Selecting sentences does not eliminate interpretation
TextRank explores graph-based ranking for keyword and sentence extraction, and query-biased summary research studies selecting content relevant to a search. Extractive summaries reuse source wording rather than generate entirely new sentences, which can make verification easier. Yet selection itself changes context. A sentence can refer to an earlier condition, a named exception or a speaker identified elsewhere. Removing those connections can make an exact quotation misleading even though none of its words were altered.
Preserve the dependencies of selected sentences
For an ABOLTUS experiment, an extracted sentence beginning this approval applies only should retain enough surrounding context to identify which approval and which conditions are involved. A quotation from a discussion should keep the speaker or document role visible. The interface can link each selected passage to its source location and show neighboring text on demand. This allows concise previews while making clear that the extract is a selection, not a complete account of everything the document says about the topic.
Evaluate misleading selections
A useful test should include pronouns, exceptions, contrasting paragraphs and quoted statements from different participants. Reviewers should check whether the extract supports the apparent meaning when read alone. The relevant research motivates sentence selection techniques, not a guarantee that copied wording is always faithful in a new context. For ABOLTUS, the practical criterion is contextual integrity: a short extract should help the user judge relevance without creating a stronger or different claim than the full source would support.