DeepSearch Model Training
Our Deep Search model uses advanced deep learning techniques to deliver faster and more accurate search results. It’s designed to understand both user intent and product relevance, making search experiences more personalized and precise.
When training your Deep Search model, you can guide its focus in two powerful ways:
Catalog targeting: Choose which parts of your product catalog the model should prioritize.
Field selection: Specify which product attributes (like title, brand, color, etc.) should influence how results are ranked.
This makes your search engine smarter, faster, and more aligned with what your customers are actually looking for.
Training Status Section
The status section gives you critical information for use in managing and debugging anything where the output is not what is expected:
Status: This tells you the status of the model (Training, Training Failure, Trained)
Last Trained: This is the date of the last model training for DeepSearch
Last Catalog Sync: This is the last time the product information was updated from the brand
Last Product Field Validation: This is the last time the product fields were validated.
Locale Training
DeepSearch supports a unique, dedicated model for each catalog or locale, ensuring that search results are tailored to the specific language, region, or product mix of that locale.
When you select a locale, DeepSearch automatically trains and links a model specifically for it, no manual setup required. This ensures your search experience stays relevant, localized, and optimized for each market.
Product Fields
DeepSearch, being a text-based engine, relies on product data as a training set. Here you can determine which fields you want DeepSearch to be trained on.
Disclaimer
Only include fields that are useful - extraneous data added into the training can reduce relevancy in results.
