maliknaik/natural_unit_conversion

Natural Unit Conversion Dataset

Dataset Overview

This dataset contains unit conversion requests, where each example includes a sentence with associated entities (in spaCy-supported format) for Named-Entity Recognition (NER) modeling. The entities represent the values and units being converted. The goal is to aid in developing systems capable of extracting unit conversion data from natural language for natural language understanding.

The data is structured with the following fields:

• text: The sentence containing the unit conversion request natural text.
• entities: A list of entity annotations. Each entity includes:
  • The start and end (exclusive) positions of the entity in the text.
  • The entity type (e.g., UNIT_VALUE, FROM_UNIT, TO_UNIT, FEET_VALUE, INCH_VALUE).

Examples

[
  {
    "text": "I'd like to know 8284 atm converted to pascals", 
    "entities": [
      [39, 46, "TO_UNIT"], 
      [17, 21, "UNIT_VALUE"], 
      [22, 25, "FROM_UNIT"]
    ]
  }, 
  {
    "text": "Convert 3'6\" to fts", 
    "entities": [
      [8, 9, "FEET_VALUE"], 
      [10, 11, "INCH_VALUE"], 
      [16, 19, "TO_UNIT"]
    ]
  },
  {
    "text": "Convert this: 4487 n-m to poundal meters",
    "entities": [
      [26, 40, "TO_UNIT"], 
      [14, 18, "UNIT_VALUE"], 
      [19, 22, "FROM_UNIT"]
    ]
  },
]

Loading Dataset

You can load the dataset maliknaik/natural_unit_conversion using the following code:

from datasets import load_dataset

dataset_name = "maliknaik/natural_unit_conversion"
dataset = load_dataset(dataset_name)

Entity Types

• UNIT_VALUE: Represents the value to be converted.
• FROM_UNIT: The unit from which the conversion is being made.
• TO_UNIT: The unit to which the conversion is being made.
• FEET_VALUE: The value representing feet in a length conversion.
• INCH_VALUE: The value representing inches in a length conversion.

Dataset Split and Sampling

The dataset is split using Stratified Sampling to ensure that the distribution of entities is consistent across training, validation, and test splits. This method helps to ensure that each split contains representative examples of all entity types.

• Training Set: 583,863 samples
• Validation Set: 100,091 samples
• Test Set: 150,137 samples

Supported Units

The dataset supports a variety of units for conversion, including but not limited to:

Length: inches, feet, yards, meters, centimeters, millimeters, micrometers, kilometers, miles, mils

Area: square meters, square inches, square feet, square yard, square centimeters, square miles, square kilometers, square feet us, square millimeters, hectares, acres, are

Mass: ounces, kilograms, pounds, tons, grams, ettograms, centigrams, milligrams, carats, quintals, pennyweights, troy ounces, uma, stones,, micrograms

Volume: cubic meters, liters, us gallons, imperial gallons, us pints, imperial pints, us quarts, deciliters, centiliters, milliliters, microliters, tablespoons us, australian tablespoons, cups, cubic millimeters, cubic centimeters, cubic inches, cubic feet, us fluid ounces, imperial fluid ounces, us gill, imperial gill

Temperature: celsius, fahrenheit, kelvin, reamur, romer, delisle, rankine

Time: seconds, minutes, hours, days, weeks, lustrum, decades, centuries, millennium, deciseconds, centiseconds, milliseconds, microseconds, nanoseconds

Speed: kilometers per hour, miles per hour, meters per second, feets per second, knots, minutes per kilometer

Pressure: atmosphere, bar, millibar, psi, pascal, kilo pascal, torr, inch of mercury, hecto pascal

Force: newton, kilogram force, pound force, dyne, poundal

Energy: kilowatt hours, kilocalories, calories, joules, kilojoules, electronvolts, energy foot pound

Power: kilowatt, european horse power, imperial horse power, watt, megawatt, gigawatt, milliwatt

Torque: newton meter, kilogram force meter, dyne meter, pound force feet, poundal meter

Angle: degree, radians

Digital: byte, bit, nibble, kilobyte, gigabyte, terabyte, petabyte, exabyte, tebibit, exbibit, etc

Fuel_efficiency: kilometers per liter, liters per100km, miles per us gallon, miles per imperial gallon

Shoe_size: eu china, usa canada child, usa canada man, usa canada woman, uk india child, uk india man, uk india woman, japan

Usage

This dataset can be used for training named entity recognition (NER) models, especially for tasks related to unit conversion and natural language understanding.

License

This dataset is available under the CC0-1.0 license. It is free to use for any purpose without any restrictions.

Citation

If you use this dataset in your work, please cite it as follows:

@misc{unit-conversion-dataset,
  author = {Malik N. Mohammed},
  title = {Natural Language Unit Conversion Dataset for Named-Entity Recognition},
  year = {2025},
  publisher = {HuggingFace},
  journal = {HuggingFace repository}
  howpublished = {\url{https://huggingface.co/datasets/maliknaik/natural_unit_conversion}}
}