XMILE-an expert system for maintenance learning from textual reports

Abstract

Software incidents are normally described in natural language (like English or Portuguese languages), because the users become free to express themselves about the incident. In this paper, we propose XMILE - an eXpert MaIntenance LEarning system based on NLP (Natural Language Processing) and machine learning techniques, that is capable of inferring the main attributes (type of intervention, maintenance action, cause and faulty zone) from textual reports of incidents. The XMILE was used on a real set of reports of maintenance incidents performed on IT systems of a Brazilian automobile enterprise, with excellent results in terms of precision and recall metrics.