Free tools / Passage relevance inspector
Passage relevance inspector: which passage answers the question?
Our passage relevance inspector splits your text into passages and ranks each against one question with Okapi BM25, a published lexical method (k1 = 1.2, b = 0.75). You see each passage's score, the question words it matches, its coverage and whether its opening sentence carries those words.
The inspector runs in this page. Nothing you enter is sent to our server. Paste up to 100,000 characters.
No email is needed, and nothing you paste is stored. A shared result stays in its link, not on our server.
What we check and why
Passages
Paragraphs separated by blank lines, with a heading line joined to the paragraph after it. A paragraph over 300 words, or every passage in window mode, is cut into 120-word windows.
Words that count
Words are lower-cased and accents removed; 82 common words are dropped. Suffixes: -ies becomes -y, -sses becomes -ss, -es after s, x, z, ch or sh is removed, a final -s is removed (not -ss), -ing is removed from words over five letters and -ed from words over four.
BM25
For each question word in a passage: idf x (f x (k1 + 1)) / (f + k1 x (1 - b + b x length / average length)), with k1 = 1.2 and b = 0.75, where f is how often the word appears in the passage. idf = ln(1 + (N - n + 0.5) / (n + 0.5)) for N passages, n of which hold the word.
Coverage and the opening sentence
Coverage is the share of distinct question words a passage holds. We also say whether its opening sentence holds any of them, since an answer that opens a passage survives being quoted alone.
What lexical scoring misses
It matches words, not meaning: a synonym scores nothing. Scores compare passages within one text and cannot be compared across texts.
Questions
What is passage relevance?
How well one section of a page answers a specific question when read on its own. Many search and answer systems retrieve passages rather than whole pages, so a page can cover a topic while no single passage answers the question.
Is BM25 how ChatGPT or Google pick passages?
We do not know how any engine retrieves passages, and none publishes it in full. BM25 is a published, repeatable lexical baseline: it shows whether a passage uses the words of the question, which is one thing you control.
Why does a passage that answers the question score low?
It answers in different words. Lexical scoring counts shared words after removing common words and suffixes, so a synonym scores nothing. Use the question's own terms where they read naturally.
Want the fix made for you?
Answer Repair completes one agreed source-backed page or fix. Choose Live or Handoff before payment. Due 48 elapsed hours after Ready. Full refund on request within 30 days of delivery.
When you want us to do the work, every offer and its published price is on our offers page. More free checks: the ChatGPT Business Hub readiness check, the content scoring tool and the raw HTML readability checker, or see all free tools. Read next: How do you write a page an AI engine can quote?