data_process.py

import random
import iso639

import language_names
import language_paraphrase
import language_translate

import pandas as pd
import datasets
from dataclasses import dataclass

import json
import uuid

class DataProcess:

    class RandomText:

        # list of random quotes
        random_quote = {
            1: '\'',
            2: '\"',
            3: '“',
            4: '῎',
            5: '`',
        }

        # provide instruction with a text; process of randomization of a text
        @staticmethod
        def randomize_text(text, original_lang=None, target_lang=None):
            templates = language_translate.random_templates_translate.get(original_lang, {}) if not ((original_lang == target_lang) and (original_lang is not None) and (target_lang is not None)) else language_paraphrase.random_templates_paraphrase.get(original_lang, {})
            template = random.choice(list(templates.values()))
            quote = random.choice(list(DataProcess.RandomText().random_quote.values()))
            original_lang_name = DataProcess.language_name(None, original_lang, original_lang)
            target_lang_name = DataProcess.language_name(None, target_lang, original_lang)
            return template.format(text=text, lang1=target_lang_name, lang2=original_lang_name, quote=quote)

    # convert to iso639_1 
    def convert_code(self, code):
        try:
            mapped_code = iso639.to_iso639_1(code)
        except:
            mapped_code = None
        return mapped_code

    # return language #1 name in language #2
    def language_name(self, lang1, lang2):
        name = language_names.language_names.get(lang1, {}).get(lang2)
        if name is not None:
            return name
        # just in case
        elif lang1 == lang2:
            iso_name = iso639.to_native(lang1)
            return(iso_name)
        else:
            return None

converter = DataProcess()

"""
EXAMPLES:

# get language name; iso639_1 code
print(converter.language_name('ru', 'en')) # Output: Russian
print(converter.convert_code("eng")) # Output: en

# convert into INSTRUCTION format: text; to; from
text = "test"
print(converter.RandomText.randomize_text(text, "uk", "fr")) # Ти можеш перекласти цей вислів: 'test'?
print(converter.RandomText.randomize_text(text, "uk", "de")) # Переклади наступний текст "test" з мови "німецька мова"
"""

@dataclass
class QnA:
    INSTRUCTION: str
    RESPONSE: str
    SOURCE: str
    METADATA: str

# format to QnA
def create_qna(row):
    # get rows; create uuid based on texts
    text = row['Text']
    translation = row['Translated text']
    lang_from = converter.convert_code(row['Original lang'])
    lang_to = converter.convert_code(row['Target lang'])
    uuid_val = uuid.uuid3(uuid.NAMESPACE_OID, str(text + translation))
    # json with language, uuid and langs-pair
    METADATA = {"language": f"{lang_to}", "uuid": f"{uuid_val}", "langs-pair": f"{lang_from}-{lang_to}"}
    metadata_str = json.dumps(METADATA)
    SOURCE = "tatoeba"
    # randomizing INSTRUCTION
    INSTRUCTION = converter.RandomText.randomize_text(text, lang_to, lang_from)
    RESPONSE = translation
    return QnA(INSTRUCTION, RESPONSE, SOURCE, metadata_str)

# load the dataset from Hugging Face
hf_dataset = datasets.load_dataset('0x22almostEvil/tatoeba-mt-llama-only', split='train')

# original is ~3M; with num_shards=55 it'll be ~65K
hf_dataset = hf_dataset.shard(num_shards=55, index=0)

print(hf_dataset)

# convert the dataset to a pandas dataframe
df = pd.DataFrame(hf_dataset)

# apply the create_qna function to each row of the dataframe to create QnA objects
qna_list = df.apply(create_qna, axis=1).tolist()

# save the QnA objects as a parquet file
qna_df = pd.DataFrame(qna_list, columns=["INSTRUCTION", "RESPONSE", "SOURCE", "METADATA"])
qna_df.to_parquet("translation-taboeba-qna-65k-oa.parquet", row_group_size=100, engine="pyarrow", index=False)