tugrulkaya/Turkish-clinicalNER

Turkish Clinical Notes NER - Annotation Guideline

Version: 1.0
Date: January 2025
Language: Turkish (clinical text) / English (documentation)
Project: Turkish Clinical Notes NER Dataset

---

1. Project Overview

1.1 Objective

To establish a standardized guideline for consistent and accurate annotation of medical entities (named entities) in Turkish clinical notes.

1.2 Scope

This guideline covers the following clinical document types:

• Hospital discharge summaries (epikriz)
• Outpatient clinic notes
• Consultation reports
• Emergency department records
• Admission/discharge summaries
• Operative notes

1.3 Target Annotators

• Medical students / Medical school graduates
• Nurses and healthcare professionals
• Medical documentation specialists
• Native Turkish speakers with medical knowledge

---

2. Entity Types

This project defines 8 main entity types:

Code	Entity Type	Color	Description
DIS	DISEASE	🔴 Red	Diseases, symptoms, complaints
MED	MEDICATION	🟢 Green	Drug names, active ingredients
DOS	DOSAGE	🟡 Yellow	Dose, amount, frequency, duration
ANA	ANATOMICAL_LOCATION	🔵 Blue	Body parts, organs
PRO	PROCEDURE	🟣 Purple	Procedures, tests, surgeries
LAB	LABORATORY	🟠 Orange	Test results, values
TEM	TEMPORAL	⚪ Gray	Temporal expressions
PAT	PATIENT_CHARACTERISTIC	⚫ Black	Demographic features

---

3. Entity Definitions and Rules

3.1 DISEASE (Hastalık/Semptom) 🔴

Definition

Diseases, syndromes, symptoms, complaints, and pathological conditions.

Include

Category	Examples (Turkish)
Diagnosis names	tip 2 diyabet, hipertansiyon, pnömoni, KOAH
Symptoms	ateş, öksürük, baş ağrısı, bulantı, kusma
Syndromes	metabolik sendrom, nefrotik sendrom, ARDS
Pathological findings	hepatomegali, ödem, ral, ronküs, üfürüm
Complications	sepsis, ARDS, DIC

Exclude

• Normal findings: "kalp sesleri doğal" (heart sounds normal), "batın rahat" (abdomen soft)
• Negated conditions: "ateş yok" (no fever), "öksürük bulunmamaktadır" (no cough)
• Risk factors (if not a disease): "obezite" (only if BMI>30, include as disease)

Examples

✅ Correct Examples:
- "Hasta [şiddetli baş ağrısı]_DISEASE şikayetiyle başvurdu."
  (Patient presented with [severe headache]_DISEASE complaint.)
- "[Tip 2 DM]_DISEASE tanısı mevcut."
  ([Type 2 DM]_DISEASE diagnosis present.)
- "Akciğer bazallerinde [ral]_DISEASE duyuldu."
  ([Rales]_DISEASE heard at lung bases.)
- "[Akut miyokard infarktüsü]_DISEASE ön tanısı konuldu."
  (Preliminary diagnosis of [acute myocardial infarction]_DISEASE.)

❌ Incorrect Examples:
- "Ateş yok" → fever NOT tagged (negation)
- "Kalp sesleri doğal" → NOT tagged (normal finding)
- "[Tip 2 diyabet tanısı]_DISEASE" → "tanısı" should not be included

Boundary Rules

Situation	Rule	Example
Severity modifier	INCLUDE	[şiddetli baş ağrısı]_DISEASE
Acute/Chronic	INCLUDE	[akut pankreatit]_DISEASE
Localization	See below*

Localization decision:

Option 1 (Simple - Preferred): [Sol alt lob pnömonisi]_DISEASE (single entity)
Option 2 (Detailed): [Sol alt lob]_ANATOMICAL [pnömonisi]_DISEASE (separate)

→ This project uses Option 1 (simple approach) for V1.

---

3.2 MEDICATION (İlaç) 🟢

Definition

Drug names, active ingredients, brand names, and specific drug classes.

