Viewing Extractions¶
Completed cases display extracted entities organized by pipeline step. The extraction results show the nine component types identified from case text.
Accessing Extraction Results¶
Extraction results are organized by pipeline step:
- Step 1 -- Contextual Framework (Roles, States, Resources)
- Step 2 -- Normative Requirements (Principles, Obligations, Constraints, Capabilities)
- Step 3 -- Temporal Dynamics (Actions, Events, causal chains, temporal relations)
- Step 4 -- Whole-Case Synthesis (Code Provisions, Precedent References, Ethical Questions, Conclusions, Decision Points, Resolution Patterns, Causal-Normative Links, Question Emergence)
Authenticated users access results via the pipeline status bar on the case detail page. Completed steps show a green filled button. The case detail page itself displays extracted entities as interactive RDF triple labels for all users.
Entity Review Display¶
Extraction results showing entities from the Facts section
Interface Layout¶
| Section | Description |
|---|---|
| Available Classes | Existing ontology classes (collapsed by default) |
| Extracted Entities | Entities identified from case text |
| Section Toggle | Switch between Facts and Discussion results |
Entity Table¶
The entity table displays extracted concepts with their labels, definitions, types, and status.
| Column | Description |
|---|---|
| Label | Short entity identifier |
| Type | Concept type (Role, State, Principle, etc.) |
| Definition | Full description from extraction |
| Status | New, Existing, or Modified |
| View Extraction | See original LLM prompt and response |
Section Toggle¶
Steps 1-2 extract from Facts and Discussion sections separately. Toggle between sections to view entities from each pass.
Step 3 extracts from the full case text in a single unified pass and does not have a section toggle.
Extraction Steps¶
Step 1: Contextual Framework¶
Identifies situational elements:
| Type | Symbol | Description |
|---|---|---|
| Roles | R | Professional positions with duties and authority |
| States | S | Situational context and conditions |
| Resources | Rs | Professional knowledge including codes and precedents |
Step 2: Normative Requirements¶
Identifies ethical guidance elements:
| Type | Symbol | Description |
|---|---|---|
| Principles | P | High-level ethical guidelines |
| Obligations | O | Specific requirements for action or restraint |
| Constraints | Cs | Inviolable boundaries on conduct |
| Capabilities | Ca | Competencies for professional practice |
Normative requirements extracted from case discussion
Step 3: Temporal Dynamics¶
Identifies action and event elements via unified LangGraph extraction:
| Type | Symbol | Description |
|---|---|---|
| Actions | A | Volitional professional interventions |
| Events | E | Occurrences outside agent control |
| Causal Chains | - | NESS test causal analysis |
| Allen Relations | - | OWL-Time temporal ordering |
| Timeline | - | Chronological sequence of actions and events |
View Extraction Details¶
Click View Extraction on any entity to see the original LLM interaction:
- Prompt - The template-rendered prompt sent to the model
- Response - Raw JSON response from the LLM
- Timestamp - When extraction occurred
- Model - Which LLM model was used
This provides transparency into the extraction process.
Available Classes¶
The "Available Classes" section shows existing ontology classes from OntServe. Expand this section to see what classes were available during extraction:
| Category | Examples |
|---|---|
| Roles | Engineer, Client, Employer |
| States | Competent, Conflicted, Authorized |
| Resources | NSPE Code, State Regulations |
| Principles | Public Safety, Competence |
| Obligations | Disclose, Verify, Report |
| Constraints | Cannot Certify, Must Not |
| Capabilities | Can Consult, May Decline |
| Events | Request, Discovery, Violation |
| Actions | Certify, Disclose, Decline |
Status Indicators¶
| Status | Icon | Meaning |
|---|---|---|
| New | Star | No matching ontology class found |
| Existing | Check | Matched to existing ontology class |
| Modified | Pencil | User-edited after extraction |
Step 4: Whole-Case Synthesis¶
Step 4 analyzes extracted entities and case text to produce 8 additional entity types:
| Entity Type | Phase | Description |
|---|---|---|
| Code Provision Reference | 2A | NSPE code sections cited in the case |
| Precedent Case Reference | 2B | BER cases referenced in discussion |
| Ethical Question | 2C | Questions posed to the Board |
| Ethical Conclusion | 2C | Board's formal determinations |
| Canonical Decision Point | Phase 3 | Points where ethical choices must be made |
| Resolution Pattern | 2E | How ethical tensions are resolved |
| Causal-Normative Link | 2E | Connections between causal factors and norms |
| Question Emergence | 2E | How ethical questions arise from case facts |
Step 4 has three sub-views accessible from the pipeline sidebar:
| View | Description |
|---|---|
| Extraction | Phase overview and re-run controls |
| Review | Entity review and OntServe commit |
| Full View | Tabbed interface: Entities (graph), Flow, Provisions, Precedents, Q&C, Analysis, Decisions, Narrative |
Related Pages¶
- Browsing Cases - Navigate the case repository
- Nine-Component Framework - Formal definitions of concept types
- Color Scheme - Visual coding for entity types