---
dataset_info:
  features:
  - name: title
    dtype: string
  - name: sentiment
    dtype: int64
  splits:
  - name: train
    num_bytes: 173969.97117694156
    num_examples: 1998
  - name: test
    num_bytes: 43536.02882305845
    num_examples: 500
  download_size: 140919
  dataset_size: 217506.0
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
  - split: test
    path: data/test-*
license: apache-2.0
task_categories:
- text-classification
language:
- zh
tags:
- finance
size_categories:
- 1K<n<10K
---

Dataset Summary
This dataset contains 2,500 news titles scraped from NBD Stocks and labeled with AI-generated sentiment annotations. Each title is categorized into one of three sentiment classes:

0: Negative

1: Neutral

2: Positive

The dataset can be used for tasks such as financial sentiment analysis, text classification, and AI model training for financial market prediction.

Data Collection
The data was collected using custom Python scripts to scrape news headlines from NBD Stocks. Copilot was used to generate sentiment labels based on headline content.

Dataset Format
The dataset is provided in CSV and Hugging Face Dataset format. It consists of:

title: The news headline

label: Sentiment category (0, 1, 2)

Example structure:

title	label
"Tech stocks fall after inflation report"	0
"Market remains steady despite global tensions"	1
"Company X reports record-breaking profits"	2
Data Splits
The dataset is randomly shuffled and split into:

Training Set: 80% (2000 samples)

Validation Set: 20% (500 samples)

Usage
This dataset can be used for:

Financial market sentiment analysis

Training AI models for stock trend predictions

Research in NLP and financial journalism