Success Stories

Analyze

Solve

Deploy

Succeed

To optimize data from different languages, we clusterized the data by DBSCAN and extracted meaningful clusters by a set of keywords or keyword sequences.

Automatic Template Prediction System
Background

The system is based on analyzing textual information of incoming e-mails and suggests the most suitable automatic answer (template message).

Challenge

A2C Cloud needed to develop a system that supports more than eight languages and offers the ability to remove redundant data/stop-words from e-mails, create template messages, etc. It had to be an independent application that linked other systems within the shared network.

Solution

Our team created a Java-based application to modify clusterization results, and provide the ability to create a new cluster, associate some existing data to the new cluster and apply filters. 

The next step was to develop a python-based application with restful a API for integration with the client's infrastructure and schedule, to train the new models when there is enough data.

When the system accumulates enough data for training, it generates a new model that continues to improve prediction accuracy.