TiLt-HS

TiLt-HS (Tests in Lithuanian, High School) is a dataset of multiple-choice question tests that are used to assess the knowledge of high school students in several academic areas.

Dataset Details

The dataset was collected in August, 2024.

Dataset Description

Point of contact: Jekaterina Novikova

Language: Lithuanian (lt)

License: CC-BY-NC-SA-4.0

Uses

Direct Use

The dataset is intended to be used as a subset of training data for the development of multilingual language models.

It can be used together with the TiLt-Pro dataset, as both follow the exact same structure.

How to Use the Dataset

Huggingface Datasets:

from datasets import load_dataset

ds = load_dataset("Jekaterina/tilt-hs")

Pandas:

import pandas as pd

df = pd.read_json("hf://datasets/Jekaterina/tilt-hs/high_school_economics_lt.json")

Dataset Structure

Data Instances

A typical data point comprises some meta data, including language of the test, the country of origin, file source and name, number of the test's specific question, and a license of the test. The main content contains the question, suggested options for the answer and the correct answer. In addition, some information about the nature of the test is provided, including test level and category in both original language and in English.

An example from the TiLt-HS dataset looks as follows:

{
"language": "lt",
"country": "Lithuania",
"file_name": "Elektroninė-versija-Ekonomikos-ir-finansu-uzduotys-su-atsakymais.pdf",
"source": "https://www.nmakademija.lt/wp-content/uploads/2020/08/",
"license": "no license",
"level": "high school",
"category_en": "microeconomics",
"category_original_lang": "mikroekonomika",
"original_question_num": 1,
"question": "Nuperkamų prekių kiekio priklausomybė nuo kainos vadinama",
"options": [
"pasiūlos funkcija",
"paklausos funkcija",
"rinkos pusiausvyra",
"sąnaudų padengimo tašku"
],
"answer": "2"
}

Data Fields

language: The language of the sample.

country: The country of the sample.

source: The ID or URL of the source.

file_name: The name of the source file.

license: License of the source.

level: The academic level of the tested knowledge. Can be one of the following: e.g. Middle School, High School, University Entrance, University, Professional etc.

category_en: The low level category according to the source IN ENGLISH, i.e. the exam name.

category_original_lang: Similarly to the category_en attribute, the low level category according to the source IN THE ORIGINAL LANGUAGE.

original_question_num: Id of the sample.

question: The text of the multi-choice question or the statement of the True-False question.

options: A list of texts of the multiple choice options. In the case of True-False task, it is left blank.

answer: The string representing the correct choice(s) (1, 2, 3, ...). It corresponds to the numbering of the respective choice columns. In the case of True-False task, this is "1" for true and "0" for false.

Data Splits

The dataset is not split and has only a train subset.

The dataset contains tests for three academic topics, with the following individual number of datapoints/test questions in each:

Train

microeconomics
71

macroeconomics
25

finances and bookkeeping, year 2018
46

Dataset Creation

Curation Rationale

This dataset was collected as a part of the AYA Expedition initiative, for the Global Exams project.

Source Data

The original tests for all the topics were downloaded from the website of the National Academy of Students. Exact link to each test question is provided in the source and file_name fields of the dataset.

Data Collection and Processing

Tests were downloaded during the period of August 25-27, 2024.

Only the questions with multiple-choice answers were selected from the tests. Questions depending on images/figures to be answered, were filtered out. Questions that required reading a paragraph & answering were not included either. Only a subset of available professional tests was included in the current version of the dataset.

Annotations

The dataset does not contain any additional annotations.

Personal and Sensitive Information

No personal, sensitive or private information included in the dataset.

Bias, Risks, and Limitations

This dataset only includes the tests for a certain limited number of academic topics. There is a risk that some of the knowledge presented in the tests changes with time, especially in the area of finance and bookkeeping, so it is important to pay attention to the date of the data collection. This dataset only contains of Lithuanian academic tests and does not necessarily generalize to other languages.

Recommendations

Users should be aware of the risks and limitations of the dataset.

Dataset Card Author

Jekaterina Novikova

Dataset Card Contact

Jekaterina Novikova