2nd/10_Wiki/Topics/AI_and_ML/Bibliometrics.md

id, title, category, status, canonical_id, aliases, duplicate_of, source_trust_level, confidence_score, verification_status, tags, raw_sources, last_reinforced, github_commit, tech_stack

id	title	category	status	canonical_id	aliases	duplicate_of	source_trust_level	confidence_score	verification_status	tags	raw_sources	last_reinforced	github_commit	tech_stack
wiki-2026-0508-bibliometrics	Bibliometrics	10_Wiki/Topics	verified	self	계량서지학 citation analysis h-index impact factor altmetrics scientometrics	none	B	0.88	applied	bibliometrics citation h-index impact-factor altmetrics semantic-scholar openalex science-of-science		2026-05-10	pending	language framework Python Semantic Scholar API / OpenAlex / Scopus

Bibliometrics

📌 한 줄 통찰

"매 knowledge 의 economics". 매 citation count + h-index + impact factor 의 quantify. 매 modern: 매 altmetrics + arXiv 의 real-time. 매 Goodhart trap — 매 metric 의 game (citation cartel).

📖 핵심

매 indicator

Citation count

• 매 paper 의 cited 횟수.
• 매 simple, 매 lag.

h-index (Hirsch)

• 매 author 의 매 h-index = 매 N 개 paper 가 매 N 회 cited.
• 매 productivity + impact 의 결합.
• 매 한계: 매 self-citation, 매 field 의 difference.

i10-index

• 매 ≥10 cite 의 paper 수.

Impact Factor (IF)

• 매 journal 의 매 2-year average citation.
• 매 publisher 의 marketing.
• 매 individual paper 의 X (variance).

Eigenfactor

• 매 PageRank-like.
• 매 quality-weighted.

Altmetrics

• 매 Twitter / blog / news mention.
• 매 immediate.
• 매 attention ≠ 매 quality.

매 modern source

• Google Scholar: 매 broad, 매 noisy.
• Scopus (Elsevier): 매 paid.
• Web of Science (Clarivate): 매 paid.
• Semantic Scholar (AI2): 매 free, 매 AI-enriched.
• OpenAlex: 매 open, 매 250M+ paper.
• CrossRef: 매 DOI registry.
• arXiv: 매 preprint.
• PubMed: 매 biomedical.

매 modern issue

Citation cartel

• 매 mutual citation.
• 매 self-citation 폭발.
• 매 retractable.

Field difference

• 매 CS vs biology vs literature 의 매 base rate 다름.
• 매 normalize 필수.

Time lag

• 매 citation 의 매 5 year 의 mature.
• 매 fast field (AI) 의 less applicable.

Predatory journal

• 매 paper mill.
• 매 IF 의 fake.

LLM 의 영향

• 매 paper volume 의 explosion.
• 매 review 의 saturate.
• 매 quality control 의 break.

매 modern alternative

• Open peer review.
• Replication score.
• Code/data availability.
• Twitter / Mastodon discussion.
• YouTube 의 explainer.
• Cited by GitHub.

매 응용

1. Hiring / promotion: 매 academic.
2. Funding: 매 grant evaluation.
3. Library: 매 journal subscription.
4. National R&D: 매 country comparison.
5. Trend analysis: 매 emerging topic.
6. Knowledge graph: 매 citation network.

💻 패턴

Semantic Scholar API

import requests

def get_paper(doi):
    r = requests.get(f'https://api.semanticscholar.org/graph/v1/paper/DOI:{doi}',
                     params={'fields': 'title,authors,year,citationCount,influentialCitationCount,references,citations'})
    return r.json()

paper = get_paper('10.48550/arXiv.2206.04615')
print(f"{paper['title']}: {paper['citationCount']} citations")

OpenAlex (open citation data)

import requests

def search(query, n=20):
    r = requests.get('https://api.openalex.org/works',
                     params={'search': query, 'per_page': n,
                             'select': 'id,title,publication_year,cited_by_count,authorships'})
    return r.json()['results']

# 매 author h-index
def author_h_index(author_id):
    r = requests.get(f'https://api.openalex.org/works',
                     params={'filter': f'author.id:{author_id}', 'per_page': 200,
                             'select': 'cited_by_count'})
    citations = sorted([w['cited_by_count'] for w in r.json()['results']], reverse=True)
    h = sum(1 for i, c in enumerate(citations) if c >= i + 1)
    return h