Include

Category	Examples (Turkish)
Generic names	metformin, amoksisilin, atorvastatin
Brand names	Glucophage, Augmentin, Lipitor
Specific drug classes	ACE inhibitörü, beta bloker (when prescribed)
Vaccines	COVID-19 aşısı, tetanoz aşısı
Blood products	eritrosit süspansiyonu, TDP

Exclude

• General treatment expressions: "antibiyotik tedavisi" (antibiotic therapy)
• Dosage information: separate entity (DOSAGE)
• Treatment modalities: "kemoterapi", "radyoterapi" → PROCEDURE

Examples

✅ Correct Examples:
- "[Metformin]_MEDICATION [1000 mg]_DOSAGE [2x1]_DOSAGE başlandı."
  ([Metformin]_MEDICATION [1000 mg]_DOSAGE [twice daily]_DOSAGE started.)
- "[Aspirin]_MEDICATION kesildi."
  ([Aspirin]_MEDICATION discontinued.)
- "Hastaya [ramipril]_MEDICATION önerildi."
  ([Ramipril]_MEDICATION recommended to patient.)

❌ Incorrect Examples:
- "Antibiyotik tedavisi planlandı" → general expression, NOT tagged
- "[Metformin 1000 mg]_MEDICATION" → dosage should be separate
- "Antihipertansif başlandı" → not specific enough

Medication + Dosage Relationship

Correct format:
[Drug_name]_MEDICATION [amount]_DOSAGE [frequency]_DOSAGE [duration]_DOSAGE [route]_DOSAGE

Example:
[Amoksisilin]_MEDICATION [500 mg]_DOSAGE [3x1]_DOSAGE [oral]_DOSAGE [7 gün]_DOSAGE

---

3.3 DOSAGE (Dozaj) 🟡

Definition

Drug dose, amount, administration frequency, duration, and route of administration.

Include

Category	Examples (Turkish)
Amount	500 mg, 1 g, 10 mL, 2 tablet, 1 ampul
Frequency	2x1, 3x1, günde 3 kez, 8 saatte bir, sabah-akşam
Duration	7 gün, 2 hafta, 1 ay boyunca
Route	oral, IV, IM, SC, topikal, inhaler
Special instructions	aç karnına, yemeklerden sonra

Examples

✅ Correct Examples:
- "[Parasetamol]_MEDICATION [500 mg]_DOSAGE [3x1]_DOSAGE [oral]_DOSAGE"
- "[5 gün]_DOSAGE süreyle kullanılacak" (to be used for [5 days]_DOSAGE)
- "[IV]_DOSAGE yoldan uygulandı" (administered via [IV]_DOSAGE route)
- "[Aç karnına]_DOSAGE alınacak" (to be taken [on empty stomach]_DOSAGE)

❌ Incorrect Examples:
- "Yüksek doz" (high dose) → vague, NOT tagged
- "Uygun dozda" (appropriate dose) → vague, NOT tagged

Note

Each DOSAGE entity should logically relate to a MEDICATION entity. This relationship should be recorded if relation annotation is performed.

---

3.4 ANATOMICAL_LOCATION (Anatomik Bölge) 🔵

Definition

Body regions, organs, anatomical structures, and systems.

Include

Category	Examples (Turkish)
Organs	karaciğer, akciğer, kalp, böbrek, beyin
Regions	sol kol, sağ alt kadran, lumbar bölge, epigastrium
Structures	femoral arter, portal ven, safra kesesi
Systems	sindirim sistemi, solunum sistemi (when specifically mentioned)

Examples

✅ Correct Examples:
- "[Sol akciğer]_ANATOMICAL bazalinde ral mevcut."
  (Rales present at [left lung]_ANATOMICAL base.)
- "[Karın]_ANATOMICAL muayenesinde hassasiyet var."
  (Tenderness on [abdominal]_ANATOMICAL examination.)
- "[Sağ alt kadran]_ANATOMICAL ağrısı mevcut."
  ([Right lower quadrant]_ANATOMICAL pain present.)

❌ Incorrect Examples:
- "Sistemik muayene" → too general
- "Genel durum" → not anatomical

Lateralization Rules

Situation	Rule
Right/Left specified	INCLUDE: [sol akciğer]_ANATOMICAL
Bilateral	INCLUDE: [her iki akciğer]_ANATOMICAL
Upper/Lower/Middle	INCLUDE: [sol üst lob]_ANATOMICAL

---

3.5 PROCEDURE (Prosedür) 🟣

Definition

Surgical procedures, diagnostic tests, therapeutic procedures, and imaging modalities.

