EKIN, a natural language voice solution for industrial maintenance

With the aim of facilitating and improving the productivity of some industrial processes, a consortium made up of Vicomtech, UPV/EHU, Ikor, IMH, UZEI and Tekniker is developing and implementing interaction mechanisms between technicians and machines to take a further step towards interaction in natural language with industrial production machines.

Thus, within the framework of the project, called EKIN, Vicomtech, UPV/EHU and Tekniker have developed a conversational assistant that allows interactive access to information from technical manuals, combining advanced Question Answering technologies and Dialogue Systems.

In addition to coordinating the project, Vicomtech has led the process of compiling and annotating technical manuals from which technology capable of interacting with them in natural language has been developed. In this line, Vicomtech has investigated how to automate the generation of dialog managers from the information contained in the manuals. On the other hand, Vicomtech has worked together with Ikor Technology Center in the optimization of activation word detection and neural voice recognition models, to be able to embed them in boards with low-cost microprocessors. In addition, the center has participated in the compilation of existing industrial technical manuals in Basque from which the most relevant terminology and expressions in the sector have been derived. They will serve as the basis for developing natural language interaction systems in Basque with industrial production machines.

In order to validate and evaluate both the quality and usability of EKIN's natural assistant in a maintenance environment, professors and students from the IMH Advanced and Digital Manufacturing Campus have come to the Tekniker facilities to assess a set of problems/situations related to robot maintenance.

The robots are associated with the information in manuals, so that, by interacting with EKIN's natural assistant, they have tried to obtain all the information necessary to solve the problems posed without having to resort to the traditional search for information in manuals.

The users, in this case, the teachers and students, have interacted in natural language by voice and the assistant has presented them with the relevant information on the screen.

This experimentation has been used to carry out the evaluation of one of the main expected results of the EKIN project: the solution for assistance in maintenance procedures based on technical manuals.

To learn more about the EKIN project, click here.

This project has been financed by SPRI within the framework of the Elkartek Program that supports collaborative research carried out by RVCTI agents in the RIS3 Euskadi strategic areas.