The Lookup Wrangle can be used to store data to be looked up later. They come in two variants: key and semantic. Key lookups find data using an exact match, such as looking up IDs or names. Semantic lookups find data using a semantic search (aka vector search), such as finding similar descriptions.
Note: Semantic Lookup wrangles are a premium feature. Contact Us to get access to this feature.
¶ Key Lookup Wrangles
Key Lookup Wrangles return columns based on the keys from the cells highlighted. These keys can be words, numbers, characters or a combination of the previous. Read below to learn more about Key Lookup Wrangles!
¶ Creating Key Lookup Wrangles
To create a Key Lookup Wrangle, first go to the Lookup tab (under the My Wrangles section of the Wrangles task pane) and click on the + icon, then select the Key option. A new sheet (named Train-Lookup) will open, this is where your Key data and Value data will be stored.
Two columns will be in the Train-Lookup sheet by default: Key and Value.
¶ Key
This is the input data that the Lookup will find. It is important to note that the first column in a lookup's training data must remain being named Key.
¶ Value
This is the data that the Lookup will return when the corresponding Key is found. The name of the column can be changed, and more Value columns can be added using the "+ Add Column" button or by simply adding new columns by hand. This allows for multiple columns to be returned when the Key is found.
Note: Values are limited to 10 values per lookup.
¶ Semantic Lookup Wrangles
Semantic Lookup Wrangles are a premium feature that allows users to find data similar to key value, using a semantic search. Contact us by email or at this link to have Semantic Lookup Wrangles enabled for your account.
¶ Creating Semantic Lookup Wrangles
Creating AI Extract Wrangles is very similar to creating Key Lookup Wrangles.
Like Key Lookups, a new sheet will appear (entitled "Train-Lookup") when you create a new Semantic Lookup Wrangle.
¶ Key
This is the input data that will be used to find the most similar record from this data. It is important to note that the first column in a lookup's training data must remain being named Key.
¶ Value
Semantic Lookup values work just the same as Key Lookup values, they are the data that the Lookup will return when the corresponding Key is found. The name of the column can be changed, and more Value columns can be added using the "+ Add Column" button or by simply adding new columns by hand. This allows for multiple columns to be returned when the Key is found.
Note: Values are limited to 10 values per lookup.
¶ Updating Lookup Wrangles
To update your Lookup Wrangle, first click on the three dots next to the play button then click Edit.
From here, simply update your Wrangle as needed then click submit. Once your Wrangle has been saved, it will be ready for use.
Note: When you retrain your Wrangle, the training sheet is now named after the Wrangle.
¶ Product Code Example (Semantic Lookup)
This is an example model for looking up product titles from product codes.
Product Codes refering to the same product can many times have discrepancies from supplier to supplier. This Semantic Lookup returns a the category for the product code, even if the formatting is not completely accurate.
¶ Selecting Multiple Columns
If you have created multiple output columns in the Lookup, you will have the option to select which columns you would like to be output.
¶ Score
When running a Semantic Lookup Wrangle, you have the option to output the similarity score between the Key and the input value. A cosine similarity calculation is done to find the Score value.