didiooi / projecthutton

Using NLP, NLG and Database to query and scan well database

Hutton: A Cognitive Geoscience Assistance

Using NLP, NLG and Database to query and scan well database.

Collaborators: Dawn Jobe (Aramco Services, Sher Didi Ooi (University of Bristol), Kamal Hami-Eddine (Paradigm), James Lowell (GeoTeric), Vikram Sen (Anadarko Petroleum)

This project won the Popular Choice Award and Executability award.

Project status: Proof-of-Concept as part of Geophysics Hackathon 2017

Motivation

A geoscientist's dream (apart from having a rock library at home, or more field work to do) is to have a cloud AI personal assistant, who will listen to our requests, query a structured database and return responses in speech form and also display regression plots derived from database

Vision

How do you make money in an open access - open source world? – Everyone has the same tech, everyone has access to top notch data scientists. Future competitive edge will be about getting insight from your data – gaining insight more efficiently. We present to you Hutton, a cognitive assistant for geoscientists. He is a friend of Alexa’s, and he is hooked into your existing data management system and he is here to help you gain smarter insights, faster.

Project Description

See Tutorial Slides for how we got started
See Google Slide on our final presentation
Unfortunately demo of it was not recorded! Questions asked during Demo on Hutton skill set:
“Alexa, ask Hutton …

1. to open the well database.
2. what part of the world is the [DB name] data set from?
3. how many wells are in the database?
4. to get wells drilled by [operator name]?
5. how many wells have cumulative production over one billion barrels?
6. which well is the most profitable?

Tech/Library Used

Alexa Skills Kit, CloudWatch Manager, R Shiny, AWS Lambda, AWS S3, DynamoDB, LASReader