In antdurrant/word.lists: A Collection Of Word Lists For ESL

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  fig.path = "man/figures/README-",
  out.width = "100%"
  )

options(knitr.kable.NA = '')

word.lists

The goal of word.lists is to provide easy access to a handful of word-frequency lists.

These lists come primarily from the researchers Charlie Browne, Brent Culligan and Joseph Phillips, and made publicly available at https://www.newgeneralservicelist.org/ and Tom Cobb, whose site https://www.lextutor.ca/ is immensely useful for vocabulary profiling.

Installation

You can install the released version of word.lists from CRAN with:

# no, you can't
# install.packages("word.lists")

And the development version from GitHub with:

# install.packages("devtools")
# just this one for at least a while
devtools::install_github("antdurrant/word.lists")

We'll see the distribution of words from the NGSL and NAWL in some arbitrary academic-ish text.

NAWL List Description

A dataset containing the New Academic Word List (NAWL) & New General Service List (NGSL). Difficulty groupings have been arbitrarily set by me, as follows: Group 1: first 500 words of NGSL by frequency & "supplementary" words - months/numbers etc Group 2: next 500 words of NGSL by frequency Group 3: next 1000 words of NGSL by frequency Group 4: remaining NGSL words by frequency (about 800 words) Group 5: academic word list (about 950 words)

library(word.lists)
library(udpipe)
library(dplyr)

What does a list look like?

list_academic %>%
  head() %>% 
  knitr::kable()

Any old text will do:

text <- "There are numerous indicators of success in any field of research. 
Though one may be valid in some context, it may be rendered without utility in another."

Annotate it with udpipe, keep only some useful bits

piped_text <- udpipe(text, object = "english") %>% 
  select(doc_id, sentence_id, token_id, token, lemma, upos) 

piped_text %>%
  head() %>%
  knitr::kable()

Show the words and where they are in the list

piped_text %>%
  left_join(list_academic) %>%
  head() %>% 
  knitr::kable()

Order by most advanced words first

joined_text <- piped_text %>%
  left_join(list_academic) %>%
  filter(upos != "PUNCT") %>%
  arrange(desc(group)) 

joined_text %>%
  head() %>%
  knitr::kable()

How many in each group?

joined_text %>%
  count(on_list, group) %>%
  arrange(desc(group))%>%
  knitr::kable()

We can see that this has a fairly high proportion of "academic" words with the majority falling under the most frequent thousand-word grouping, which is totally normal.

Auto-generate translated wordlists

If I am teaching Japanese-speaking kids:

I want to get a wordlist

piped_text %>%
  word.lists::get_wordlist(language = "jpn")%>%
  knitr::kable()

Okay, but I know my students know the first thousand or so words

piped_text %>%
  word.lists::get_wordlist(language = "jpn") %>%
  left_join(list_academic) %>%
  filter(upos != "PUNCT",
         group > 2) %>%
  select(on_list, group, token, lemma, upos, translation) %>%
  knitr::kable()

Or maybe I am teaching Finnish-speaking kids who just need academic words:

piped_text %>%
  word.lists::get_wordlist(language = "eus") %>%
  left_join(list_academic) %>%
  filter(upos != "PUNCT",
         group == 5) %>%
  select(on_list, group, token, lemma, upos, translation) %>%
  knitr::kable()

The translations only go English > OTHER. They all come from the Open Multilingual Wordnet, which itself is an international project using the fundamental work done by the Princeton wordnet. Where there is no entry in the Open Multilingual Wordnet, the translation will come out blank. It is not perfect, but it is a heck of a lot easier than going through texts one word at a time.

The translation work runs through Python's nltk module.

word.lists::nltk_languages %>%
  knitr::kable()

TODO: Add IPA

TODO: Bring this functionality to an app

antdurrant/word.lists documentation built on July 20, 2023, 3:57 p.m.