Annotation Guidelines for Data Extraction
When you are reviewing documents, it is crucial to annotate them carefully, as Orby learns not only from the text of the entities but also from their positions in the document. Here are two things to keep in mind:
Entity Text: It is crucial to annotate the text for an entity exactly as it appears in the document. This ensures accurate data extraction, which is vital for optimal performance. Altering the text or using different wording could confuse Orby. However, it's worth noting that future versions may include features that allow for more flexibility in text annotation.
Entity Location : Use a bounding box to highlight the entity's location on the page. If the bounding box is incorrect, Orby could learn the wrong information, compromising the quality of its future tasks.
Remember, poor annotation can send Orby down the wrong learning path, leading to less-than-optimal performance.