A little room to wonder
Search & Indexes
Build a small word-to-record lookup and check how its rules differ from the meaning or trustworthiness of the results.
How can a list of words help locate a record without rereading everything?

An index is a prepared route to records
An index organises information so that a later lookup can follow a prepared structure. In a simple word index, each word points to the IDs of records containing it. This reverses a document-to-words description into a word-to-documents lookup.
An index must be built from something. Choosing titles alone, full text or manually supplied tags changes which matches are possible. Faster lookup does not automatically mean broader coverage; the prepared representation determines what can be found.
Make word rules explicit
A paper index can lower-case letters and separate words at spaces, provided those rules are stated. More complex systems may handle punctuation, language and word variants differently. “Contains these exact words” is a different rule from guessing a related topic or predicting what someone intended.
The teaching index here is fictional. It is not a description of Horizon’s internal search code or a request to upload queries or build a database. The same conceptual question can be explored with a few records and a pencil.
Worked example: intersect two lists
Use three invented titles: A, Blue Paper Boats; B, Paper Robot; C, Blue Leaf Map. The word blue points to A and C; paper points to A and B. Requiring both words leaves their intersection, A. Allowing either word produces the union, A, B, C.
There are two matches for each individual word, one for the both-words rule and three for the either-word rule. The different totals reflect a stated query rule. They do not show that A is more accurate, more useful or more trustworthy than the other records.
An index needs maintenance and judgement
If a title changes but its index entry does not, a lookup can return an outdated result or miss a new one. Recording stable IDs helps distinguish a record’s identity from its current display name. Removing a record also requires maintaining any pointers to it.
Retrieval and evaluation remain different tasks. A matching word is evidence of a text relationship, not proof that a source is correct. Open the record, read its context and check the claim rather than treating a search position as an endorsement.
Only if you feel like trying
Build a three-record paper index
- List the words and record IDs from the three supplied titles.
- Answer blue AND paper and blue OR paper using intersection and union.
- Rename one title and identify every index entry that would need review. No real search history or personal records are needed.
Optional. Use paper or talk through the supplied fictional example. No purchase, personal information, account, public sharing or practical experiment is needed.
Look at the sources.
These references offer further reading on the ideas here. Check who published them, what they support and what you would ask next.
- Computer Science Education Research Group / University of CanterburySearching algorithms
- PostgreSQL Global Development GroupPreferred Index Types for Text Search
Research checked 1 November 2026. External pages have their own content and privacy practices.
Where could this connect?
These are editorial links between subjects. You choose which, if any, to follow.
Computing
Search indexes show how computing prepares data for a particular kind of lookup.
Follow this connectionData Tables & Sorting
Stable record IDs help keep lookup pointers attached to the right records when display order changes.
Follow this connectionKnowledge & Evidence
Finding matching words is the start of checking an idea, not proof that the matched claim is true.
Follow this connection