Forcepoint DSPM AI models
Small Language Model (SLM)
-
Transforms unstructured text into salient document vectors with 10-30 million parameters that effectively captures the context of the documents.
For more information, see the Large and Small Language Models.
Deep Neural Net Model
Performs sentiment analysis using document vectors. They are compact in size, fewer than 100,000 parameters which ensures rapid processing.
For more information, see the Deep Neural Networks.
'Bag-of-Words' Model
This concept is used for topic detection. These models enhance classification accuracy, which will help guide you to the correct models quickly.
For more information, see the Bag-of-Words Model
Filters & Evaluators
- It includes some regex (unintelligible) Python code to specify known data.
- This allows Forcepoint DSPM to get extremely accurate and helps to get a verdict quickly and accurately.
- They are another layer of the overall mesh.
Mapping Models
Bayesian inference combines output models into meaningful outputs which are viewed in the UI.
When a document is classified using AI and ML, the administrator can click through each object, identifying how the mesh got that verdict. This will list the detectors and classifications that were flagged in the document.
All of the vectors within the mesh are clearly identified to explain and understand how and why a classification verdict was assigned.
For more information, see the https://en.wikipedia.org/wiki/Bayesian_inference
Move to KBA