AI ile Literatür Taraması: Akademik Araştırma Hızlandırması

Literatür taraması = araştırmanın temel taşı. Eskiden haftalar, AI ile günler. Anahtar: doğru tool + verify.

Tool Karşılaştırma

Tool	Use Case	Fiyat
Perplexity	Quick research + source	$20/ay
Elicit	Systematic review	$10-30/ay
Consensus	Q&A from research	$9/ay
Scite	Citation analysis	$20/ay
Semantic Scholar	Free search engine	Free
ResearchRabbit	Visual citation network	Free
Connected Papers	Paper relationship	Free
Litmaps	Visualize literature	$10/ay
Iris.ai	Pro research	Enterprise
PubMed	Bio/medical free	Free

Detay: Perplexity

Perplexity - Hızlı Soru

Use case: "Yapay zeka eğitimde etki son 3 yıl?"

Perplexity:
1. Soru yaz
2. 'Focus on Academic' (toggle)
3. Cevap + 10+ source
4. Her source linked
5. Follow-up question depth

Avantaj:
- 30 sn → 10 paper summary
- Source verify edilebilir
- Pro tier: GPT-4o + Claude options
- Spaces (collaborative)

Elicit - Sistematik Review

Workflow:

1. elicit.com
2. Research question yaz
3. AI 8-10 paper buluyor
4. Summary table:
   - Title, Year, Methodology, Sample, Outcome
5. Filter (peer-reviewed, year range)
6. Extract data points (custom)
7. Export (CSV, RIS)

Sistematik review yardımcı (PRISMA flow):
- Identification (AI)
- Screening (AI suggest, human decide)
- Eligibility (manual)
- Inclusion

Consensus - Q&A

Soru: "Does intermittent fasting improve cognitive function?"

Consensus:
- 'Yes': %45 of papers
- 'No': %20
- 'Possibly': %35
- Confidence: moderate

10 paper summary with verdict.

Use case:
- Clinical question (evidence base)
- Quick fact-check
- Meta-perspective
- Debate prep

Scite - Citation Analysis

Paper: "Smith (2023) AI in education"

Scite:
- Cited by 50 papers
- Of those:
  - Supporting: 35
  - Mentioning: 12
  - Contrasting: 3
- See contrasting → tartışmalı bulguları

Use case:
- Argumentation strength
- Find counter-evidence
- Track citation evolution
- Identify highly contested findings

Search Strategy

"Research question'dan PICO:

P (Population): Türk üniversite öğrencileri
I (Intervention): AI tool kullanımı
C (Comparison): Geleneksel öğrenme
O (Outcome): Akademik başarı

Search string:
("Turkish" AND "university students") 
AND ("AI" OR "ChatGPT" OR "artificial intelligence")
AND ("learning" OR "education" OR "academic")
AND ("performance" OR "achievement")

Database:
- Google Scholar (broad)
- ERIC (education)
- Scopus (multidisciplinary)
- Web of Science
- PubMed (if health-related)
- DergiPark (TR akademik)
- ULAKBİM (TR)
"

Paper Discovery

"Bu makale ile ilgili daha fazla:

Seed paper: 'AI Adoption in Higher Education'

Tools:
1. Connected Papers
   - 40 similar papers (similarity)
   - Visual network
2. ResearchRabbit
   - Citation graph
   - Author network
3. Litmaps
   - Citation timeline
4. Scholarcy
   - Summary cards

Outcome: 50 related paper, 10 must-read.

Reading Strategy

"100 paper okumak imkansız. AI prioritize:

Triage:
1. Skim title (1000 → 200)
2. Skim abstract (200 → 50)
3. Skim figures + tables (50 → 20)
4. Full read top 20

AI yardım:
- Elicit: abstract extract
- ChatGPT: summarize uzun pdf
- Scholarcy: keypoint cards
- NotebookLM: Q&A on uploaded papers
"

Detay: NotebookLM

Citation Manager

Zotero setup:

1. Zotero install
2. Browser connector (1-click save)
3. Tag system (kategori)
4. Collections (project bazlı)
5. PDF annotation
6. Word plugin (citation insert)
7. Format styles (APA, MLA, etc.)
8. Sync cloud (300MB free)

