|
|
--- |
|
|
license: apache-2.0 |
|
|
task_categories: |
|
|
- text-classification |
|
|
language: |
|
|
- tr |
|
|
size_categories: |
|
|
- n<1K |
|
|
--- |
|
|
# Türkçe Metin Sınıflandırma Veri Seti / Turkish Text Classification Dataset |
|
|
|
|
|
**[TR]** Bu repoda, Doğal Dil İşleme (NLP) görevleri için toplanmış ve kategorize edilmiş 1000 adet Türkçe metin örneği içeren bir veri seti bulunmaktadır. Veri seti, özellikle metin sınıflandırma için makine öğrenmesi modellerini eğitmek, test etmek ve ince ayar (fine-tuning) yapmak amacıyla küçük ama çok yönlü bir kaynak olarak tasarlanmıştır. |
|
|
|
|
|
**[ENG]** This repository contains a dataset of 1,000 Turkish text samples, collected and categorized for Natural Language Processing (NLP) tasks. The dataset is designed as a small but versatile resource for training, testing, and fine-tuning machine learning models, particularly for text classification. |
|
|
|
|
|
## Genel Bakış / Overview |
|
|
|
|
|
**[TR]** Veri seti, çeşitli Türkçe web sitelerinden (haber portalları, ansiklopediler, forumlar vb.) alınmış haber metinleri, tarihsel açıklamalar ve genel ilgi alanlarına yönelik içerikler gibi çeşitli metin girişlerinden oluşur. Her bir kayıt; başlık, kaynak URL, yayınlanma tarihi ve önceden atanmış bir kategori gibi meta verileri içerir. |
|
|
|
|
|
Bu örnek veri seti aşağıdaki amaçlar için idealdir: |
|
|
- Metin sınıflandırma modelleri için prototip oluşturma. |
|
|
- Büyük Dil Modellerini (LLM'ler) Türkçe'ye özgü bir görev için ince ayar (fine-tuning) yapma. |
|
|
- Doğal Dil İşleme alanında eğitim amaçlı kullanım. |
|
|
|
|
|
**[ENG]** The dataset consists of various text entries, including news articles, historical descriptions, and general-interest topics sourced from a range of Turkish websites. Each entry is accompanied by metadata such as its title, source URL, publication date, and a pre-assigned category. |
|
|
|
|
|
This sample is ideal for: |
|
|
- Prototyping text classification models. |
|
|
- Fine-tuning Large Language Models (LLMs) on a Turkish-specific task. |
|
|
- Educational purposes in the field of NLP. |
|
|
|
|
|
## Veri Toplama Yöntemi / Data Collection Method |
|
|
|
|
|
**[TR]** Veriler, **Google Gemini API**'si ile güçlendirilmiş özel bir otonom agent (yazılım ajanı) kullanılarak toplanmıştır. Bu agent, aşağıdaki görevleri yerine getirmek üzere programlanmıştır: |
|
|
|
|
|
1. **Gezinme:** Önceden belirlenmiş Türkçe web sitelerini tarama. |
|
|
2. **Tespit Etme:** İlgili metin içeriklerini (başlık ve ana metin) çıkarma. |
|
|
3. **Analiz Etme:** Gemini'nin anlama ve muhakeme yeteneklerini kullanarak içeriği gerçek zamanlı olarak analiz etme. |
|
|
4. **Sınıflandırma:** Her metni önceden tanımlanmış kategorilerden birine atama. |
|
|
5. **Depolama:** Kaynak URL ve zaman damgası dahil olmak üzere yapılandırılmış veriyi kaydetme. |
|
|
|
|
|
Bu 1000 satırlık veri seti, agent'ın yeteneklerini göstermek amacıyla test için oluşturulmuş küçük bir örnektir. |
|
|
|
|
|
**[ENG]** The data was collected using a custom-built autonomous agent powered by the **Google Gemini API**. The agent was programmed to: |
|
|
|
|
|
1. **Browse** a curated list of Turkish websites (news portals, encyclopedias, forums, etc.). |
|
|
2. **Identify** and extract relevant textual content (titles and body text). |
|
|
3. **Analyze** the content in real-time using Gemini's reasoning capabilities. |
|
|
4. **Classify** each text into a predefined set of categories. |
|
|
5. **Store** the structured data, including the source URL and timestamp. |
|
|
|
|
|
This 1,000-row dataset is a small sample generated for testing and demonstration purposes, showcasing the agent's capabilities. |
|
|
|
|
|
## Veri Seti Yapısı / Dataset Structure |
|
|
|
|
|
**[TR]** Veri seti, `testdata3.csv` adlı tek bir CSV dosyasında sunulmaktadır. |
|
|
|
|
|
- **Format:** CSV (Virgülle Ayrılmış Değerler) |
|
|
- **Karakter Kodu:** UTF-8 |
|
|
- **Satır Sayısı:** 1000 (artı başlık satırı) |
|
|
- **Dil:** Türkçe |
|
|
|
|
|
**[ENG]** The dataset is provided in a single CSV file: `testdata3.csv`. |
|
|
|
|
|
- **Format:** CSV (Comma-Separated Values) |
|
|
- **Encoding:** UTF-8 |
|
|
- **Rows:** 1,000 (plus a header row) |
|
|
- **Language:** Turkish |
|
|
|
|
|
### Sütunlar / Columns |
|
|
|
|
|
| Sütun Adı (TR) | Column Name (EN) | Veri Tipi / Data Type | Açıklama / Description | |
|
|
|:---------------|:-----------------|:----------------------|:---------------------------------------------------------------------------------------------------------------------------------------------------| |
|
|
