This repo uses Uppsala Persian Corpus (UPC) to construct a lexicon of 70528 unique words. With all the excitement around game Wordle, we also extracted words with different length (2, 3, 4, ..., 10) and stored them to separate files for easier access. Please note that these files might contain offensive words, I have not check them manually.

GetWords.py can read these files and return words as a list of strings.

The main lexicon (data/persian-words.txt) is build very liberally; we only filter out words that contain ASCII characters or Arabic numerals.

More conservative filtering has been applied to files with fixed word length. We drop all words that contain any of the following characters:

• All forms of hamze (همزه).
• All forms of tanvin (تنوین).
• All forms of short vowels.
• Tashdid (تشدید).

Also when calculating length of words we do not take into account Zero-width non-joiner (نیم‌فاصله) characters. For example, "پیش‌بینی‌" length is consider to be 7 rather than 8.

After applying these filters, we ended up with these number of words per file:

• 2 letter words: 316 unique words
• 3 letter words: 2389 unique words
• 4 letter words: 7236 unique words
• 5 letter words: 11024 unique words
• 6 letter words: 12507 unique words
• 7 letter words: 11753 unique words
• 8 letter words: 9335 unique words
• 9 letter words: 6113 unique words
• 10 letter words: 3648 unique words

Uppsala Persian Corpus (UPC) is annotated with the following 31 part-of-speech tags:

You can choose to include only words with certain tags in the output lexicon files. For example, the following command:

python3 Cleanup.py N_PL N_SING  

will generate the output files that only include singular and plural nouns.