Include

Category	Examples (Turkish)
Surgeries	apendektomi, kolesistektomi, KABG, TDP
Imaging	toraks BT, batın USG, kraniyal MR, PAAG
Endoscopic	gastroskopi, kolonoskopi, ERCP
Interventional	biyopsi, parasentez, torasentez
Cardiac	EKG, EKO, koroner anjiyografi
Lab tests (names)	hemogram, biyokimya, TİT, idrar kültürü

Exclude

• Result values → LABORATORY
• General expressions: "tetkik istendi" (tests ordered), "görüntüleme yapıldı" (imaging done)

Examples

✅ Correct Examples:
- "[Toraks BT]_PROCEDURE çekildi." ([Chest CT]_PROCEDURE performed.)
- "Hastaya [laparoskopik kolesistektomi]_PROCEDURE yapıldı."
  ([Laparoscopic cholecystectomy]_PROCEDURE performed.)
- "[EKG]_PROCEDURE'de ST elevasyonu görüldü."
  (ST elevation seen on [ECG]_PROCEDURE.)
- "[Hemogram]_PROCEDURE ve [biyokimya]_PROCEDURE istendi."

❌ Incorrect Examples:
- "Tetkikleri istendi" → general, NOT tagged
- "Görüntülemesi var" → not specific

---

3.6 LABORATORY (Laboratuvar) 🟠

Definition

Test results, measurement values, laboratory findings, and vital signs.

Include

Category	Examples (Turkish)
Numeric values	HbA1c: 7.2, Hgb: 10.5 g/dL, WBC: 15000
Qualitative values	pozitif, negatif, yüksek, düşük, normal
Vital signs	TA: 140/90 mmHg, ateş: 38.5°C, nabız: 95/dk
Imaging findings	infiltrasyon, konsolidasyon, kitle

Format Rules

Test name + value tagged together:
✅ [HbA1c: 7.2]_LABORATORY
✅ [WBC: 15000/mm³]_LABORATORY
✅ [TA: 140/90 mmHg]_LABORATORY

Qualitative only:
✅ CRP [pozitif]_LABORATORY
✅ Sedimentasyon [yüksek]_LABORATORY

Examples

✅ Correct Examples:
- "[WBC: 15000/mm³]_LABORATORY saptandı."
- "[HbA1c %8.5]_LABORATORY ile diyabet regülasyonu kötü."
- "Troponin [pozitif]_LABORATORY bulundu."
- "[TA: 160/100 mmHg]_LABORATORY ölçüldü."
- "BT'de [sağ alt lobda konsolidasyon]_LABORATORY görüldü."

❌ Incorrect Examples:
- "Laboratuvar değerleri normal" → not specific
- "Tetkikleri normal sınırlarda" → not specific

---

3.7 TEMPORAL (Tarih/Süre) ⚪

Definition

Dates, durations, temporal expressions, and frequency indicators.

Include

Category	Examples (Turkish)
Exact dates	15.03.2024, Mart 2024, 2019 yılında
Relative durations	3 gündür, 2 haftadır, son 1 ayda
Starting points	dün gece, bu sabah, geçen hafta
Frequency	haftada bir, ayda 2 kez, yılda bir
Age reference	5 yaşından beri, çocukluktan beri

Examples

✅ Correct Examples:
- "[3 gündür]_TEMPORAL devam eden ateş" (fever ongoing for [3 days]_TEMPORAL)
- "[2019 yılından]_TEMPORAL beri diyabet tanılı" (diabetic since [2019]_TEMPORAL)
- "Şikayetler [son 1 haftada]_TEMPORAL arttı" (complaints increased in [last 1 week]_TEMPORAL)
- "[Dün gece]_TEMPORAL başlayan göğüs ağrısı" (chest pain starting [last night]_TEMPORAL)

❌ Incorrect Examples:
- "Uzun süredir" (for a long time) → vague
- "Bir süredir" (for some time) → vague

---

3.8 PATIENT_CHARACTERISTIC (Kişisel Özellik) ⚫

Definition

Patient's demographic features, habits, and medically relevant personal information.

Include

Category	Examples (Turkish)
Age	45 yaşında, 65 y, 3 aylık
Gender	erkek, kadın, E, K
Age + Gender	45 yaşında erkek
Smoking	20 paket/yıl sigara, sigara içiyor
Alcohol	alkol kullanıyor, sosyal içici
Occupation (medically relevant)	madenci, çiftçi, sağlık çalışanı
Obstetric	G3P2, nullipar, postmenopozal

Examples

✅ Correct Examples:
- "[45 yaşında erkek]_PATIENT hasta başvurdu."
  ([45-year-old male]_PATIENT patient presented.)
- "Hasta [20 paket/yıl sigara]_PATIENT öyküsü mevcut."
  (Patient has [20 pack-year smoking]_PATIENT history.)
- "[G2P1]_PATIENT gebe hasta." ([G2P1]_PATIENT pregnant patient.)
- "[Emekli madenci]_PATIENT hastada silikozis şüphesi."
  (Silicosis suspected in [retired miner]_PATIENT patient.)

