All businesses receive a big number of messages from different sources and human operators must manually respond to repetitive similar queries from different business domains that requires different knowledge and skills. To automate the task, we prepared a platform utilizing natural language processing (NLP) to classify incoming messages and to answer them automatically.
Our platform based on historical data "learn" to understand context of a request written in natural language, classify the request into categories, pull data needed for answers from internal or 3rd party sources and prepare text of the answer to be automatically sent or proposed to human agent for sending.
System is ready to be deployed on-premise or as a service in cloud. The application contains the configuration part which allows definition of answers to the individual message types. The system is designed to support multiple languages based on your needs.
Solution prepared ready-to-send answers for 75% of received messages (measured in PoC)
Architecture Building Blocks
Language Detector detects the language of the incoming messages to pass to correct modules.
Text Understanding understands the written text, classifies into defined categories.
Entity/Intent Recognition extracts important entities from the incoming message, provides sentiment analysis.
Data Sources/Batch Import pulls incoming messages from external sources automatically (e.g. email servers) or allows a batch of incoming messages to be sent or imported into the system.
3rd parties Integration allows to pull necessary data from the back-office systems, suggests and performs follow-up actions in the back-office system.
Data Manager orchestrates all data flows in the system, sends incoming data to AI modules.
REST API Interface provides all necessary functionality for returning the results, suggests possible responses to the incoming messages.