Natural Language Processing for Troubleshooting

Natural Language Processing for Troubleshooting
Summary

The Experiments is conducted by Intellimech Laboratory. 

It provides a troubleshooting tool capable of helping both operators and enterprisesindiagnosis and maintenance procedures.

This troubleshooting system will be divided into the user interface subsystem, the elaboration subsystem, and the knowledge subsystem. The user interface is responsible to handle the interactions with the operator, the elaboration subsystem processes the operator’s inputs to find the most probable failure and the knowledge subsystem represents the available data that the system refers to for solving the incoming issue. The experiment is aimed at improving the current troubleshooting system, including its basic user interface that is currently based on closed-ended questions, and its limited source of knowledge. NLP based tools will be exploited to achieve a troubleshooting system featuring advanced human-machine interaction based on free text to speech dialogue, capable of accommodating mistakes and with a self-learning mechanism, that will automatically enrich its knowledge overtime by analysing the ongoing issues.

Moreover, three possible scenarios have been identified in the experiment. The first scenario involves a smart contact centre, that processes and redirects the inquiries to the corresponding expert. The second one covers a troubleshooting system that is able to deduce the damaged component. Finally, the third scenario includes the failure mode, by conducting a dialogue with the operator.

Results type(s)
Unfold all
/
Fold all
Attached files
File Type
EXP-02-IMECH.pdf PDF
More information & hyperlinks
Country: IT
Address: Via Stezzano 87, Bergamo
Geographical location(s)
Structured mapping
Unfold all
/
Fold all
Comment:

The troubleshooting system implemented in the Experiment is an AI-driven advanced solution that improved an already existing system, totally deterministic. The new solutions instead is based on a probabilistic approach. Compared to the previous one, it is:

  • more robust: each answer modifies the probability associated with each component without deleting or discarding any of them
  • more flexible: the system also includes the ability to skip a specific question
  • more efficient: the tool can determine the best question based on the probability associated with each component.