Training Data for Product and Issue Understanding

It is the process of labeling product references, reported problems, and their relationships in source records. The resulting examples support downstream models for product entity recognition, issue classification, and relationship understanding.

Use CSV records containing product feedback, service notes, support tickets, or other relevant text fields. In row mode, each record becomes a separate annotation task with its original fields preserved.

Yes. Entity labels can be created inside any text column. Each span retains its source column and exact character offsets, including when product and issue mentions occur in different fields.

Create the product and issue entities, then connect them with a relation defined in your ontology. For example, a has_issue relation can point from the product mention to the reported issue in the same record.

The tabular relation overlay connects labeled content across fields within one record. This workflow does not create joins or entity relationships between arbitrary CSV rows.

Yes. Shared record properties can capture context such as severity, issue category, and resolution state using the property types and choices defined for your project.

Define shared entity classes, property choices, and relation types in an ontology. Instructions, required-property checks, and review help annotators use those definitions consistently.

Reviewers inspect labels and relationships in the original record. They can correct the work, add contextual comments, or reject it through a configured rework path before approval.

Yes. JSONL exports retain record values, column schema, entity spans, relations, and supported record properties, so the annotated evidence remains connected for downstream model development.