AI integration:
- ZotFile + AI plugin
- ARIA (read PDF semantic)
- Mendeley desktop AI
"

NotebookLM Akademik

Use case:

1. notebooklm.google.com
2. 50 PDF upload (papers)
3. NotebookLM her papers'a sahip
4. Sor:
   - "Methodology trends?"
   - "Yang's paper'a karşı çıkan?"
   - "5 paper'da common limitation?"
5. Source link her cevapta

Avantaj:
- Hayali citation YOK (sadece uploaded)
- Multi-source synthesis
- Audio overview (5 dk özet ses)
- Free tier (15 source)
- Plus tier: 200 source
"

Tema Sentezi

"50 paper özetinden tema:

[paste 50 abstract]

Çıktı:
1. Cluster (5-7 tema)
2. Her tema:
   - Definition
   - Key authors (3)
   - Main findings
   - Methodology trend
   - Gap
3. Paradigm shift (varsa)
4. Methodological evolution
5. Contested findings
6. Future direction (3)
"

Critical Appraisal

"Paper kritiği:

Paper: [...]

Çıktı:
1. Research question clarity
2. Methodology appropriateness
3. Sample size + selection
4. Validity (internal + external)
5. Reliability
6. Statistical analysis appropriate?
7. Findings supported by data?
8. Discussion proportional?
9. Limitations acknowledged?
10. Implications justified?

Score: 1-10 + reasoning.
"

Gap Analiz

"Literatür gap analizi:

Çalışılan: 
- Population A, B, C
- Method X, Y
- Context 1, 2

Gap (eksik):
- Population D (TR researchers?)
- Method Z (mixed methods?)
- Context 3 (small business?)
- Time period: 2023+?
- Cross-cultural compare?
- Longitudinal study?

Çıktı: 5 potential research direction.
"

PRISMA Diagram

"PRISMA flowchart:

Identification:
- Database 1 (PubMed): n=1000
- Database 2 (Scopus): n=500
- Database 3 (WoS): n=300
- Other: n=50
Total: 1850
Duplicates removed: 350
Total unique: 1500

Screening (title/abstract):
- Excluded: 1200 (reasons)
- Included: 300

Eligibility (full text):
- Excluded: 250 (reasons)
- Included: 50

Final: 50 included.

Tool: Rayyan / Covidence for screening.
"

Türkçe Akademik Kaynaklar

TR akademik:
- DergiPark (8000+ TR dergi)
- ULAKBİM TR Dizin
- YÖK Tez Merkezi
- Ankara Üniversite e-kitap
- TÜBİTAK ULAKBİM EKUAL
- Trdizin

International + TR:
- Google Scholar (TR filter)
- ResearchGate (free if author shares)
- Academia.edu

Erişim:
- Üniversite VPN
- Sci-Hub (legal gray)
- Author email request
- Interlibrary loan
"

Bibliographic Software

Tool	Free	Sync	Plugin
Zotero	✅	300MB	Word, Google Doc
Mendeley	✅	2GB	Word
EndNote	Trial	Cloud	Word
Paperpile	$3/ay	Google integrated	Docs
Citavi	Trial	Local	Word

Reading Workflow

Sabah 30 dk:
- Top 3 paper skim
- Note taking (Obsidian, Notion)

Öğleden sonra 60 dk:
- 1 paper deep read
- Critical notes
- Add to lit review section

Hafta:
- 7 paper read
- Weekly synthesis
- Update gap analysis

Ay:
- 30 paper coverage
- Monthly write-up
"

Yaygın Hatalar

1. Vanilla ChatGPT citation: Hayali
2. No verification: Sahte kaynak
3. Single database: Bias
4. Recent only: Foundational paper miss
5. No critical lens: Sadece kabul
6. PDF dump: Read yok, just organize
7. No synthesis: List = literature review değil

Sonraki Adımlar

• Akademik Makale
• Tez Yazımı
• NotebookLM
• Perplexity
• Halüsinasyonu Önleme

Özet

Literatür AI = haftalar → günler. Perplexity (hızlı) + Elicit (sistematik) + Consensus (Q&A) + NotebookLM (deep). Anahtar: source verify, critical thinking, sentez = human work. AI organizing yardım, judgment sende.