Changes for page Architecture
Last modified by Robert Schaub on 2025/12/24 21:46
Summary
-
Page properties (1 modified, 0 added, 0 removed)
Details
- Page properties
-
- Content
-
... ... @@ -68,6 +68,70 @@ 68 68 == 3. AKEL Architecture == 69 69 {{include reference="FactHarbor.Specification.Diagrams.AKEL_Architecture.WebHome"/}} 70 70 See [[AI Knowledge Extraction Layer (AKEL)>>FactHarbor.Specification.AI Knowledge Extraction Layer (AKEL).WebHome]] for detailed information. 71 + 72 +== 3.5 Claim Processing Architecture == 73 + 74 +FactHarbor's claim processing architecture is designed to handle both single-claim and multi-claim submissions efficiently. 75 + 76 +=== Multi-Claim Handling === 77 + 78 +Users often submit: 79 +* **Text with multiple claims**: Articles, statements, or paragraphs containing several distinct factual claims 80 +* **Web pages**: URLs that are analyzed to extract all verifiable claims 81 +* **Single claims**: Simple, direct factual statements 82 + 83 +The first processing step is always **Claim Extraction**: identifying and isolating individual verifiable claims from submitted content. 84 + 85 +=== Processing Phases === 86 + 87 +**POC Implementation (Two-Phase):** 88 + 89 +Phase 1 - Claim Extraction: 90 +* LLM analyzes submitted content 91 +* Extracts all distinct, verifiable claims 92 +* Returns structured list of claims with context 93 + 94 +Phase 2 - Parallel Analysis: 95 +* Each claim processed independently by LLM 96 +* Single call per claim generates: Evidence, Scenarios, Sources, Verdict, Risk 97 +* Parallelized across all claims 98 +* Results aggregated for presentation 99 + 100 +**Production Implementation (Three-Phase):** 101 + 102 +Phase 1 - Extraction + Validation: 103 +* Extract claims from content 104 +* Validate clarity and uniqueness 105 +* Filter vague or duplicate claims 106 + 107 +Phase 2 - Evidence Gathering (Parallel): 108 +* Independent evidence gathering per claim 109 +* Source validation and scenario generation 110 +* Quality gates prevent poor data from advancing 111 + 112 +Phase 3 - Verdict Generation (Parallel): 113 +* Generate verdict from validated evidence 114 +* Confidence scoring and risk assessment 115 +* Low-confidence cases routed to human review 116 + 117 +=== Architectural Benefits === 118 + 119 +**Scalability:** 120 +* Process 100 claims with ~3x latency of single claim 121 +* Parallel processing across independent claims 122 +* Linear cost scaling with claim count 123 + 124 +**Quality:** 125 +* Validation gates between phases 126 +* Errors isolated to individual claims 127 +* Clear observability per processing step 128 + 129 +**Flexibility:** 130 +* Each phase optimizable independently 131 +* Can use different model sizes per phase 132 +* Easy to add human review at decision points 133 + 134 + 71 71 == 4. Storage Architecture == 72 72 {{include reference="FactHarbor.Specification.Diagrams.Storage Architecture.WebHome"/}} 73 73 See [[Storage Strategy>>FactHarbor.Specification.Architecture.WebHome]] for detailed information.