| `Başlık` | `Title` | `string` | **[TR]** Makalenin veya metin girişinin başlığı. <br> **[ENG]** The title of the article or text entry. | |
|
|
| `Metin` | `Text` | `string` | **[TR]** Makalenin ana içerik metni. <br> **[ENG]** The main body content of the article. | |
|
|
| `Türü` | `Type` | `string` | **[TR]** Giriş türü için bir meta veri alanı (bu örnekte tümü `high`). <br> **[ENG]** A metadata field for the entry type (consistently `high` in this sample). | |
|
|
| `Site` | `Site` | `string` | **[TR]** İçeriğin tam kaynak URL'si. <br> **[ENG]** The full source URL of the content. | |
|
|
| `Tarih` | `Date` | `string` | **[TR]** ISO 8601 formatında yayınlanma zaman damgası. <br> **[ENG]** The publication timestamp in ISO 8601 format (`YYYY-MM-DDTHH:MM:SS.sssZ`). | |
|
|
| `Kategori` | `Category` | `string` | **[TR]** Gemini agent tarafından atanan kategori. <br> **[ENG]** The category assigned by the Gemini agent. | |
|
|
|
|
|
### Kategoriler / Categories |
|
|
|
|
|
**[TR]** `Kategori` sütununda aşağıdaki etiketler bulunmaktadır: |
|
|
**[ENG]** The `Kategori` (Category) column includes the following labels: |
|
|
|
|
|
- `Tarih` (History) |
|
|
- `Gündem` (Agenda/Current Events) |
|
|
- `Teknoloji` (Technology) |
|
|
- `Kültür` (Culture) |
|
|
- `Bilim` (Science) |
|
|
- `Spor` (Sports) |
|
|
- `Sanat` (Art) |
|
|
- `Ekonomi` (Economy) |
|
|
- `Politika` (Politics) |
|
|
|
|
|
## Potansiyel Kullanım Alanları / Potential Use Cases |
|
|
|
|
|
**[TR]** |
|
|
- **Metin Sınıflandırma:** Türkçe haber ve makaleleri kategorize etmek için model eğitme. |
|
|
- **Büyük Dil Modellerini İnce Ayarlama (Fine-Tuning):** Gemma, Llama gibi modelleri veya diğer Türkçe destekli modelleri, alana özgü daha iyi bir anlayış için bu veri setiyle eğitme. |
|
|
- **Duygu Analizi:** Duygu etiketleri olmasa da, metinler duygu analizi projeleri için bir temel olarak kullanılabilir. |
|
|
- **Eğitim Aracı:** Türkçe NLP alanına yeni başlayan öğrenciler ve araştırmacılar için harika bir başlangıç noktasıdır. |
|
|
|
|
|
**[ENG]** |
|
|
- **Text Classification:** Train models to categorize Turkish news and articles. |
|
|
- **Fine-Tuning LLMs:** Use this dataset to fine-tune models like Gemma, Llama, or other Turkish-supported models for improved domain-specific understanding. |
|
|
- **Sentiment Analysis:** Although not labeled for sentiment, the text can be used as a base for sentiment analysis projects. |
|
|
- **Educational Tool:** A great starting point for students and researchers venturing into Turkish NLP. |
|
|
|
|
|
## Nasıl Kullanılır? / How to Use |
|
|
|
|
|
**[TR]** Aşağıda, Python ve `pandas` kütüphanesi kullanarak veri setini nasıl yükleyip inceleyebileceğinize dair basit bir örnek bulunmaktadır. |
|
|
|
|
|
**[ENG]** Here's a simple example of how to load and explore the dataset using Python and the `pandas` library. |
|
|
|
|
|
```python |
|
|
import pandas as pd |
|
|
|
|
|
# Load the dataset from the CSV file |
|
|
try: |
|
|
df = pd.read_csv('testdata3.csv') |
|
|
except FileNotFoundError: |
|
|
print("Error: testdata3.csv not found. Make sure the file is in the correct directory.") |
|
|
exit() |
|
|
|
|
|
# Display the first 5 rows |
|
|
print("--- First 5 Rows ---") |
|
|
print(df.head()) |
|
|
|
|
|
# Display dataset information and columns |
|
|
print("\n--- Dataset Info ---") |
|
|
df.info() |
|
|
|
|
|
# Display the distribution of categories |
|
|
print("\n--- Category Distribution ---") |
|
|
print(df['Kategori'].value_counts()) |
|
|
``` |
|
|
|
|
|
## Notlar ve Sorumluluk Reddi / Notes & Disclaimers |
|
|
|
|
|
**[TR]** |
|
|
- Bu veri seti otomatik bir agent tarafından oluşturulmuştur ve yalnızca **eğitim ve araştırma amaçlıdır**. |
|
|
- Kategorizasyon tamamen otomatiktir ve hatalar içerebilir. İnsan tarafından bir doğrulama yapılmamıştır. |
|
|
- Veri seti, agent'ın işlevselliğini göstermek amacıyla kısa bir sürede toplandığı için tekrar eden kayıtlar içermektedir. Model eğitimi için veriyi ön işleyerek bu kopyaları kaldırmanız önerilir. |
|
|
- İçeriklerin hakları `Site` sütununda belirtilen orijinal yayıncılara aittir. |
|
|
|
|
|
**[ENG]** |
|
|
- This dataset was generated by an automated agent and is intended for **educational and research purposes only**. |
|
|
- The categorization is fully automated and may contain inaccuracies. Human verification has not been performed. |
|
|
- The dataset contains duplicate entries, as it was collected over a short period to demonstrate the agent's functionality. For model training, it is recommended to preprocess the data and remove these duplicates. |
|
|
- The content belongs to the original publishers, as indicated by the `Site` column. |