❌ Incorrect Examples:
- "Hasta" alone → NOT tagged
- "Bilinç açık" (conscious) → examination finding, not patient characteristic

---

4. General Annotation Rules

4.1 Tagging Format: BIO Schema

This project uses the BIO (Beginning-Inside-Outside) tagging schema.

Tag	Meaning	Usage
B-XXX	Beginning	First token of an entity
I-XXX	Inside	Continuation tokens of an entity
O	Outside	Non-entity tokens

Example

Text: "45 yaşında erkek hasta 3 gündür şiddetli baş ağrısı şikayetiyle başvurdu"
(45-year-old male patient presented with severe headache complaint for 3 days)

Token:     45       yaşında  erkek   hasta  3      gündür   şiddetli  baş     ağrısı  şikayetiyle  başvurdu
BIO Tag:   B-PAT    I-PAT    I-PAT   O      B-TEM  I-TEM    B-DIS     I-DIS   I-DIS   O            O

4.2 Tokenization Rules

• Words are separated by whitespace
• Punctuation marks are treated as separate tokens
• Number + unit may stay together: "500mg" → single token or "500 mg" → two tokens

Example tokenization:
"HbA1c: 7.2%" → ["HbA1c", ":", "7.2", "%"] or ["HbA1c:", "7.2%"]

Recommended: ["HbA1c", ":", "7.2", "%"] (detailed tokenization)

4.3 Boundary Rules

Rule 1: Minimum Span Principle

Tag the smallest meaningful unit, exclude unnecessary words.

✅ "[Tip 2 diyabet]_DISEASE tanısı var" ([Type 2 diabetes]_DISEASE diagnosis present)
❌ "[Tip 2 diyabet tanısı]_DISEASE var"

✅ "[Metformin]_MEDICATION tedavisi alıyor" (receiving [Metformin]_MEDICATION therapy)
❌ "[Metformin tedavisi]_MEDICATION alıyor"

Rule 2: Maximum Specificity Principle

If an entity is fully expressed, include the complete expression.

✅ "[Kronik obstrüktif akciğer hastalığı]_DISEASE" (Chronic obstructive pulmonary disease)
❌ "[Kronik]_? [obstrüktif akciğer hastalığı]_DISEASE"

✅ "[Laparoskopik kolesistektomi]_PROCEDURE"
❌ "[Laparoskopik]_? [kolesistektomi]_PROCEDURE"

Rule 3: Modifier Inclusion

Modifiers that change medical meaning are included in the entity.

Modifier Type	Example	Rule
Severity	şiddetli, hafif, orta	INCLUDE
Temporal	akut, kronik, subakut	INCLUDE
Lateralization	sağ, sol, bilateral	INCLUDE
Localization	üst, alt, anterior	INCLUDE

✅ [şiddetli baş ağrısı]_DISEASE (severe headache)
✅ [akut miyokard infarktüsü]_DISEASE (acute myocardial infarction)
✅ [sol akciğer]_ANATOMICAL (left lung)
✅ [kronik böbrek yetmezliği]_DISEASE (chronic kidney failure)

---

5. Special Cases

5.1 Negation

Basic Rule

Negated entities are NOT TAGGED.

❌ "Ateş yok" (no fever) → fever NOT tagged
❌ "Öksürük şikayeti bulunmamaktadır" (no cough complaint) → cough NOT tagged
❌ "DM tanısı yoktur" (no DM diagnosis) → DM NOT tagged
❌ "Baş ağrısı tarif etmiyor" (does not describe headache) → headache NOT tagged

Turkish Negation Expressions

Expression	Example
yok	"Ateş yok" (no fever)
bulunmamaktadır	"Ödem bulunmamaktadır" (no edema)
mevcut değil	"Patoloji mevcut değil" (no pathology)
saptanmadı	"Kitle saptanmadı" (no mass detected)
izlenmedi	"Patoloji izlenmedi" (no pathology observed)
tarif etmiyor	"Ağrı tarif etmiyor" (does not describe pain)
-mIyor/-mAdI	"Öksürmüyor" (not coughing), "Kusmadı" (did not vomit)

Exception: History Inquiry

Disease names in family history or past history may be tagged even if negative:

✅ "Ailede [diyabet]_DISEASE öyküsü yok" (no family history of [diabetes]_DISEASE) → acceptable
✅ "Geçirilmiş [MI]_DISEASE öyküsü sorgulandı, yoktu" (history of [MI]_DISEASE inquired, negative) → acceptable

Note: A negation attribute (neg=True/False) may be added in future versions.

