Hyperedge Nodes & Event Projection

Hyperedge Nodes & Event Projection

Problem Statement

In narrative texts like the Mahabharata, speaker attribution is often repetitive and needs to be rendered differently based on the target audience or language.

Source:

`sanjaya `uvacha

Target (English):

Sanjaya said:

Target (Sanskrit):

सञ्जय उवाच

Currently, we model this as two separate commands: a speaker entity (sanjaya) and an action (uvacha). However, this relies on linear processing and doesn’t capture the event effectively.

Proposed Solution: EventHeader

The Vyasa syntax supports Event Headers, which act as hyperedge nodes connecting multiple entities in an event tuple.

Syntax:

( sanjaya uvacha )

This parses into a Node::EventHeader containing a tuple of commands.

Semantic Meaning

This node represents a Hyperedge connecting:

• Subject: sanjaya (Entity)
• Verb/Action: uvacha (Action)
• Context: The simpler linear stream implies this event applies to the following block.

Projection Strategy

To support multi-lingual output, the Projector can be updated to handle EventHeader nodes by matching them against a template, similar to standard commands.

1. Template Definition

We can define a template for the event concept, or speculatively, for the specific tuple signature.

A more flexible approach is to map the Action to a template that accepts the Subject.

Template (uvacha):

`template { target="html" name="uvacha" } [
    `div { class="speaker-attribution" } [
        `span { class="en" } [ $.subject.label_en said: ]
        `span { class="sa" } [ $.subject.label_dev ]
    ]
]

2. Compiler Logic

When the compiler encounters ( sanjaya uvacha ):

1. Resolve Entities: It resolves sanjaya to the entity defined in the registry.
2. Resolve Action: It identifies uvacha as the action.
3. Apply Template: It looks for a template matching the action (uvacha).
4. Inject Context: It injects the subject’s properties ($.subject.*) into the template.

This allows the source text to remain minimal and semantic (( sanjaya uvacha ) or just sanjaya uvacha if we infer the tuple), while the output can be rich and localized.