Skip to Content
BayesiaLabBayesiaLab User GuideMenusHellixiaEmbedding Generator

Embedding Generator

Overview

The Embedding Generator is a Hellixia feature that automatically creates embedding vectors for selected nodes. The generator builds a continuous, high-dimensional vector for each node by reading its Node Name, Long Name, and/or Comments (these are the Embedding Source Fields). The choice of which fields to use, and the richness of the Comment in particular, directly shapes the resulting embedding.

The output vectors can have various dimensionalities (e.g., 3,072, 1,536, or 1,024 dimensions, depending on the model chosen). Furthermore, Hellixia supports Embedding Concatenation: embeddings from several different models can be joined together (e.g., 3,072 + 1,536 + 1,024 dimensions) to enrich the representation of each node. This produces richer node representations and, in turn, more informative semantic correlations in the learned network.

The purpose of generating embeddings is to make the nodes comparable in a continuous space where semantic similarity corresponds to geometric proximity. This comparability is what enables the subsequent unsupervised structural learning step to find semantic correlations and build a Semantic Network.

Usage

You start the Embedding Generator by selecting Menus > Hellixia > Embedding Generator.

Step 2 of the Semantic Network Generator workflow includes the Embedding Generator function.

Workflow

Prepare the network, source material, selected nodes, selected arcs, or Knowledge Files required by the function.
Confirm that Hellixia provider settings and model access are configured.
Open Hellixia > Embedding Generator.
Review the available options and add General Context when the model needs domain-specific guidance.
Run the function and inspect the generated graph, comments, priors, embeddings, translations, or images.
Dimension ElicitorComments