vislupus/alpaca-bulgarian-jokes-multilingual-prompts

Bulgarian Jokes Dataset

Overview

The Bulgarian Jokes Dataset is a collection of Bulgarian-language jokes gathered and prepared for use in training and fine-tuning natural language processing (NLP) models. This dataset is designed to help researchers and developers build models capable of understanding and generating humorous content in Bulgarian.

Dataset Structure

The dataset is structured in a format suitable for NLP training and fine-tuning tasks, such as the Alpaca format. Each data entry contains fields that provide instructions, input, and expected output to facilitate supervised training of language models. The dataset also features prompts that vary in language, including combinations of Bulgarian, English, Macedonian, and Serbian, providing a rich multilingual training resource.

Examples

{
  "instruction": "fun stuff za Австралия",
  "input": "",
  "output": "- Иванчо, погледни глобуса!\r\n- Гледам, госпожо!\r\n- Добреее... А сега ми кажи в кое полукълбо се намира Австралия!\r\n- В северното, госпожо!\r\n- Е, как в северното, бе, човек. Погледни хубаво. Ти, въобще знаеш ли къде е север?\r\n- Да, госпожо! Знам това много добре!\r\n- Тогава ти давам последен шанс да коригираш своя отговор! И така, в кое полукълбо се намира Австралия?\r\n- В северното, госпожо!\r\n- А, бе, ти тъп ли си или какво? Ще ти пиша двойка да знаеш!\r\n- Ами, пишете ми, госпожо! Аз - какво съм виновен, че държите глобуса наопаки!"
},
{
  "instruction": "урнебесне шале за Пиянски",
  "input": "",
  "output": "- Скъпи, сега вече сме женени и можеш да пиеш по-малко бира!\n- Започваш да звучиш като бившата ми жена!\n- Ооо, не знаех, че си бил женен!\n- Не съм бил!!!"
},
{
  "instruction": "смешни случки за работа",
  "input": "",
  "output": "Шефът: „Тази година няма да ви повишим заплатата, защото кризата удря всички.“\nСлужителят: „Тогава нека кризата ми върши и работата!“"
}

Features

• Language: Bulgarian
• Categories: 599 categories covering a wide range of themes and topics
• Format: Suitable for training in Alpaca-like supervised formats
• Multilingual Prompts: Includes prompts in Bulgarian, English, Macedonian, and Serbian to introduce variety and support cross-linguistic model capabilities.

Usage

This dataset can be used for various NLP tasks, including:

• Training and fine-tuning large language models (LLMs) to generate jokes in Bulgarian.
• Evaluating the performance of language models in humor generation and understanding.
• Creating NLP applications that require humorous content processing or generation.

Sources

The jokes in this dataset are collected from various sources such as humor websites, user-submitted content, and other public domain materials. Each entry was curated to ensure quality and diversity.