Datasets:

joshdavham
/

multilingual-frequency-lists

license: agpl-3.0
language:
  - it
  - fr
  - ja
  - pt
  - es
tags:
  - frequency-list
  - multilingual
  - lemma
  - netflix
pretty_name: Multilingual Frequency Lists
size_categories:
  - 10K<n<100K
configs:
  - config_name: fr
    data_files:
      - split: train
        path: fr/freq_list.tsv
  - config_name: ja
    data_files:
      - split: train
        path: ja/freq_list.tsv
  - config_name: es
    data_files:
      - split: train
        path: es/freq_list.tsv
  - config_name: it
    data_files:
      - split: train
        path: it/freq_list.tsv
  - config_name: pt
    data_files:
      - split: train
        path: pt/freq_list.tsv

Multilingual Frequency Lists

This dataset contains multiple word-frequency lists in various languages such as French, Japanese, Spanish, Italian and Portuguese. Specifically, these are frequency lists of lemmas, meaning, for example, that words like 'run', 'runs' and 'running' are counted together as occurences of the same lemma 'run'.

These frequency lists were generated from ~1GB of subtitles scraped from a variety of Netflix shows and films and parsed using relevant spacy models for each language. Each frequency list contains between 15k and 20k of the top most frequent words.

Quickstart

from datasets import load_dataset
from pprint import pprint

language = "ja" # japanese
dataset = load_dataset("joshdavham/multilingual-frequency-lists", language)

pprint(dataset["train"][:5])
# > {'count': [181501, 97107, 88598, 81836, 70747],
#  'lemma': ['ない', 'です', 'てる', 'ます', 'から'],
#  'rank': [1, 2, 3, 4, 5]}

License

This dataset is licensed under the AGPL.