5.2 Uncertainty / Suspicion

Suspected or probable diagnoses ARE TAGGED.

✅ "[Pnömoni]_DISEASE şüphesi ile yatırıldı" (admitted with suspicion of [pneumonia]_DISEASE)
✅ "Olası [akut koroner sendrom]_DISEASE" (probable [acute coronary syndrome]_DISEASE)
✅ "[Apandisit]_DISEASE ekarte edilemedi" ([appendicitis]_DISEASE could not be ruled out)
✅ "Ayırıcı tanıda [pankreatit]_DISEASE düşünüldü" ([pancreatitis]_DISEASE considered in differential)

5.3 Coordination (And/Or Conjunctions)

Entities connected by conjunctions are tagged SEPARATELY.

✅ "[Ateş]_DISEASE ve [öksürük]_DISEASE şikayeti var"
   ([Fever]_DISEASE and [cough]_DISEASE complaint present)
✅ "[Metformin]_MEDICATION ve [gliklazid]_MEDICATION kullanıyor"
   (Using [Metformin]_MEDICATION and [gliclazide]_MEDICATION)
✅ "[Karaciğer]_ANATOMICAL ve [dalak]_ANATOMICAL normal"

❌ "[Ateş ve öksürük]_DISEASE" → INCORRECT
❌ "[Metformin ve gliklazid]_MEDICATION" → INCORRECT

5.4 Abbreviations

Abbreviations are tagged THE SAME WAY as their full forms.

Abbreviation	Full Form (Turkish)	Entity
DM	Diyabetes Mellitus	DISEASE
HT	Hipertansiyon	DISEASE
KAH	Koroner Arter Hastalığı	DISEASE
KOAH	Kr. Obstrüktif Akciğer Hast.	DISEASE
KBY	Kronik Böbrek Yetmezliği	DISEASE
MI	Miyokard İnfarktüsü	DISEASE
SVO	Serebrovasküler Olay	DISEASE
BT	Bilgisayarlı Tomografi	PROCEDURE
MR	Manyetik Rezonans	PROCEDURE
USG	Ultrasonografi	PROCEDURE
EKG	Elektrokardiyografi	PROCEDURE

✅ "[DM]_DISEASE tanılı hasta" ([DM]_DISEASE diagnosed patient)
✅ "[Diyabetes mellitus]_DISEASE tanılı hasta"

5.5 Nested Entities

This project uses the FLAT approach (V1).

Text: "Sol akciğer alt lob pnömonisi" (Left lower lobe pneumonia)

Flat Approach (V1 - Preferred):
→ [Sol akciğer alt lob pnömonisi]_DISEASE

Nested Approach (V2 - Future):
→ [[Sol akciğer alt lob]_ANATOMICAL pnömonisi]_DISEASE

5.6 Discontinuous Entities

Discontinuous entities are tagged SEPARATELY.

Text: "Sağ ve sol akciğerde ral duyuldu" (Rales heard in right and left lung)

→ [Sağ]_ANATOMICAL ve [sol akciğerde]_ANATOMICAL [ral]_DISEASE duyuldu

Text: "Karaciğer ve dalak büyük palpe edildi" (Liver and spleen palpated enlarged)

→ [Karaciğer]_ANATOMICAL ve [dalak]_ANATOMICAL büyük palpe edildi

---

6. Annotation Workflow

6.1 Step-by-Step Process

┌─────────────────────────────────────┐
│ STEP 1: Read Text Completely        │
│ - Understand context                │
│ - Grasp overall structure           │
└──────────────┬──────────────────────┘
               ▼
┌─────────────────────────────────────┐
│ STEP 2: First Pass                  │
│ - DISEASE (most prominent)          │
│ - MEDICATION (easy to identify)     │
└──────────────┬──────────────────────┘
               ▼
┌─────────────────────────────────────┐
│ STEP 3: Second Pass                 │
│ - ANATOMICAL_LOCATION               │
│ - PROCEDURE                         │
│ - LABORATORY                        │
└──────────────┬──────────────────────┘
               ▼
┌─────────────────────────────────────┐
│ STEP 4: Third Pass                  │
│ - TEMPORAL                          │
│ - PATIENT_CHARACTERISTIC            │
│ - DOSAGE                            │
└──────────────┬──────────────────────┘
               ▼
┌─────────────────────────────────────┐
│ STEP 5: Negation Check              │
│ - Remove negated entities           │
│ - Check exceptions                  │
└──────────────┬──────────────────────┘
               ▼
┌─────────────────────────────────────┐
│ STEP 6: Final Review                │
│ - Boundary consistency              │
│ - Missing entity check              │
│ - Tag accuracy                      │
└─────────────────────────────────────┘

