This double process cycle is iteratively applied until an optimal balance of differences and commonalities between stakeholders are reached that meets the semantic integration requirements.
This approach is based on research on community-based ontology engineering[1][2] that is validated in European projects, government and industry.
Semantic reconciliation is a process cycle constituted of four subsequent activities: scope, create, refine, and articulate.
Next, relevant facts are collected from documentation such as, e.g., natural language descriptions, (legacy) logical schemas, or other metadata and consequently decomposing this scope in elicitation contexts.
This corresponds to Robert Meersman's statement that semantics are "a (set of) mapping(s) from your representation language to agreed concepts (objects, relationships, behavior) in the real-world".
In the mind of reader or writer), hidden away in the implementation itself (e.g., in a database table or column code) or informally captured in textual documentation.
In that respect, it is similar to Enterprise Information Integration (EII) with the added benefit that the shared models are not described in technical terms but in a way that is easily understood by the business.