
![Azerbaijani LLM Leaderboard Banner](../images/posts/az-llm-leaderboard/image.png)

## Bridging Technology and Language

The LLM Leaderboard Project for the Azerbaijani Language is a pioneering initiative that marks a significant milestone in natural language processing for Azerbaijani. Launched on Hugging Face under the leadership of Kapital Bank, this project addresses a critical gap in language technology.

## Project Overview

The project brings together a collaborative network of companies and universities, united by a shared vision: creating a robust evaluation framework for Azerbaijani-language large language models (LLMs).

### Key Objectives

- Establish a comprehensive evaluation standard for Azerbaijani LLMs
- Promote innovation in language technology
- Highlight the unique linguistic characteristics of Azerbaijani

## Technical Approach

Our methodology involves:
- Rigorous benchmarking of existing models
- Developing specialized evaluation metrics
- Encouraging continuous improvement in Azerbaijani language AI

## Impact and Significance

This initiative is more than a technical project—it's a statement about the importance of linguistic diversity in artificial intelligence. By focusing on Azerbaijani, we're ensuring that technological advancements are inclusive and representative.

## Explore the Leaderboard

[Visit the Azerbaijani LLM Leaderboard on Hugging Face](https://huggingface.co/spaces/LLM-Beetle/azerbaijani-llm-leaderboard)

## Future Directions

We see this project as the beginning of a broader conversation about language technology, representation, and innovation.

---

*Disclaimer: This project represents a collaborative effort to advance language technology for Azerbaijani.*

The Azerbaijani LLM Leaderboard Project


![Azerbaijani Banks QA Dataset Banner](../images/posts/azbanks-dataset/image.png)

## Overview
This project comprises question-answer dataset from five prominent banks in Azerbaijan: Kapital Bank, Unibank, YeloBank, ABB Bank, and LeoBank. Dataset contains over 25,000 samples of question-answer pairs scraped from the comments section of the respective bank's Instagram page. These dataset are valuable resources for training and evaluating question answering systems tailored to the banking sector in Azerbaijan.

### Dataset Contents
azbanks-qadata.csv contains 4 columns, question, answer, bank, datetime

### Data Collection
The comments were scraped from the official Instagram pages of the respective banks. Comments were filtered to retain only those containing questions and their corresponding answers.

### Data Format
Each question-answer pair is provided as a separate entry in the dataset.


### Usage
Researchers and developers can utilize these datasets for training and evaluating question answering models, natural language understanding systems, and other related applications within the banking domain in Azerbaijan. The datasets can be used individually or combined for more comprehensive training and evaluation.

### Download
The dataset is available for download on the following link: 

[Azerbaijani Banks QA Dataset](https://huggingface.co/datasets/arzumanabbasov/azbanks-qadata)



The Azerbaijani Banks QA Dataset


![Azerbaijani Govenmental QA Dataset Banner](../images/posts/azgov-dataset/image.png)


## Overview
This dataset contains over 5000 samples of question-answer pairs scraped from the comments section of the Instagram page of AsanXidmat, a government organization in Azerbaijan dedicated to providing services to Azerbaijani citizens. The dataset is intended for use in training and evaluating question answering systems, particularly those focused on understanding and responding to inquiries related to government services in Azerbaijan.

### Dataset Contents
The dataset is organized as follows:
> This dataset consists of 2 columns 'text' : question and 'answer':answer by asanxidmat

### Data Collection
The comments were scraped from the official Instagram page of AsanXidmat. The comments were filtered to retain only those containing questions and their corresponding answers.

### Data Format
Each question-answer pair is provided as a separate entry in the dataset. The questions and answers are stored in csv format.

### Usage
Researchers and developers can utilize this dataset for training and evaluating question answering models, natural language understanding systems, and other related applications. The dataset can be used as-is or combined with other datasets for more extensive training and evaluation.

### Download

The dataset is available for download on the following link:

[Azerbaijani Government Services QA Dataset](https://huggingface.co/datasets/arzumanabbasov/azerbaijani-gov-qa)

Azerbaijani Government Services Question Answering Dataset

<p><img src="../images/posts/azbanks-dataset/image.png" alt="Azerbaijani Banks QA Dataset Banner"></p>
<h2>Overview</h2>
<p>This project comprises question-answer dataset from five prominent banks in Azerbaijan: Kapital Bank, Unibank, YeloBank, ABB Bank, and LeoBank. Dataset contains over 25,000 samples of question-answer pairs scraped from the comments section of the respective bank’s Instagram page. These dataset are valuable resources for training and evaluating question answering systems tailored to the banking sector in Azerbaijan.</p>
<h3>Dataset Contents</h3>
<p>azbanks-qadata.csv contains 4 columns, question, answer, bank, datetime</p>
<h3>Data Collection</h3>
<p>The comments were scraped from the official Instagram pages of the respective banks. Comments were filtered to retain only those containing questions and their corresponding answers.</p>
<h3>Data Format</h3>
<p>Each question-answer pair is provided as a separate entry in the dataset.</p>
<h3>Usage</h3>
<p>Researchers and developers can utilize these datasets for training and evaluating question answering models, natural language understanding systems, and other related applications within the banking domain in Azerbaijan. The datasets can be used individually or combined for more comprehensive training and evaluation.</p>
<h3>Download</h3>
<p>The dataset is available for download on the following link:</p>
<p><a href="https://huggingface.co/datasets/arzumanabbasov/azbanks-qadata">Azerbaijani Banks QA Dataset</a></p>


The Azerbaijani Banks QA Dataset

Overview

Dataset Contents

Data Collection

Data Format

Usage

Download

More research posts

The Azerbaijani LLM Leaderboard Project

Azerbaijani Government Services Question Answering Dataset