research: Riker's deep analysis — publishability assessment + literature review

[gauravsurtani]

Research Analysis Complete

Commander Riker conducted a deep analysis of the Email-Link codebase and its competitive landscape to assess publishability for O-1 evidence.

Branch

research/riker-analysis-2026-02-17 — view branch

Files Added (1,473 lines across 6 files)

File	Lines	What It Covers
research/README.md	33	Overview + next steps checklist
research/analysis/introspection.md	607	Every module analyzed — architecture, correctness, gaps
research/analysis/reading-list.md	376	20 papers annotated with threat level + citation strategy
research/analysis/gap-analysis.md	213	Novelty claims ranked + 5 vulnerability scenarios with mitigations
research/analysis/landscape-summary.md	186	Competitive map + venue strategy
research/analysis/executive-summary.md	91	Publishability verdict + O-1 assessment

Key Findings

1. The gap is real and unoccupied. Zero papers combine email data + KG construction + TransE embeddings.
2. MAILEX (EMNLP 2023) is the closest threat — email event extraction, but no KG, no embeddings. They can't model cross-email event-entity relationships. We can.
3. Evaluation is inflated. Link prediction tested on 100 triples (~2% of 24,483 entities). Needs full-entity evaluation.
4. Paper overclaims two things: "LLM-powered interface" (it's keyword dispatch) and "ML classifiers" (it's regex + spaCy).
5. Strongest angle: TransE predicting missing event participants from graph structure — novel for the email domain.
6. TransE is not outdated — SparseTransX (MLSys 2025) proves active research. RotatE is natural future work.

Publishability Verdict

Strong. With 3 targeted weeks of work:

• Expand LP evaluation to full entity set
• Add BM25/TF-IDF baseline comparison
• Address the two overclaims
  → Publishable arXiv preprint in a domain with thin prior work.

Target Venues (ranked)

1. arXiv preprint (immediate, plant the flag)
2. EMNLP 2026 workshop
3. CIKM 2026
4. IEEE BigData 2026

Next Steps

• Fix link prediction evaluation (full entity set)
• Add baseline comparisons (BM25, TF-IDF)
• Remove or implement the LLM agent claim
• Write Related Work section (cite MAILEX + Guo 2020)
• Draft arXiv preprint

---

Analysis by Commander Riker (research agent) — February 17, 2026