any other tool picked from the market?
Customized platform allowed us to go much deeper in terms of sentiment analytics vis-Ã -vis any text analytics tool. No Natural Language Processing tool can process Hinglish (Mixture of Hindi and English as a language). A sizable proportion of the responses have Hindi words interspersed in English or vice versa. The customized platform with human intervention ensured that the finer contours of the sentiments were captured in tagging. Plus, there is more to emotional expression than just the sequence of words — only human beings can understand the deeper hues of emotions.
The fact that this was a TV show meant other factors too?
Yes, given the time between the show on Sunday and Asar (the show on Friday) where impact stories were needed, fast processing was key. Implementing newer revisions to text analytics rules takes time to implement and test, with the human approach — half an hour of instructions is enough to change the course. Also, the show topics were closely guarded and couldn’t be disclosed much in advance which meant that the tool based approach had challenges. Contextual comments and comparisons with other shows/ insinuations about current affairs etc. cannot be programmed within time. For example, It is ironical that Skoda ad plays during the commercial break. (This comment was in reference to the Skoda Rapid TV commercial based on Big Fat Indian Wedding)
The show is about change. Did the platform help in creating 'impact' points?
Yes. The platform enabled us to cull responses which needed immediate attention — for example, people not able to donate online or when the show did not have subtitles in some languages. In some instances respondents were directed to appropriate help by the Production team based on the contact information shared by the respondents. As insights partners our job was limited to sharing the insights with the production team — actions had to be taken by them.