Conversation Classification at its Finest
A Hugging Face & AWS Case Study
Summary
Kustomer is the omnichannel SaaS CRM platform reimagining enterprise customer service to deliver standout experiences.
Built with intelligent automation, Kustomer scales to meet the needs of any contact center and business by unifying data from multiple sources and enabling companies to deliver effortless, consistent and personalized service and support through a single timeline view.
Kustomer IQ is a set of tools that contextualizes conversations, eliminates time-consuming tasks and deflects repetitive questions, so everyone is less distracted and customer services reduce cost per contact.
Challenge & Solution
The process of classifying and analyzing incoming tickets is currently a manual process for agents. This of course is time consuming, error-prone and not utilizing agents to their best abilities. Their Conversation Classification pipeline makes agents more efficient through automation. Kustomer’s current feature is made of three steps:
- 1. Clean up existing data
- 2. Choose the conversation attribute that you want your new custom model to predict.
- 3. When active, the model detects specific language in all first inbound email conversations and automatically makes predictions based on this data.
Kustomer model training runs in SageMaker by fine-tuning a pre-trained model, using Hugging Face transformers. Kustomers current Conversation Classification pipeline fine-tunes BERT-base models using their customers data, more specifically, Hugging Face's official bert-base-cased or bert-base-multilingual-cased (for orgs handling messages in languages other than English).
The training and evaluation process relies on Hugging Face transformers, and the inference is then served by SageMaker endpoints.
In other features they also fine-tune other *compressed* language models versions, such as distilbert-base-cased, distilbert-base-multilingual-cased and Distilroberta-base.
After training, a formal evaluation is conducted. Upon activation, the model is deployed as a SageMaker endpoint.
About Kustomer
Kustomer enables customers to resolve easy questions themselves, and empowers agents to deliver convenient, informative experiences that build loyalty. It's a new type of CRM for customer service that drives smarter data-driven processes so you resolve more conversations quickly.
“If you are a brand where responsive, personal, differentiated customer service is a priority, you have to use Kustomer. There is nothing close to it on the market. It was like they read our minds.” — Leslie Voorhees Means, Co-Founder & CEO, Anomalie
Benefits
The Benefits of a Hugging Face & AWS Backed Solution
Resource Efficiency
- Freeing up valuable time for Kustomer resources to focus on their core business tasks.
Scalability
- Using Hugging Face & SageMaker allows Kustomer to move away from a manual workflow, to a fully automated pipeline.
Cost Savings
- Combining an automated workflow and resource efficiency allows Kustomer to save costs.
Hugging Face & AWS, a Better Together Strategy
Why Hugging Face?
Hugging Face allows Kustomer to easily test, train and deploy models for all their Machine Learning use cases. Kustomer chooses Hugging Face due to their reliability and wide range of the models available on the Hugging Face hub.
Kustomer huge fans of open source and the community around Hugging Face. Hugging Face is making transformers - which are traditionally hard to use - easy to handle.
When it comes to building custom text classifiers, fine-tuning a pre-trained generic Hugging Face model with their customers’ own data is an easy process.
Why AWS?
With high-performance compute options powered by machine learning, Amazon Web Services (AWS) enables organizations to undergo broad digital transformations with modern, cloud-native solutions. Improved processes, increased efficiency, and accelerated innovation are just some of the benefits realized from the inclusion of machine learning in business operations.
Offering a broad set of machine learning services and supporting cloud infrastructure, AWS enables organizations to tailor their machine learning solution to meet the unique needs of their business. Organizations are already realizing great value from AWS, enabling them to provide new experiences for their customers and drive business growth.
Improved processes, increased efficiency, and accelerated innovation are just some of the benefits realized from the inclusion of machine learning in business operations.
