🏛️ AI Archaeology Lab — Architecture

Academic paper analysis and intellectual genealogy mapping. SvelteKit + Deno + Supabase + Ollama (Mac Mini) + PDF parsing.

🛠️ Tech Stack

Component	Technology	Why
Frontend	SvelteKit	Paper upload, genealogy visualizer, literature review builder
Backend	Deno 2	PDF parsing, citation extraction, analysis orchestration
Database	Supabase (PostgreSQL)	Papers, citations, genealogy trees, user analyses
AI	Ollama (Mac Mini)	Paper deep-reading, novelty detection, literature synthesis
PDF Parsing	PdfParse + custom OCR	Extract text and citations from academic PDFs
Citation Data	Semantic Scholar API	Forward citations, paper metadata, citation context

🔬 Analysis Pipeline

1. PDF → Structured Text

Parse PDF: extract abstract, body text, references section. Identify citation patterns. Build a structured representation of the paper.

2. Deep Read (Ollama)

Ollama reads full paper. Identifies: research question, methodology, key results, limitations, actual (vs stated) contribution. Not surface-level — understands what the paper actually did.

3. Citation Trace

Extract all citations. Query Semantic Scholar API for each cited paper → get metadata, abstract, citations count. Build backward genealogy.

4. Novelty Detection

Compare paper's claims against cited work. Ollama identifies: what does this paper add that's new, how does it extend/refute prior work, what's the intellectual delta.

5. Forward Citation Map

Query Semantic Scholar for papers that cite this one. Categorize: who extended the methodology, who applied to new domains, who criticized. Build forward genealogy.

6. Synthesis

All outputs combined: genealogy map, novelty summary, significance explanation, plain-language overview. Presented in the web UI.

← Requirements | All Projects | Presentation →