6.2 Annotation Tools

Tool	Features	Recommendation
Argilla	HF integration, collaborative	⭐ Recommended
Label Studio	Self-hosted, free	Alternative
Prodigy	Fast, active learning	Paid
Doccano	Simple, open source	For beginners

---

7. Quality Control

7.1 Inter-Annotator Agreement (IAA)

• Minimum 2 annotators should independently tag the same text
• Target Cohen's Kappa ≥ 0.80
• Disagreements resolved by expert adjudicator

IAA Calculation:
- Entity exact match: Strict F1
- Entity partial match: Relaxed F1
- Token level: Cohen's Kappa

7.2 Quality Metrics

Metric	Target	Description
Exact Match F1	≥ 0.85	Entity boundaries exactly match
Partial Match F1	≥ 0.90	Entity overlap sufficient
Label Accuracy	≥ 0.95	Correct entity type

7.3 Common Errors Checklist

Error Type	Example	Correction
Over-tagging	Tagging normal findings	Only tag pathological conditions
Under-tagging	Missing abbreviations	DM, HT, KAH should always be tagged
Boundary error	Including "tanısı" in entity	"Tip 2 diyabet" is sufficient
Negation error	Tagging "ateş" in "Ateş yok"	Don't tag negated entities
Wrong type	Symptom → PROCEDURE	Check entity definitions
Coordination	"A ve B" as single entity	Tag separately

7.4 Review Checklist

For each completed annotation:

• All diseases/symptoms tagged?
• All medications tagged?
• Dosages associated with relevant medications?
• Negation rules applied?
• Abbreviations tagged?
• Boundaries correct (minimum span)?
• Coordinated entities tagged separately?

---

8. Full Annotation Examples

Example 1: Emergency Department Note

Raw Text:

45 yaşında erkek hasta, 3 gündür devam eden şiddetli göğüs ağrısı 
şikayetiyle acil servise başvurdu. Özgeçmişinde tip 2 DM ve HT mevcut. 
Metformin 1000 mg 2x1 ve ramipril 5 mg 1x1 kullanıyormuş. Sigara: 25 
paket/yıl. Yapılan EKG'de V1-V4'te ST elevasyonu saptandı. 
Troponin I: 2.5 ng/mL (normal <0.04) bulundu. TA: 150/95 mmHg. 
Akut anterior MI ön tanısıyla kardiyolojiye yatırıldı.

English Translation:

45-year-old male patient presented to emergency department with severe 
chest pain ongoing for 3 days. Past medical history includes type 2 DM 
and HT. Using Metformin 1000 mg twice daily and ramipril 5 mg once daily. 
Smoking: 25 pack-years. ECG showed ST elevation in V1-V4. 
Troponin I: 2.5 ng/mL (normal <0.04). BP: 150/95 mmHg. 
Admitted to cardiology with preliminary diagnosis of acute anterior MI.

Annotated:

[45 yaşında erkek]_PATIENT hasta, [3 gündür]_TEMPORAL devam eden 
[şiddetli göğüs ağrısı]_DISEASE şikayetiyle acil servise başvurdu. 
Özgeçmişinde [tip 2 DM]_DISEASE ve [HT]_DISEASE mevcut. 
[Metformin]_MEDICATION [1000 mg]_DOSAGE [2x1]_DOSAGE ve [ramipril]_MEDICATION 
[5 mg]_DOSAGE [1x1]_DOSAGE kullanıyormuş. [Sigara: 25 paket/yıl]_PATIENT. 
Yapılan [EKG]_PROCEDURE'de [V1-V4'te ST elevasyonu]_LABORATORY saptandı. 
[Troponin I: 2.5 ng/mL]_LABORATORY bulundu. [TA: 150/95 mmHg]_LABORATORY. 
[Akut anterior MI]_DISEASE ön tanısıyla kardiyolojiye yatırıldı.

BIO Format:

45          B-PAT
yaşında     I-PAT
erkek       I-PAT
hasta       O
,           O
3           B-TEM
gündür      I-TEM
devam       O
eden        O
şiddetli    B-DIS
göğüs       I-DIS
ağrısı      I-DIS
...

