January 1st, 2023,
The Coala European Project is proud to present two of its assets made available on the European AI-on-demand (AIOD) platform that seeks to bring together the AI community while promoting European values. The platform is a facilitator of knowledge transfer from research to multiple business domains.
COALA is developing a solution for cognitive assistance that consists of a composition of trustworthy AI components with a voice-enabled digital intelligent assistant as an interface. The solution will support workers that need to use analytics tools and new workers that perform on-the-job training, and the “WHY Engine” is a key component of the solution.
The two assets made publicly available on the platform are:
The WHY-Engine is a prototype to explain Prevention’s predictions being developed in COALA EU Project.
The architecture of the COALA solution for the Digital Intelligent Assistant (DIA) is structured into four layers: 1) manufacturing resources, 2) data management and integration, 3) AI-based reasoning and planning, and 4) the DIA dialog management.
The AI-based reasoning and planning layer provides the solution with artificial intelligence. In COALA, we focus on components for prescriptive quality analytics, an expert system for on-the-job training of machine operators and production line managers, a component that answers the users why questions about the AI’s decisions and the provided information (e.g. regarding the information quality) –we call this component “WHY engine”.
Main Characteristic
The WHY-Engine has a data-sharing interface for receiving the prediction value for a feature, a prediction model, and a list of all features. The explanation is computer-readable but hard to understand by humans (especially without a data science background). Therefore, the assistant’s dialog management must integrate its results into text messages for human users.
The WHY engine prototype is a public demonstrator therefore it is available for public usage. We shared the source code and deployed images of docker containers to do it. This prototype can be executed in two ways: 1. Cloning or pulling the source code and having a pythonx environment adequate (version >= 3.8), and perlxi interpreter (version >= 5.30); 2. Using docker desktop and the docker-compose to run the publicly available containers; The operation system used in the docker container was linuxxii, but the software can run in any operating system that supports python3 and perl.
The Web app was developed to simplify the interaction with the WHY Engine REST API.
Background:
The architecture of the COALA solution for the Digital Intelligent Assistant (DIA) is structured into four layers:
1) manufacturing resources,
2) data management and integration,
3) AI-based reasoning and planning, and
4) the DIA dialog management.
Detailed Description
The COALA project aims to develop a human-centered AI-based digital assistant to optimize manufacturing processes quality and training.
The COALA solution integrates prescriptive quality analytics, an AI system to support on-the-job training of new workers, and a novel explanation engine – the WHY engine. The objective is to provide a more proactive and pragmatic approach to support operative situations characterized by cognitive load, time pressure, and little or zero tolerance for quality issues. One of the key concepts involved in COALA’s approach is explainability. To be trusted by users, the assistant will have to provide an explanation of its predictions and recommendations to workers on the shop floor and to new workers in training. This is the function of the Why engine: to produce such explanations. This is an exciting new application of explainability in the manufacturing context. It will also introduce new challenges and new solutions.
The Why Engine Architecture consists of:
- The Frontend App;
- The Backend Service;
The Backend Service has 2 parts:
- The API;
- The WHY engine Service.
The API provides a way to send the files to the service and to run the WHY engine, therefore can be used by the other COALA modules for testing purposes. The WHY engine service has two stages; the first converts the classifier into an internal common model, and the second search for the explanations of an instance using the information of the internal model.
The WHY engine prototype is a public demonstrator therefore it is available for public usage. We shared the source code and deployed images of docker containers to do it. This prototype can be executed in two ways:
1. Cloning or pulling the source code and having a pythonx environment adequate (version >= 3.8), and perlxi interpreter (version >= 5.30);
2. Using docker desktop and the docker-compose to run the publicly available containers; The operation system used in the docker container was linuxxii, but the software can run in any operating system that supports python3 and perl.
COALA has received funding from the European Union’s Horizon 2020 Research and Innovation Programme
under grant agreement no. 957296