Shotnest · Library
Understand a substring match
In the Shotnest search tool, “order” matches “preorder reference” because search looks for substrings, while the classifier checks word boundaries.
Substring search versus keyword boundaries
The search tool keeps a record when its normalized text contains the normalized query anywhere, even inside a longer word. Order sits inside preorder, so record 1 matches and record 2, “delivery note”, does not. The classifier is stricter: for Latin and Cyrillic keywords it requires that no letter comes right before or after the keyword. A shelf decision and a search hit therefore follow different rules.
The app search is also substring-based for each word you type, and it checks a second form of the text in which words split across a line break are joined, so a hyphenated word at the end of a line can still be found.
Normalization applies before the substring test, so a search for cafe also finds café inside a longer word.
Try the original example
- Supplied records
- preorder reference — delivery note
- Query
- order
- Matching record numbers
- 1
Narrow a broad match
- Search order and note that record 1 matches through preorder.
- Read the whole matched word before drawing a conclusion.
- Add a distinctive part, such as an order number: “order #4471” matches only a record that contains it.
- Open the screenshot to see the context around the match.
- Query order · records preorder reference, delivery note
- matching record: 1
- Query order #4471 · with a third record “Order #4471 shipped”
- matching record: 3
Very short queries spread widely. Searching pay matches both “Repayment schedule” and “Pay now”, so the tool returns records 1 and 2.
Misreading substring hits
- Treating a match on preorder as a record of an actual order.
- Assuming a short query like “pay” finds only payments; it also finds words that contain it.
- Expecting the search hit to change the shelf the classifier chose.
- Giving up when a word is split across lines in the copied text, instead of searching a shorter part.
A match is not a classification
A substring hit tells you where some characters occur. It does not tell you what the record is about. Use the search tool linked below to find candidates and read each one.
A substring match also ignores which line the characters are on. Read the matched line itself in the screenshot rather than the whole record.