Example 2: Outpatient Clinic Note

Raw Text:

62 yaşında kadın hasta, 2 haftadır olan halsizlik ve nefes darlığı 
şikayetiyle başvurdu. Bilinen KOAH ve KKY tanıları var. Düzenli 
spiriva 18 mcg 1x1 inhaler ve furosemid 40 mg 1x1 kullanıyor. 
Muayenede her iki akciğer bazalinde ince raller duyuldu. 
Pretibial ödem +2. SpO2: %89. PAAG'de kardiyomegali ve pulmoner 
venöz konjesyon izlendi. KKY alevlenmesi düşünüldü.

English Translation:

62-year-old female patient presented with weakness and shortness of breath 
for 2 weeks. Known diagnoses of COPD and CHF. Regularly using spiriva 
18 mcg once daily inhaler and furosemide 40 mg once daily. 
Examination revealed fine rales at both lung bases. 
Pretibial edema +2. SpO2: 89%. Chest X-ray showed cardiomegaly and 
pulmonary venous congestion. CHF exacerbation considered.

Annotated:

[62 yaşında kadın]_PATIENT hasta, [2 haftadır]_TEMPORAL olan 
[halsizlik]_DISEASE ve [nefes darlığı]_DISEASE şikayetiyle başvurdu. 
Bilinen [KOAH]_DISEASE ve [KKY]_DISEASE tanıları var. Düzenli 
[spiriva]_MEDICATION [18 mcg]_DOSAGE [1x1]_DOSAGE [inhaler]_DOSAGE ve 
[furosemid]_MEDICATION [40 mg]_DOSAGE [1x1]_DOSAGE kullanıyor. Muayenede 
[her iki akciğer bazalinde]_ANATOMICAL [ince raller]_DISEASE duyuldu. 
[Pretibial ödem +2]_DISEASE. [SpO2: %89]_LABORATORY. 
[PAAG]_PROCEDURE'de [kardiyomegali]_LABORATORY ve 
[pulmoner venöz konjesyon]_LABORATORY izlendi. 
[KKY alevlenmesi]_DISEASE düşünüldü.

Example 3: Negation Examples

Raw Text:

Hasta ateş, öksürük, balgam şikayeti tarif etmiyor. Göğüs ağrısı yok. 
Nefes darlığı mevcut. Ailede diyabet öyküsü sorgulandı, yok. 
Sigara kullanmıyor. Muayenede ral ronküs yok. Batın rahat, defans yok.

English Translation:

Patient does not describe fever, cough, sputum complaints. No chest pain. 
Shortness of breath present. Family history of diabetes inquired, negative. 
Does not smoke. No rales or rhonchi on examination. Abdomen soft, no guarding.

Annotated:

Hasta ateş, öksürük, balgam şikayeti tarif etmiyor. Göğüs ağrısı yok. 
[Nefes darlığı]_DISEASE mevcut. Ailede [diyabet]_DISEASE öyküsü 
sorgulandı, yok. Sigara kullanmıyor. Muayenede ral ronküs yok. 
[Batın]_ANATOMICAL rahat, defans yok.

Explanation:

• "ateş, öksürük, balgam" → negated, NOT TAGGED
• "Göğüs ağrısı" → negated with "yok", NOT TAGGED
• "Nefes darlığı" → confirmed with "mevcut", TAGGED
• "diyabet" → family history inquiry, TAGGED (exception)
• "ral ronküs" → negated with "yok", NOT TAGGED
• "defans" → negated with "yok", NOT TAGGED

---

9. Frequently Asked Questions (FAQ)

Q1: How do I distinguish between disease and symptom?

A: Both are tagged under the DISEASE entity type. No distinction needed.

Q2: Should "tedavisi düzenlendi" (treatment adjusted) be tagged?

A: No. If no specific medication or procedure is mentioned, don't tag.

Q3: How is "Akciğerlerde patoloji yok" (no pathology in lungs) tagged?

A: Nothing is tagged. Both "akciğerler" and potential pathology are negated.

Q4: Which category for vital signs (BP, pulse, temperature)?

A: LABORATORY category. Example: [TA: 140/90 mmHg]_LABORATORY

Q5: Should "kronik hastalık" (chronic disease) alone be tagged?

A: No. Too general. Specific disease name must be mentioned.

Q6: How is drug allergy tagged?

A: "Penisilin alerjisi var" → [Penisilin]_MEDICATION (as allergen)

Q7: Should ICD-10 codes be tagged?

A: No. Tag disease names, not codes.