Citation network (NetworkX)

import networkx as nx

def build_citation_network(seed_paper_id, depth=2):
    G = nx.DiGraph()
    queue = [(seed_paper_id, 0)]
    seen = set()
    
    while queue:
        pid, d = queue.pop(0)
        if pid in seen or d > depth: continue
        seen.add(pid)
        paper = get_paper(pid)
        G.add_node(pid, title=paper['title'], year=paper['year'])
        
        for ref in paper.get('references', []):
            G.add_edge(pid, ref['paperId'])
            queue.append((ref['paperId'], d + 1))
    
    return G

# 매 PageRank 의 influence
pageranks = nx.pagerank(G)
top_influential = sorted(pageranks.items(), key=lambda x: -x[1])[:10]

Altmetrics

# 매 Altmetric API
import requests

def altmetric(doi):
    r = requests.get(f'https://api.altmetric.com/v1/doi/{doi}')
    if r.status_code != 200: return None
    data = r.json()
    return {
        'score': data.get('score'),
        'twitter': data.get('cited_by_tweeters_count'),
        'news': data.get('cited_by_msm_count'),
        'blog': data.get('cited_by_feeds_count'),
    }

Field-normalized citation

def field_normalized_citation_score(paper_citations, field_avg, field_year_avg):
    """매 field + year 의 normalize."""
    expected = field_year_avg
    return paper_citations / max(expected, 1)
    
# 매 RCR (Relative Citation Ratio) — NIH 의 metric

Trend detection

def emerging_topic(papers_by_year, recent_years=3):
    """매 recent 의 acceleration 의 detect."""
    recent_count = sum(papers_by_year.get(y, 0) for y in range(2024, 2027))
    older_count = sum(papers_by_year.get(y, 0) for y in range(2020, 2024))
    
    growth = (recent_count - older_count) / max(older_count, 1)
    return growth > 1.5  # 매 2.5× growth → 매 emerging

Predatory journal detector

PREDATORY_INDICATORS = [
    'fee mentioned upfront',
    'no peer review',
    'bogus impact factor',
    'misleading scope',
    'spam emails',
]

def assess_journal(journal):
    score = 0
    if journal.has_apc and journal.apc < 100: score += 1  # too cheap
    if journal.peer_review_time < 7: score += 1  # too fast
    if journal.editorial_board_overlap > 50: score += 1
    if journal.in_doaj: score -= 2  # whitelist
    return 'predatory' if score >= 2 else 'legitimate'

🤔 결정 기준

사용	Indicator
Single paper	Citation + altmetric + influential citations
Author	h-index + i10 + field-normalized
Journal	Eigenfactor (NOT IF)
Trend	Year-over-year growth
Country	Field-normalized + collaboration
Hiring	Mix + qualitative review

기본값: OpenAlex / Semantic Scholar (free) + 매 multi-metric + 매 qualitative.

🔗 Graph

• 변형: Citation-Analysis · Altmetrics · Scientometrics
• 응용: H-Index · Impact-Factor
• Adjacent: Goodharts-Law

🤖 LLM 활용

언제: 매 literature review. 매 trend detection. 매 author / journal evaluation. 매 knowledge graph 구축. 언제 X: 매 single citation count 의 quality conclusion. 매 cross-field comparison without normalize.

❌ 안티패턴

• IF 의 individual paper 적용: 매 misleading.
• h-index 만: 매 manipulation.
• No field normalize: 매 cross-field unfair.
• Self-citation 의 ignore: 매 inflated.
• Recent paper 의 IF 평가: 매 lag X.
• Predatory 의 trust: 매 fake metric.
• Citation 의 quality 의 conflate: 매 controversial paper 의 high citation.

🧪 검증 / 중복

• Verified (Hirsch h-index, NIH RCR, San Francisco DORA declaration).
• 신뢰도 B.
• Related: Awards · Benchmarks · Goodharts-Law · Open-Science.

🕓 Changelog

날짜	변경
2026-05-08	Phase 1
2026-05-10	Manual cleanup — indicator + altmetric + 매 OpenAlex / Semantic Scholar code + predatory detector