Moderation

Clients offering personalization services often have needs to prevent the abuse and violation of various agreements. Typical uses case include, for example:

• Validating that product text personalization does not contain swear words

• Validating that product text personalization does not contain prohibited brands or related copy

• Validating that image personalization does not contain nudity or gore

• Validating that image personalization does not contain embedded prohibited text

• Validating that image personalization does not contain brand names, licensed characters, etc.

Moderating text and images during the personalization time allows to validate concerns early in the process, before an order is placed and sent to OMS, thus reducing the customer service costs and increasing the customers' satisfaction.

To assist with such use cases, Customizer platform offer moderation services, some of which are available with core platform subscription fees and more advanced services that require a separate subscription.

Included text moderation services

Customizer platform include a text moderation engine out of the box, which is available for customers to use without any additional fees or provides. The built-in moderation service is a high-performance, fault-tolerant solution that is tuned for speed and aims to cover basic text moderation needs.

The out-of-the-box moderation service is based on explicit allow/block list architecture.

• Customers can optionally start with the provided community-maintained, multilingual (20+ languages) block list

• Customers can upload three more additional block and allow lists

  • For example, it is possible to create a list of prohibited brands and an additional list of prohibited words.

The service does not impose strict usage limits. The system usage is monitored to prevent system abuse and perform troubleshooting but is not explicitly limited by the total number of calls or rate limited.

This allows the implementation of more flexible use cases, for example, validating the text while a customer is still typing the text.

The standard text moderation can be a good fit for simpler use cases:

• Short or limited text personalization

  • For example, 3 letter monogramming has limited room for abuse

• Explicit list of prohibited brands or licenses

• Fast moderation speed requirements while explicitly allowing some other personalization types

  • For example, a business card or greeting card customizer would explicitly include person’s emails or names.

Advanced text moderation

Some customers may require more advanced text moderation needs:

• PII detection, such as SSNs, emails, names, credit card numbers, etc

• Profanity detection

• Toxic language

• Sentiment analysis

The AI services come preloaded with common

Customizer and Cartographer include a more advanced text moderation engine based on a set AI pre-trained models. Using the advanced engine over a separate implementation has several advantages:

• Simple API integrated with Customizer services

• Prebuilt integrations

• Managed through the Customizer Admin interface

The AI-based services impose usage limits and rate limits. It may be required to review and revise some of the UX/UI flows to accommodate:

• Slower response types

• Higher usage fees

For example, instead of type-ahead validation, when using these types of services it may be preferred to perform the validation on the

Image moderation services

Customers who allow personalization by allowing end customers to upload their own images may be interested in the image moderation services. Typical use cases include:

• Nudity and graphic image detection

• Brand and licensed characters recognition

• Text OCR and image validation

Privacy concerns arise when reviewing and analyzing User Generated Content images. By using a trusted infrastructure provider, bundling the services, and tightly controlling the life of users uploaded content, Drive is able to provide a privacy-centric, GDPR and CCPA-compliant, image moderation service.

Drive uses AI hosting and interference services provided by third party platforms, such as Microsoft Azure or OpenAI. The respective providers are named as a data subprocessors on the Drive DPA agreement, and data concerns are governed by the Drive respective MSAs.

Image moderation services are tuned out of the box for the above named use cases. Image moderation services, for example, could be combined with the built-in or advanced text moderation services:

• Capture the text embedded in images

• Moderate the captured text for profanity or brand name inclusion.

Moderation services comparison

Text moderation

Image moderation