If the document text has poor-quality OCR and needs to be updated, what is the best method to reflect this new text in the index?

Running a full build of the index is the most effective method to reflect updated text, especially when the document has poor-quality OCR (Optical Character Recognition). When OCR quality is inadequate, it can lead to misinterpretations of the text, rendering the content less useful or accurate in the database.

A full index build processes all the documents in their entirety and regenerates the index based on the latest, corrected text. This ensures that any updates to the document—resulting from fixing OCR errors—are captured across the entire index. It allows for comprehensive re-evaluation of the content, which is essential for maintaining accuracy in search results and ensuring that users can find the correct information.

Incremental builds typically only update portions of the index based on changes, which wouldn’t adequately address the issues stemming from previously inaccurate OCR text. Adding new repeated content filters or clustering documents focus more on organizing or enhancing search efficiency or grouping documents rather than correcting underlying text quality issues. Hence, in the context of needing to address poor-quality text specifically, a full rebuild is the most appropriate approach.