Citation Chaining Guide

Step 2: Backward Chaining (Reference Mining)

Examine the reference list of each seed paper and identify which cited works are relevant.

import requests

HEADERS = {"User-Agent": "ResearchPlugins/1.0 (https://wentor.ai)"}

def get_references(work_id):
    """Get all references of a paper via OpenAlex."""
    url = f"https://api.openalex.org/works/{work_id}"
    response = requests.get(url, headers=HEADERS)
    paper = response.json()
    ref_ids = paper.get("referenced_works", [])

    references = []
    for ref_id in ref_ids:
        ref = requests.get(f"https://api.openalex.org/works/{ref_id.split('/')[-1]}", headers=HEADERS).json()
        if ref.get("title"):
            references.append(ref)
    return references

# Get references of a seed paper
seed_id = "W2741809807"
references = get_references(seed_id)

# Sort by citation count to find the most influential foundations
references.sort(key=lambda p: p.get("cited_by_count", 0), reverse=True)
for ref in references[:15]:
    print(f"[{ref.get('publication_year', '?')}] {ref['title']} ({ref.get('cited_by_count', 0)} citations)")

Step 3: Forward Chaining (Citation Tracking)

Find all papers that have cited your seed paper.

def get_citations(work_id, limit=200):
    """Get papers citing a given paper via OpenAlex."""
    all_citations = []
    page = 1
    while len(all_citations) < limit:
        response = requests.get(
            "https://api.openalex.org/works",
            params={
                "filter": f"cites:{work_id}",
                "sort": "cited_by_count:desc",
                "per_page": min(200, limit - len(all_citations)),
                "page": page
            },
            headers=HEADERS
        )
        results = response.json().get("results", [])
        if not results:
            break
        all_citations.extend(results)
        page += 1
    return all_citations

citations = get_citations(seed_id)
# Filter for recent, well-cited papers
recent_impactful = [c for c in citations if c.get("publication_year", 0) >= 2022 and c.get("cited_by_count", 0) >= 5]
recent_impactful.sort(key=lambda p: p.get("cited_by_count", 0), reverse=True)

Step 4: Co-Citation and Bibliographic Coupling

Two advanced techniques extend basic citation chaining:

def find_co_cited_papers(paper_ids, min_co_citation_count=3):
    """Find papers frequently co-cited with the given papers."""
    from collections import Counter
    reference_counts = Counter()

    for pid in paper_ids:
        refs = get_references(pid)
        for ref in refs:
            ref_id = ref.get("paperId")
            if ref_id and ref_id not in paper_ids:
                reference_counts[ref_id] += 1

    # Papers cited by multiple seeds are co-cited candidates
    co_cited = [(pid, count) for pid, count in reference_counts.items()
                if count >= min_co_citation_count]
    co_cited.sort(key=lambda x: x[1], reverse=True)
    return co_cited

Step 5: Iterative Expansion

Repeat the process with the most relevant papers discovered in each round:

1. Round 1: Start with 3-5 seed papers
2. Round 2: Run backward + forward chaining on seeds, identify 10-15 new relevant papers
3. Round 3: Run chaining on the new papers from Round 2
4. Saturation: Stop when new rounds yield diminishing returns (i.e., the same papers keep appearing)

Tools for Citation Chaining

Common Pitfalls

• Citation bias: Highly cited papers are not always the best or most relevant. Pay attention to less-cited but methodologically sound papers.
• Recency bias: Forward chaining favors recent papers with fewer citations. Allow time for citation accumulation or use Mendeley readership as a proxy.
• Field boundaries: Citation chains tend to stay within disciplinary silos. Combine with keyword searches in adjacent-field databases to break out.
• Incomplete coverage: No single database indexes all citations. Cross-check with at least two sources (e.g., OpenAlex + Google Scholar).