Fig. 5

Re-ranking chunks—‘NuA4’ and ‘meiosis’ are keywords extracted from a query: Find all results that connect NuA4 with meiosis. Each row represents a paragraph or chunk from research papers, with columns displaying both individual and total frequencies of the keywords. The upper table shows chunks that are initially ranked according to cosine similarity computed between the query and each chunk in the vector DB. The bottom table depicts the re-ranking of the same chunks based on their keyword frequencies. The ranking starts at 0, indicating that the topmost relevant chunk is ranked at the 0th position