Semantic Platform Newsletter - May 2024

Clubic uses Semantic Platform!

Thanks to the Semantic Platform solution, Clubic improved its internal linking and made several thousand articles visible again. Click on the button to see the full case study.

New features

⚙️ API - Classification

A new feature lets you asynchronously classify a batch of documents.

⚙️ API - Named entity recognition

We have completely overhauled the automatic recognition engine for named entities (discovery mode), which now boasts even greater precision. In addition, a fourth type of entity is now available: communities. It can be used to highlight groups of people sharing common characteristics (geographical, linguistic, etc.).

A number of new features have also been added to our engine:

• ability to return only the 1st occurrence of a named entity that appears several times in the document
• parameterization of previous option by entity type (place, person, organization, community)
• customize links created on recognized entities and add metadata to these links (in dictionary mode)
• ability to import dictionaries and variants by batch
• when an entity has been recognized by dictionary, the name of the dictionary is specified in the return message
• ability to blacklist named entities

💻 Customer portal - New

Most of our customers want semantic enhancements to be submitted to the editor for validation. In some cases, this is not desired or not possible.

The Semantic Platform portal allows you to view the enrichments you have made, and tointervene when our AI requires human validation (confidence rate too low) or when an enrichment requires correction. For the moment, the portal only manages automatic classification.

🧩 WordPress Connector Extension - Improvements

The following improvements have been made to theWordPress extension:

• ability to prohibit the creation of new named entities suggested by the AI to certain users
• setting the WordPress taxonomies to be used

Features under development

⚙️ API - Suggested folder to create

We're working on a new feature which, by analyzing the editorial base, will suggest folders to be created.

⚙️ API - Named entity recognition

Currently, the named entity recognition call returns all entity types, and the threshold at which an entity is detected by the AI is common to all. It will soon be possible to choose which entity types to return, and to set each threshold freely.

⚙️ API - Classification

We will soon be offering a hierarchical classification engine, capable of associating content with an intermediate level of the classification plan (currently, association is only made with the last levels).

💻 Customer portal - Improvements

We'll soon be adding the ability to control other semantic enrichments (named entity detection and read-only links). The portal will also soon make it possible to import a batch of documents for semantic enrichment, without having to script an API call.

🧩 WordPress Connector Extension - Improvements

TheWordPress extension will soon be able to automatically initialize the solution's dictionaries from existing WordPress taxonomies.

R&D features

⚙️ API - LLM features

We are working on adding new functionalities enabled by the integration of a Large Language Model, and our current research focuses on :

• acceleration and parameterization support for current functions
• text generation functionalities, in compliance with the Charter of Paris, as we have committed to do