Meridian licenses several standard charterparty templates (e.g., a common industry form) and each contract begins with an identical, several-hundred-word 'definitions and interpretation' preamble common to that template. Across hundreds of contracts using the same template, what data-preparation step improves retrieval quality?

Detecting and deduplicating or down-weighting the identical shared preamble is correct because this is the same underlying problem seen with repeated customs disclaimers and reissued circulars in this domain: near-identical boilerplate text repeated across a large number of documents crowds the index with redundant vectors and can surface as an irrelevant top result, diluting the value of retrieval for the clauses that actually differ contract to contract. Indexing the identical preamble in full inside every contract's chunk set unchanged is wrong because it directly reproduces that crowding problem at scale - across hundreds of contracts, the same near-identical text would occupy a large share of the index while contributing little discriminating information for any specific query. Deleting every contract that uses a standard template is a wildly disproportionate response - it discards the genuinely unique, substantive clauses those contracts contain (the actual charter terms, rates, and conditions) purely because they share a common preamble, throwing out valuable content to solve a problem that only affects one shared section. Translating the preamble into a different language does nothing to address duplication - it produces another version of the same repeated boilerplate content in a different language, which does not reduce redundancy and adds unnecessary translation cost and complexity that the scenario never called for. The exam-relevant pattern is recognizing shared-template boilerplate as the same deduplication problem as other repeated-text cases in this domain, regardless of document type.