research-lit

Generating a literature review from arXiv and Semantic Scholar requires aggregating multi-source results, applying deduplication, and using citation-network analysis to synthesize key methods and open questions. This process automatically expands query variants to maximize coverage.

Citation-network snowball expansion works by tracing references and citing papers to uncover hidden connections across domains. It applies gap-driven expansion to find missing literature, then structures the results into a ranked set of papers for related-work synthesis.

Local PDFs and Zotero metadata can be used for gap analysis in robotics research. The system identifies and aggregates relevant literature from these local sources alongside web results to support comprehensive discovery and cross-domain exploration.

The best way to deduplicate research papers across Obsidian vaults and web results is to aggregate them into a single pipeline. This pipeline applies metadata extraction and variant generation before deduplication, ensuring a clean set for structured synthesis.

This literature review approach fully supports cross-domain topics in machine learning. It applies automatic query-variant expansion and cross-domain exploration to maximize coverage, specifically targeting robotics, ML, and broader interdisciplinary subjects.

Query-variant expansion is needed for aggregating academic papers because initial search terms often miss relevant literature. By generating multiple query variants, the system maximizes discovery coverage across arXiv, Semantic Scholar, and local databases.