Development of an automation system for processing maintenance requests based on NLP methods

Presents the development and implementation of a software system for automating the processing of requests for maintenance of corporate infrastructure using artificial intelligence methods. A modular solution architecture is proposed that integrates machine learning models for hardware type classification and named entity extraction (NER) from unstructured text messages. The key stages of implementation are de-scribed: preparation of a marked-up dataset, training of transformer-based models (RuBERT), building an API interface and integration with ITSM systems (using the example of the Jira Service Desk) for the automatic creation of service and diagnostic requests. The effectiveness of the solution was evaluated using the metrics of accuracy, completeness and F1-measure, as well as an analysis of its advantages compared to manual processing. The developed software system demonstrates high practical applicability for reducing incident processing time, minimizing errors and freeing up the resources of technical specialists.

Authors: A. G. Glushchenko

Direction: Informatics, Computer Technologies And Control

Keywords: system analysis, information processing, artificial intelligence, machine learning, application processing, entity extraction (NER), ITSM integration, RuBERT

---

View full article