ykankaya / ptbr-wordlist

A selection of Brazilian Portuguese words that are good enough to be used as a BIP-0039 wordlist

Brazilian Portuguese Wordlist

Objectives

The idea is to build a selection of Brazilian Portuguese words that are good enough to be used on computer-generated mnemonic pasphrases, such as the ones described in the BIP-0039 draft and adopted by some cryptocurrencies clients and wallets.

Attributes of a "good enough" wordlist

Following the footsteps of other languages wordlists currently listed in this BIP39 document, specially the French and Spanish ones, the selection of words for this list must consider the following features/attributes:

• High priority on simple and common portuguese words.
• Only words with 5-8 letters.
• Words can be uniquely determined typing the first 4 characters (sometimes less).
• No identical words with the Spanish, English or French wordlists

In addition to that, some other attributes like:

• No accented words (we avoid words containing á,é,ê,ç… to keep it simple)
• High priority to words that could be suggested by a mobile OS (words with a high frequency score on the Android dictionary, for example)

This list of attributes is still a work in progress, feedback is welcome (see the Discussion section below).

Discussion

There is an ongoing discussion about it at the bitcointalk forum, please drop by and say hello if you want to contribute :)

Build

To build all lists use:

$ npm test

Development Scripts

The package.json file have some npm-scripts that are shortcuts for shell command lines, if you have node.js installed you can run then with:

$ npm run <script_name>

The script names are listed bellow with their equivalent command lines:

getdict:bip39

Download the bip39 recommended wordlists for Spanish, French and English

Equivalent shell:

$ cd ./downloads
$ curl -O https://raw.githubusercontent.com/fczuardi/bips/master/bip-0039/english.txt
$ curl -O https://raw.githubusercontent.com/fczuardi/bips/master/bip-0039/spanish.txt 
$ curl -O https://raw.githubusercontent.com/fczuardi/bips/master/bip-0039/french.txt

getdict:android

Download the pt_BR dictionary csv from Google's Android repository

Equivalent shell:

$ mkdir -p ./downloads
$ cd ./downloads
$ curl -O https://android.googlesource.com/platform/packages/inputmethods/LatinIME/+archive/master/dictionaries.tar.gz
$ tar -zxvf dictionaries.tar.gz pt_BR* && gunzip pt_BR*.gz

csv

Write a file named lists/all.csv that is the downloaded Android dictionary filtered to display only:

• words without any of the following letters: áéíóúâêôãõç
• that has 5-8 letters
• without any flags (flags=,), so, no adult or offensive words
• sorted by the 4th column (originalFreq)

Equivalent shell:

$ cat downloads/pt_BR_wordlist.combined | \
grep -E 'word=[^áéíóúâêôãõç,]{5,8},' | \
sed -e 's:originalFreq=::' -e 's: word=::' -e 's:f=.*,::'| \
sort -t, -k2 -rn >lists/all.csv

filter:en

Write a file named lists/all-no_en.csv that is lists/all.csv without the lines containing words of downloads/english.txt

Equivalent shell:

$ cat downloads/english.txt|grep '^a'| \
sed -e 's:^\(.*\):\^\1,:' > /tmp/all_en_pattern
$ grep -vf /tmp/all_en_pattern lists/all.csv > lists/all-no_en.csv

filter:sp

Write a file named lists/all-no_en-no_sp.csv that is lists/a-no_en.csv without the lines containing words of downloads/spanish.txt

Equivalent shell:

$ cat downloads/spanish.txt|grep '^a'| \
sed -e 's:^\\(.*\\):\\^\\1,:' > /tmp/all_sp_pattern 
$ grep -vf /tmp/all_sp_pattern lists/all-no_en.csv > lists/all-no_en-no_sp.csv

filter:fr

Write a file named lists/all-no_en-no_sp-no_fr.csv that is lists/all-no_en-no_sp.csv without the lines containing words of downloads/french.txt

Equivalent shell:

$ cat downloads/french.txt|grep '^a'| \
sed -e 's:^\\(.*\\):\\^\\1,:' > /tmp/all_fr_pattern 
$ grep -vf /tmp/all_fr_pattern lists/all-no_en-no_sp.csv > lists/all-no_en-no_sp-no_fr.csv

crop10k

Write a file named lists/filtered-top-10k.csv that is the top 10,000 words in frequency from lists/all-no_en-no_sp-no_fr.csv

Equivalent shell:

$ cat lists/all-no_en-no_sp-no_fr.csv| \
head -n 10000 > lists/filtered-top-10k.csv

a-z

Writes two separated csv files for each letter from a to z, each file contains the words from that letter one sorted alphabetically (alpha-a.csv, alpha-b.csv…) and the other sorted by frequency (score-a.csv, score-b.csv…).

Equivalent shell:

$ for x in {a..z}; \
do grep "^$x" lists/filtered-top-10k.csv > lists/score-$x.csv &&\
cat lists/score-$x.csv | sort > lists/alpha-$x.csv; 
done