Semantish

The Semantish layer is an optional feature layer within the engine that can be enabled by your project manager.

Semantish uses pattern recognition techniques, plus domains specific learning, and applies that to user searches to transform them into machine friendly queries. Domain specific words are identified through user behavior which allows the engine to isolate search intent from metadata and creates highly converting, personalized relevance. Natural language behaviors are identified using inference analysis.

We breakdown the search to parse out the search intent and get rid of the ambient noise, to inform the search engine and bring back the correct results in the best possible order.

The additional functionality and possible transformations are outlined below.

You need to consult your project manager to identify which gestures make the most sense to enable for your requirements.

Using Semantish

By default the Semantish layer is available to you, but is disabled as it requires a technical review of your data before it can be turned on. You can use the Caching portion of Semantish without waiting for the technical review.

Header structure

To start using Semantish you need to make two changes:

1. Use the -cors version of your endpoint
2. Specify that you don’t want to skip Semantish

Sample query:

curl  -H "Skip-Semantish: false" -d '{ "query": "your query", "clientKey": "{$$.env.apikey}" }' "https://{$$.env.customerID}-cors.groupbycloud.com/api/v1/search"

Caching

The Semantish layer has caching that can be configured to be 5 to 30 minutes in duration. Check with your project manager on what is the configuration for your site.

You can skip the caching layer by setting the header: -H "Skip-Caching: true"

curl -H "Skip-Caching: true" -H "Skip-Semantish: false" -d '{ "query": "your query", "clientKey": "{$$.env.apikey}" }' "https://{$$.env.customerID}-cors.groupbycloud.com/api/v1/search"

This is useful for configuring Staging and Preview environments.

If you have made changes to your data in Production and need to update the cached response for the query, we recommend that you make it possible for your search requests to skip cache. This forces the engine to perform a cache refresh and display the most up-to-date content on the next time that request is made.

Gestures

Any Semantish gestures that would have been applied will be noted in the siteParams section of the response. You can evaluate this portion of the response via the Reference Application to understand what transformations were applied.

Available gestures

Field specific and customer specific intelligence is applied to build out the patterns that are used by the Semantish layer. All of these are intended to distill the query to the search intent and associated metadata, such that recall and relevancy are optimal.

Brands

Semantish can pick out Brand specific searches, and convert them to a refinement so that your Brand experiences are merchandised once.

Price point

The engine can recognize the intent behind terms like “popular”, “cheapest”, “off-brand” and transform those into “sort” and “refine” actions.

Units

We can teach the system to recognize units that are specific to your domain, such that the engine returns the size of products that the customer is looking for.

Intelligent biasing

Depending on your customer journey and specific refinements, the Semantish layer can transform meaningful metadata within the search into intelligent biasing of the results to increase your relevancy. For example, if your customers search by using temperature modifiers (such as “warm” or “cold”), the engine can transform this into an appropriate bias for items with different levels of insulation.

Category matching

If your customers search by categories and demonstrate intent to hit a category, rather than a keyword search, Semantish can support an automatic transition of keyword searches into the correct categories.

Part number search

Semantish can automatically recognize part number searches, and disable auto-correct so that your recall is more precise.

Measurements

If you are interested in recognizing measurement driven searches, you can leverage the Semantish layer to transform the variety of ways your customers search into meaningful gestures for your individual data set.

Others

We continuously release additional gestures. Contact your Project Manager to learn how these can be useful to you.