Figuring out who is a human and who is a robot

Gatsby
  • Text analysis
  • Pattern recognition
Problem

Problem

Separating human users from machines is tedious and takes a long time

Solution

Solution

Automatically recognizing human users based on their behaviour data

How does it work

Results

  • No time wastedAutomated human user recognition saves time when humans don’t have to look through data
  • Increased accuracyLess errors in the identifying process
  • Behaviour data guides developmentWhen human users are found, the service can be better developed based on correct behaviour data

Filtering human users for better developer experience and added business value

Gatsby offers a command line tool to developers, which is also utilized by a significant number of machine users. In order to develop their service experiences and better understand the use cases, Gatsby needs to know which use sessions are created by humans, and which by machine. The two can be identified from session data, but because of high user volumes, it is extremely tedious for humans to do.

In order to make this recognition easier and more efficient, we developed an automated solution for Gatsby to recognize human users! Our solution is powered by machine learning and it analyses the Gatsby command line tool session data. More specifically the model analyses user behavioural data, run environment information and mistakes in language usage, latter of which is a very accurate way to identify a human user. Output of the model is an estimate on which users are likely humans. This information is utilized by Gatsby to develop their user experience further to suit human users’ needs.

Summary

What we did
  • Encoded user behaviour into embeddings using deep neural network based model
  • Visualized the data and helped Gatsby to understand their users
  • Built classification model for categorising user behaviour
Data science
  • Keras
  • Tensorflow
  • Numpy
Language & Tooling
  • Python
  • Apache Airflow
  • Docker
  • Kubernetes
  • Google BigQuery

Ask for more

Teemu Heikkilä

CEOteemu@emblica.com
Teemu Heikkilä
Do you have a probem you would like us to create a solution for? Contact us and lets talk more!

See more projects