# ScholarLens: Methodology and Research Integrity

## 1. Source retrieval
Crossref `/works` metadata and DOI lookups, or OpenAlex `/works` results, supply titles, publication years, DOI, author metadata, source venue and (where available) citation counts. Coverage differs by provider. No full-text scraping occurs.

Relevant official documentation:
- https://www.crossref.org/documentation/retrieve-metadata/rest-api/
- https://www.crossref.org/documentation/retrieve-metadata/rest-api/rest-api-filters/
- https://help.openalex.org/api/

## 2. Filtering and provenance
A keyword query, provider, publication range, count limit, endpoint URL and query time are recorded in the exportable session. Result lists are capped in the interface. Provider request failures display explicit messages without fabricated results.

## 3. Data normalization and deduplication
DOIs are normalized to lowercase DOI strings. Publications are deduplicated by valid DOI or normalized title and year. If neither is sufficient, some duplicates can remain.

## 4. Manual evidence workflow
User supplies reading notes or short excerpts from a work they checked. The engine separates sentences, calculates literal word overlaps with the query, and assigns simple marker-based categories: methods, findings, limitations, objectives, or unclassified. Results retain a reference to the saved source record and sentence position. There is no scientific entailment model or hallucination-free guarantee.

## 5. Research patterns
Publication timeline = count by publication year. Keyword plot = frequency of selected non-stopword title tokens. Research overlap = Jaccard similarity of title/keyword tokens. Metadata coverage = presence or absence of five documented fields. None of these measures is equivalent to research novelty, consensus, evidence strength or paper quality.

## 6. Reproducibility
Session exports include curated records, annotations, and search provenance. CSV and bibliographic exports are generated from saved records; JSON reimport restores a compatible saved session. Output requires manual review before academic citation.

## 7. External dependencies / privacy
The app runs locally in the browser, but live searches send the query/DOI to the selected public provider. It does not transmit pasted reading notes to those APIs. Private notes remain in memory and are lost on page reload unless explicitly exported. The demonstration consists of six fictional sample records and is unmistakably marked as synthetic.

## 8. Limitations
Crossref/OpenAlex may contain inaccurate, missing or inconsistent metadata; counts change over time; APIs may throttle requests. Neither citation counts nor word similarity measure research quality. No generative LLM is used, no full-text PDF extraction is implemented, and citations have not been independently checked against original works.