A matching model compares an incoming message against your own master data records and returns the single best match. Use it when the job is to connect free text to a specific record in your data - for example matching a customer's free-text product description to the correct product code in your catalog, or a company name to the right account in your customer list.
This article explains what a matching model does and the concepts you will work with. To build one, see Set up your first matching model.
What it does
You import a set of records (your master data) and the model matches each incoming message to the one record that fits best. Rather than choosing from a short list of fixed categories, a matching model can search across a large reference set and pick the right entry, returning the matched record together with a short explanation of why it was chosen.
This makes it different from a classification model: classification chooses from labels you define by hand, while matching chooses from records you import and can update at any time.
Master data records
The records are the reference data the model matches against - for example a product catalog, a list of accounts, or a set of part numbers. You import them as a file, and you designate one column as the unique identifier that identifies each record.
To learn how to import and maintain them, see Managing master data records.
Domain knowledge
Real-world data has quirks: abbreviations, synonyms, and special cases that are not obvious from the records alone. Domain knowledge is a written context document where you capture these, so the model knows how your team actually refers to things.
To learn how to use it, see Using domain knowledge to guide matching.
The result
For each message the model returns the single best-matching record plus a short piece of reasoning. Your team can confirm or correct the match, and those confirmations become validated examples that improve the model.
Related articles
- For an overview of all model types, see What are models on Tekst?.
- To build a model, see Set up your first matching model.
- To understand scoring, see How matching accuracy is measured.
0 comments
Please sign in to leave a comment.