---

10. Version History

Version	Date	Changes
1.0	January 2025	Initial release

---

11. Contact

For questions and feedback:

• GitHub Issues: [repo-link]
• Email: [email]
• Hugging Face: [dataset-link]

---

Appendix A: Entity Summary Table

Entity	Code	Include Examples	Exclude Examples
DISEASE	DIS	DM, ateş, pnömoni	Normal findings, negated
MEDICATION	MED	Metformin, Aspirin	"Antibiyotik tedavisi"
DOSAGE	DOS	500 mg, 2x1, oral	"Yüksek doz"
ANATOMICAL	ANA	Akciğer, sol kol	"Genel durum"
PROCEDURE	PRO	EKG, BT, apendektomi	"Tetkik istendi"
LABORATORY	LAB	HbA1c:7.2, pozitif	"Normal" (if not specific)
TEMPORAL	TEM	3 gündür, 2024	"Uzun süredir"
PATIENT	PAT	45 yaşında erkek	"Hasta" alone

---

Appendix B: Turkish Medical Abbreviations Dictionary

Abbreviation	Full Form (Turkish)	Full Form (English)	Entity
DM	Diyabetes Mellitus	Diabetes Mellitus	DISEASE
HT	Hipertansiyon	Hypertension	DISEASE
KAH	Koroner Arter Hastalığı	Coronary Artery Disease	DISEASE
KOAH	Kr. Obst. Akciğer Hast.	COPD	DISEASE
KKY	Konjestif Kalp Yetmezliği	Congestive Heart Failure	DISEASE
KBY	Kronik Böbrek Yetmezliği	Chronic Kidney Failure	DISEASE
MI	Miyokard İnfarktüsü	Myocardial Infarction	DISEASE
SVO	Serebrovasküler Olay	Cerebrovascular Event	DISEASE
DVT	Derin Ven Trombozu	Deep Vein Thrombosis	DISEASE
PE	Pulmoner Emboli	Pulmonary Embolism	DISEASE
BT	Bilgisayarlı Tomografi	Computed Tomography	PROCEDURE
MR	Manyetik Rezonans	Magnetic Resonance	PROCEDURE
USG	Ultrasonografi	Ultrasonography	PROCEDURE
EKG	Elektrokardiyografi	Electrocardiography	PROCEDURE
EKO	Ekokardiyografi	Echocardiography	PROCEDURE
PAAG	PA Akciğer Grafisi	Chest X-Ray	PROCEDURE
TİT	Tam İdrar Tahlili	Urinalysis	PROCEDURE
TA	Tansiyon Arteriyel	Blood Pressure	LABORATORY
WBC	White Blood Cell	White Blood Cell	LABORATORY
Hgb	Hemoglobin	Hemoglobin	LABORATORY

---

Appendix C: Dataset Schema

Hugging Face Dataset Format

{
    "id": "string",           # Unique identifier (e.g., "clinical_001")
    "text": "string",         # Original clinical text in Turkish
    "tokens": ["string"],     # Tokenized text
    "ner_tags": ["string"],   # BIO tags for each token
    "metadata": {
        "source": "string",   # Data source (synthetic, textbook, etc.)
        "department": "string", # Medical department
        "document_type": "string", # Document type
        "annotator_id": "string",  # Annotator identifier
        "annotation_date": "string" # ISO date format
    }
}

Example Entry

{
    "id": "clinical_001",
    "text": "45 yaşında erkek hasta, 3 gündür şiddetli baş ağrısı şikayetiyle başvurdu.",
    "tokens": ["45", "yaşında", "erkek", "hasta", ",", "3", "gündür", "şiddetli", "baş", "ağrısı", "şikayetiyle", "başvurdu", "."],
    "ner_tags": ["B-PAT", "I-PAT", "I-PAT", "O", "O", "B-TEM", "I-TEM", "B-DIS", "I-DIS", "I-DIS", "O", "O", "O"],
    "metadata": {
        "source": "synthetic",
        "department": "emergency",
        "document_type": "admission_note",
        "annotator_id": "ann_001",
        "annotation_date": "2025-01-15"
    }
}

---

Appendix D: Entity Tag Mapping

Entity Type	BIO Tags
DISEASE	B-DIS, I-DIS
MEDICATION	B-MED, I-MED
DOSAGE	B-DOS, I-DOS
ANATOMICAL_LOCATION	B-ANA, I-ANA
PROCEDURE	B-PRO, I-PRO
LABORATORY	B-LAB, I-LAB
TEMPORAL	B-TEM, I-TEM
PATIENT_CHARACTERISTIC	B-PAT, I-PAT
Outside	O

---

This document was prepared for the Turkish Clinical Notes NER project.

© 2025 - Licensed under CC-BY-SA 4.0