Predictive Text Classification for a large AdTech Company

Predictive Text Classification for a large AdTech Company

Location
  • the USA
Industries
  • AdTech
Technologies
Client

The client is a large and innovative big data management agency from USA that collects valuable first-party data on different websites and offers ads to visitors based on their tastes. The company works mostly with musicians selling its bid data services.

Business needs

The client already collected data from simple websites with one niche – music. However, there was a need to expand the focus and analyze complex websites, where one can find lots of different informational categories and target audiences of various ages. The categories are eligible to extract valuable textual data about website visitors and allow to serve advertising in a more accurate way. Thus, the client needed to have a model, which could analyze millions of these web pages per day and categorize the content precisely following the valuable context results.

inoxoft Solution
Inoxoft has supported the client’s intentions and offered him a hand of help by delivering an ML model to perform categorization. Every word was given an IAB category in the model that could analyze websites and predict consumers’ preferences further. The model was developed as a multilingual classifier giving the possibility to keyword websites written in any other language.

imageProject duration
1 Year
imageTeam Composition
  • 1 — Project Manager
  • 2 — Data Scientists
  • 1 — Machine Learning Lead
imageTechnologies
  • Tensorflow
  • Selenium
  • Docker Compose + AWS
  • Scrapy

Predictive Text Classification for a large AdTech Company
Predictive Text Classification for a large AdTech Company
Key features
  • IAB categorization.
  • Custom Python Scraping Backend.
  • Custom Tensorflow training and model deployment.
Challenges

Data collection and labeling of data was initially a blocker, but with some creative thinking we were able to collect a robust dataset that was suitable for training a powerful categorization model.

Results

Inoxoft’s team has produced the first-class machine learning model for the client allowing him to:

  • Analyze complex websites with different content.
  • Process 2 million websites per day.
  • Obtain the inference in under 10 milliseconds.
  • Get precise categorization of words and context.
  • Successfully sell the agency’s services by promoting the right ads to the right audiences.

Thank You!

Thank you