Universal Classification
Brief
A customer working on a sales platform has a variety of text-based inputs to their systems. They aggregate products from multiple websites, however the product metadata is not standardised. They are able to use the TrueState platform to achieve a unified ontology, allowing them to process the metadata more effectively.
The customer has a set of features that they can display about the product, such as 'power windows, push button start, LED headlights'. However, from their data source they are receiving a set of product features with arbitrary names, such as 'Electric windows with one-touch operation'. Without any NLP techniques, the customer would have to manually map these features to their own ontology.
The customer was able to use the universal classification model to automatically map these features to their ontology. The model takes the customer's ontology as input, and does an entity match against each of the provided features. The model will return a score based on its confidence that the records are a match.
Tutorial
-
Go to the Workflows - My Workflows page and create a new workflow.
-
Set the project name and description.
-
Select the ‘+’ icon in the middle of the screen and create a new Dataset node.
-
Under ‘Selection Mode’ select Upload a new Dataset and name and upload the dataset.
-
Once the dataset has uploaded, click the ‘+’ again and click ‘Use a model’.
-
Set this model action type to be ‘Entity Match’. Add the named entity for the input features and set the column name to the text field that we want to search.
-
Click the ‘+’ arrow to create another dataset node. Leave the selection mode to ‘Automatically create dataset on run’ and call it ‘Output Example Identified Features’.
-
Click the ‘+’ arrow to create another action. Click on ‘Transform Data’ and write the necessary queries to query the data.
-
Click the ‘+’ arrow to create another dataset node. Leave the selection mode to ‘Automatically create dataset on run’ and call it ‘processed--features’.
-
Drag the node edges to join them so that the example feature inputs data feeds into the entity match model, which then feeds into the output identified features dataset. The output identified features dataset will then feed data into the data transform model which will generate the processed features dataset.
-
Click ‘save’ in the top menu, then return to the Workflows page with the ‘exit button’.
-
Click ‘Actions’ then ‘Run’ next to the workflow name. You can click ‘Actions - Log’ to check the status of the run. It would also show a pop-up that says ‘Solutions Run Started.’ after clicking ‘Run’.
-
After the run is successful, you can click ‘Actions’ then ‘Edit’ next to the workflow name to view the workflow. You can then double click on the output results dataset to see the dataset preview. (You can click on the ‘expand’ button on the upper right corner to expand the preview)
