Code with SAP Labs India - Discover, Design, Deliver

Data for Storytelling

Overview: As per Gartner, the analytics and business intelligence platform market has transitioned from the visual data discovery era to the augmented era. Data and analytics leaders should begin piloting capabilities that enable the “augmented consumer”.

With the technology advancements, the organisation today has the advantage of taking data driven decisions and create their planning, forecasts based on the same. Many business users do not have time to analyze the data and then get insights. And there are gaps between how the tool produces an output and how the business user can use it to interpret it. In addition, it needs a good domain knowledge to build business insights from data. Not every user is a business expert. Given a snapshot of data, we would like to build a system which can tell a story from the data. The story includes the automation in the sense of being driven by the data, context and personal preferences. In this case this solves both the problems of the tool usage as well as guiding the user with the data driven intelligence to make business decision. The whole result is driven by outcome and effectiveness.

Challenge: Propose an AI based tool that can take a data set, identify patterns in the data, can interpret the result, and can then produce an output story that is understandable to a business user based on the context. System should be able to pro-actively analyse data on behalf of users and generate smart feeds using natural language generation techniques which can then be consumed easily by business users with very less efforts; the story can be told in any combination of plain English supporting visualizations, which however should be kept rather elementary given the intended target user.

Task: Given data/analytics output, create an AI based tool 

• that is able to turn the data into interactive data stories.
• has a generic architecture; that is, the tool should not be hardcoded to any specific input data. However, may be given additional meta data describing the structure of an input data set (such as descriptions of the columns and how to treat the values, structural or logical relationships between columns, etc).
• to generate deep insights, infer pattern and help in business decisions.
• have personalization profiles; these could be represented as meta data describing what would be of interest to a given user.
• understandable to a business user.
• with interactive and intuitive interface.
• [Bonus] to have context and interactive personalization.

Data:

Please use any publicly available dataset where the fields are labeled. For example:

• https://www.kaggle.com/manjeetsingh/retaildataset OR 
• https://www.kaggle.com/unsdsn/world-happiness OR  
  • (World Happiness Report - The reports review the state of happiness in the world today and show how the new science of happiness explains personal and national variations in happiness; you can consider year over year and trends across all years).
• Any dataset of your choice from the publicly available dataset.

Overview Session Recording

Click here to view the PPT for understanding this problem statement