NLP Configuration

Overview

Natural Language Processing (NLP) leverages descriptive text from product catalog feeds to enhance recommendations in scenarios with limited co-purchase data. It helps in cross-selling by recommending relevant products, improving the effectiveness of cross-sell strategies. Algonomy Omnichannel Personalization utilizes real-time sorting based on two deep learning NLP models to provide more relevant recommendations.

NLP Similarity: A deep learning model that recommends similar products based on text descriptions and viewed together behavioral data.

NLP Cross Sell: A deep learning model that recommends cross-sell products based on text descriptions and bought together behavioral data.

The system displays NLP affinity sort options within Algonomy recommendation strategies only when NLP is activated for the site.

You can access the NLP configuration page at:

https://portal.richrelevance.com/rrportal/nlpConfiguration

Property Name

Description

Enable NLP Strategies

Activates NLP strategies for a specific site.

Use Separate Descriptions File

Allows the upload of a separate file for descriptions. Specify the location for the separate descriptions file upload.

The following input fields are to be filled:

·        File Directory: Enter the directory path in your FTP account. The default value is '/'.

·        File Name Format: Specify the file format. For example, 'NLPDescriptions*.csv'.

·        File Delimiter: Choose the file's delimiter. Options include pipe (|), semi-colon (;), comma (,), and Ctrl+A.

Product catalog attribute contains description

Allows you to configure a description from an attribute that already exists in the catalog. You need to enter the name of an existing attribute in the product catalog to be used as the description for the product.

Note: By default, the standard attribute will be used for the description, when ‘Enable NLP Strategies’ is enabled, unless another option is specified.

NLP Config API

The NLP Config API allows users to configure description of products from attributes and prepare the data used to train and evaluate NLP models. This API returns a list of product attributes to be used for NLP strategies.

Get NLP config for a site

This API request returns NLP configuration details for a specific site.

HTTP Method: GET

URL: https://portal.richrelevance.com/rrportal/nlpConfig

Payload:

Copy
{
    "available_attributes": {
        248875: "gender",
        248876: "dept"}
    "enabled": false,
    "separate_descriptions_file": false,
    "product_attributes": [],
    "cooc_join_type": "pseudoSession",
    "available_cooc_join_types": ["userGuid", "pseudoSession", "userGuidChronological", "externalSessionIdChronological"]
}

Field

Description

available_attributes

List of active product attributes for the current site (attribute_id → external_id).

Enabled

Enables or disables input catalog files generation for NLP training.

separate_descriptions_file

Defines whether to use a separate feed file for product descriptions or extract them from the standard catalog feed file.

product_attributes

List of selected attribute IDs to include in the NLP products input file.

Note: Values are always numerical, derived from the attribute_id column.

cooc_join_type

Specifies the NLP co-occurrence join type.

available_cooc_join_types

Enumerates available co-occurrence join types.

Update NLP config for a site

This API request updates NLP configuration details for a specific site.

HTTP Method: PUT

URL: https://portal.richrelevance.com/rrportal/nlpConfig

Payload:

Copy
{
    "enabled": false,
    "separate_descriptions_file": false,
    "product_attributes": [248875],
    "cooc_join_type": "pseudoSession"
}

Field

Description

Enabled

Enables or disables the generation of input catalog files for NLP training.

separate_descriptions_file

Defines whether to use a separate feed file for product descriptions or extract them from the standard catalog feed file.

product_attributes

List of selected attribute IDs to include in the NLP products input file.

Note: Values are always numerical, derived from the attribute_id column.

cooc_join_type

Specifies the NLP co-occurrence join type.