Emotion Analysis of TV Show Content
Integer1 is a first of its kind content analytics platform for the media industry. The platform has in-depth analysis of more than 300 popular TV shows across all the channels in the US for it’s emotional signature and emotional resonance with the audiences. All popular TV shows have a strong emotional connect with the audiences and integer1 platform measures precisely that emotional engagement level. Higher emotional engagement translates to higher viewership, popularity and ratings on social media.
The platform is used by media companies and TV channels to make decisions on which shows to air and when. This is done by comparing the emotional similarity of the new show against the existing shows. Content producers, media houses and TV channel companies can then make informed decisions on how the content has to be altered to resonate with the audiences.
What we do
- First of its kind content analytics platform for emotional analysis of TV shows
- Correlates show’s emotional resonance with perceived audiences’ emotions
- Compares different shows (and new shows) for emotional signature to make informed decisions on which shows to be aired or ads to be inserted during the show
There are several independent processes that make up the entire ecosystem. An enumeration of the key ones is below –
Data Ingestion – All analytics begins with data. A data ingestion module was developed to fetch TV shows’ script data from various sources. This was stored as text files with segregation for seasons and episodes. Developed in Python as a daily job with logic to read only the newly added shows/seasons/episodes.
- Emotional Analysis – Text analytics library to derive numeric emotional values of the TV show content. Each episode was split into 10 parts giving us a good indication of how the emotions have progressed through the episode. This library was built in Python.
- TV show metadata – A separate interface library to connect to 3rd party subscribed services to fetch metadata for TV shows. The data was dumped to a MySQL database. This too was a Python library executed as a daily cron job
- Visualization – Stunning UI with a carousel type landing page, show details page, search and to compare shows for emotional resonance. All this was developed in Angular and charting using HighCharts.
- Both frontend and backend is deployed on a dedicated AWS EC2 instance. Frontend was built in Angular 2 stack while backend was a Java Spring boot based server application
- AWS RDS MySQL for all metadata of the shows
- S3 as file storage for show scripts
- Python cron jobs for ingesting TV show content