Case Study

Reducing 95% costs by automating data-labeling to create training data

Published on 05th January 2022 | By Jaxon

For

A top 10 US retailer

Tags

Automated Data Labeling
Training Machine Learning
Creating Synthetic Data

Challenges

Jaxon’s client was manually looking at over 400k return comments in order to:

  • Predict and prevent returns
  • Suggest resolutions for customer service reps
  • Classify customer intent in chatbots for Sales and Customer Service

  • Solutions

    Jaxon trained natural language understanding (NLU) models in support of the client’s chatbots and conversational AI systems. A typical NLU attempts to classify intents (what is a customer trying to accomplish?) and slots (easy to answer questions such as location) contained within user utterances. Training data, quickly provided by Jaxon, consisted of sample utterances that were annotated to identify these intents and slots. Because Jaxon used the company’s own data and did not require fully pre-labeled sample utterances, it was ideal for training domain-specific NLU models.

    Key Metrics

    82%

    Achieved 82% ML accuracy in a 9% bump over the custom ML model

    95%

    Achieved 95% reduction in overall cost

    Similar Case Studies

    For
    download
    Industrial Automation
    Predictive Maintenance
    Reduced Machine Downtime
    Condition Monitoring
    Real-time Monitoring
    For
    A Fortune 50 brand
    Customer Data
    Data Analysis
    Automated Insights
    Behavioral Analysis
    Ethical Facial Analysis
    For
    Employee Productivity
    Employee Engagement
    Employee Training