Finnish model-reviewed
1,333 pairs · 4 packs
Open source
Search vocabulary, in the open.
Searchmaker Language Packs are free synonym data for web-shop search, one pack per language and shop vertical. Use them in Searchmaker or anywhere else. Credit us and you are done.
searchmaker / searchmaker-language-packsPublic
Open synonym packs for web-shop search. Finnish, English, Swedish, German, Estonian. More coming.
- 6,653 pairs
- 20 packs
- 5 languages
- CC BY 4.0
Languages
Five languages today. More to come.
Each language ships in four verticals. Dashed languages are planned; the pipeline that makes a pack is the same for all of them.
English generated
1,503 pairs · 4 packs
Swedish generated
757 pairs · 4 packs
German generated
1,886 pairs · 4 packs
Estonian generated
1,174 pairs · 4 packs
Danish
planned
Norwegian
planned
French
planned
Spanish
planned
Dutch
planned
Polish
planned
Italian
planned
What is inside
Real pairs, from the real packs.
Mostly compound split-and-join variants (very common in Finnish, German and Estonian), plus true synonyms, abbreviations and misspellings. Terms are lowercase and accent-folded so they match indexed tokens directly.
Narrower and broader product types are deliberately excluded: searching "drill" should not suddenly show every power tool.
| Lang | Searched | Also finds |
|---|---|---|
| fi | akkupora | akkuporakone |
| fi | astianpesukone | tiskikone |
| fi | hakanen | sinkila |
| en | backpack | rucksack |
| en | bookcase | bookshelf |
| en | c-clamp | g-clamp |
| sv | akgrasklippare | akklippare |
| sv | ansiktsvisir | visir |
| sv | avjamningsmassa | golvspackel |
| de | aa-batterie | mignonbatterie |
| de | akku-motorsage | akkukettensage |
| de | arbeitsplattform | arbeitspodest |
| et | aareloikur | servaloikur |
| et | aiakultivaator | mullakultivaator |
| et | akukettsaag | akumootorsaag |
How a pack is made
Generated, filtered, reviewed, published.
1. Seeds
Per language and shop vertical (tools, general, home, garden), sub-categories and everyday product words are listed.
2. Proposals
A model proposes related search terms for each seed: synonyms, compound variants, abbreviations, misspellings.
3. Strict filter
A separate pass judges every pair for shoppers’ intent. Model numbers, over-long phrases and narrower or broader product types are dropped.
4. Review
Rejected pairs go to a denylist and never return. Finnish has had a model review pass; native-speaker review is what we need most.
Status: the packs are machine-generated and not yet reviewed by native speakers. Finnish had one model review pass; the others are unreviewed. Expect a small share of wrong or too-broad pairs, mostly among the synonyms. Every pack says so in
review.native_speaker_checked. Contribute
Native speakers wanted.
- Fix a pair: delete or correct it in the pack JSON and open a pull request.
- Review a pack: if you checked a whole pack, mark it reviewed so everyone benefits.
- Add a language or vertical: open an issue first; the validator checks the format.
License
CC BY 4.0
Use, modify and redistribute the data, including commercially, as long as you give credit. Suggested line:
Synonym data from Searchmaker Language Packs (github.com/searchmaker/searchmaker-language-packs), licensed CC BY 4.0.
The data comes with no warranty. The Searchmaker plugin and engine are commercial products; the packs are the open part.