pranali96/Indian_Financial_News

Dataset Card for Dataset Name

<!-- Provide a quick summary of the dataset. -->

The FinancialNewsSentiment_26000 dataset comprises 26,000 rows of financial news articles related to the Indian market. It features four columns: URL, Content (scrapped content), Summary (generated using the T5-base model), and Sentiment Analysis (gathered using the GPT add-on for Google Sheets). The dataset is designed for sentiment analysis tasks, providing a comprehensive view of sentiments expressed in financial news.

Dataset Description

<!-- Provide a longer summary of what this dataset is. -->

• —Curated by: Khushi Dave
• —Language(s): English
• —Type: Text
• —Domain: Financial, Economy
• —Size: 112,293 KB
• —Dataset: Version: 1.0
• —Last Updated: 01/01/2024

Dataset Sources

<!-- Provide the basic links for the dataset. -->

• —Repository: https://huggingface.co/datasets/kdave/IndianFinancialNews

Uses

<!-- Address questions around how the dataset is intended to be used. -->

Sentiment Analysis Research: Ideal for exploring sentiment nuances in Indian financial news.

NLP Projects: Enhance NLP models with diverse financial text for improved understanding.

Algorithmic Trading Strategies: Study correlations between sentiment shifts and market movements.

News Aggregation: Generate concise summaries with sentiment insights for financial news.

Educational Resource: Hands-on examples for teaching sentiment analysis and financial text processing.

Ethical AI Exploration: Analyze biases in sentiment analysis models for ethical AI research.

Model Benchmarking: Evaluate and benchmark sentiment analysis models for financial text.

Note: Use cautiously; do not rely solely on model predictions for financial decision-making.

Dataset Creation

• —Format: String
• —Columns: URL: URL of the news article

Content: Scrapped content of the news article

Summary: Summarized version using T5-base

Sentiment Analysis: Sentiment labels (Positive, Negative, Neutral) gathered using the GPT add-on

Data Collection

<!-- This section describes the data collection and processing process such as data selection criteria, filtering and normalization methods, tools and libraries used, etc. -->

Source Selection: Aggregation of Indian financial news articles from reputable sources covering a range of topics.

URL Scrapping: Extraction of URLs for each article to maintain a connection between the dataset and the original content.

Content Scrapping: Extraction of article content for analysis and modeling purposes.

Summarization: Utilization of the T5-base model from Hugging Face for content summarization.

Sentiment Annotation: Manual sentiment labeling using the GPT add-on for Google Sheets to categorize each article as Positive, Negative, or Neutral.

Data Processing:

Cleaning and Tokenization: Standard preprocessing techniques were applied to clean and tokenize the content, ensuring uniformity and consistency.

Format Standardization: Conversion of data into a structured format with columns: URL, Content, Summary, and Sentiment Analysis.

Dataset Splitting: Given the subjective nature of sentiments, the dataset was not split into training, validation, and testing sets. Users are encouraged to customize splits based on their specific use cases.

Tools and Libraries:

Beautiful Soup: Used for web scraping to extract content from HTML. Hugging Face Transformers: Employed for summarization using the T5-base model. GPT Add-on for Google Sheets: Facilitated manual sentiment annotation. Pandas: Utilized for data manipulation and structuring.

Citation

<!-- If there is a paper or blog post introducing the dataset, the APA and Bibtex information for that should go in this section. -->

@dataset{AuthorYearFinancialNewsSentiment_26000,
  author = {Dave, Khushi},
  year = {2024},
  title = {IndiaFinanceSent Corpus},
  url = {[https://huggingface.co/datasets/kdave/Indian_Financial_News]},
}

Dataset Card Authors

Khushi Dave, Data Scientist