Synonyms and descriptions

1. Synonyms and descriptions

Users speak business language; your database speaks column names. Let's bridge that gap.

2. The language gap problem

Business users speak in broad outcomes; databases speak in granular IDs. The "Customer" trap: a user asks for "top clients," but your table is sales_customers and the column is customerID. Without synonyms, Genie looks for "client" and fails. "Money" ambiguity: users say "revenue" or "sales," but your data is labeled totalPrice or gross_amt. "Location" confusion: users say "stores," but the technical table is sales_franchises. Genie doesn't automatically know these terms are synonyms. That's the concept-versus-column gap.

3. Before and after curation

Consider the before-and-after scenario shown here. A finance manager asks "What's our EBITDA?" Without synonyms, Genie finds nothing. It can't map the business term "EBITDA" to the actual columns that hold earnings data. The fix is straightforward. In the space configuration, you add synonyms that map "EBITDA" and "profit" to the correct revenue and cost columns. After adding these mappings, the same question returns accurate results. Same business language, but now Genie understands what the user is really asking for.

4. Column descriptions

Beyond synonyms, you can add column descriptions that help Genie understand context. Unity Catalog descriptions flow into Genie automatically. If you documented tables in Unity Catalog, Genie starts with a head start. You can override or add Genie-specific synonyms inside the space config without affecting Unity Catalog metadata. Instead of just "paymentMethod," add a description like "Identifies the specific card issuer network used to process the payment." This helps Genie make smarter decisions.

5. Hiding irrelevant columns

You can also hide columns that shouldn't be exposed to users. Hide technical metadata like new_id, review_uri, and chunk_id from media_customer_reviews and media_gold_reviews_chunked. These are useless for business questions and confuse the AI. Hide redundant IDs: if you have franchiseID and name, expose the name and keep the ID hidden; Genie still uses it for joins. Expose human-readable columns like product, quantity, and review_date. Fewer columns means less confusion and better accuracy.

6. Let's practice!

Compare curated and uncurated spaces to see the power of synonyms firsthand. Let's practice!

Create Your Free Account

By continuing, you accept our Terms of Use, our Privacy Policy and that your data is stored in the USA.