Release Notes: OMIn Dataset v1.0.0

We are excited to announce the release of the Operations and Maintenance Intelligence (OMIn) Dataset, a key contribution from our recent research publication. OMIn is the first open-source dataset curated specifically for knowledge extraction (KE) in the operation and maintenance domains.

Overview

The OMIn Dataset is based on raw FAA Accident/Incident data and has been meticulously curated to support KE in operations and maintenance. It features detailed textual descriptions of maintenance incidents, with particular emphasis on mentions of aircraft systems and domain-specific shorthand. This dataset includes:

• Gold Standards: Prepared for Named Entity Recognition (NER), Coreference Resolution (CR), and Named Entity Linking (NEL).
• Textual Descriptions: Rich narratives of maintenance incidents, long enough to provide context and valuable information for KE tasks.
• Structured Data: Information on aircraft details, failure codes, and dates, enabling future work on integrated KE approaches that combine structured data with natural language text.

Significance

OMIn expands the portfolio of resources available in the operation and maintenance domains by offering comprehensive records on a variety of subject matters. It is designed to be a valuable baseline dataset due to its:

• Variety and Depth: The dataset covers a wide range of incidents, each described in detail to offer context and facilitate knowledge extraction.
• Versatility: While OMIn is rooted in aviation maintenance incident data, its structure and content make it applicable to many records or logs in the operation and maintenance sectors.

Community Invitation

By releasing OMIn, we aim to support the broader community in the maintenance and manufacturing domains. We invite researchers, practitioners, and developers to collaborate on this open-source KE dataset, helping to build a robust resource that advances the state of the art in the field.

How to Access

You can download the OMIn dataset directly from this repository or access it through Zenodo, where it has been assigned a DOI for reference in academic publications.