leksara 0.0.8

Library pemrosesan teks Bahasa Indonesia untuk domain e-commerce (cleaning, PII masking, review mining, pipeline).

Leksara

Description

Leksara is a Python toolkit designed to streamline the preprocessing and cleaning of Indonesian text data for Data Scientists and Machine Learning Engineers. It focuses on handling messy and noisy Indonesian text from various domains such as e-commerce reviews, social media posts, and chat conversations. The tool helps clean text by handling Indonesian-specific challenges like slang words, regional expressions, informal abbreviations, and mixed language content, while also providing standard cleaning features like punctuation and stopword removal. This makes it an essential tool for Indonesian text analysis and machine learning model preparation.

Key Features

• Basic Cleaning Pipeline: A straightforward pipeline to clean raw text data by handling common tasks like punctuation removal, casing normalization, and stopword filtering.
• Advanced Customization: Users can create custom cleaning pipelines tailored to specific datasets, including support for regex pattern matching, stemming, and custom dictionaries.
• Preset Options: Includes predefined cleaning presets for various domains like e-commerce, allowing for one-click cleaning.
• Slang and Informal Text Handling: Users can define their own custom dictionaries for slang terms and informal language, especially useful for Indonesian text.

Usage Examples

Basic Usage: Basic Cleaning Pipeline

This example demonstrates how to clean e-commerce product reviews using a pre-built preset.

from Leksara  import Leksara 

df['cleaned_review'] = Leksara(df['review_text'], preset='ecommerce_review')
print(df[['review_id', 'cleaned_review']])

Input Data (df):

review_id	review_text
1	<p>brgnya ORI & pengiriman cepat. Mantulll 👍</p>
2	Kualitasnya krg bgs, ga sesuai ekspektasi...

Output Data:

review_id	cleaned_review
1	barang nya original pengiriman cepat mantap
2	kualitasnya kurang bagus tidak sesuai ekspektasi

Advanced Usage: Custom Cleaning Pipeline

Customize the pipeline to mask phone numbers and normalize whitespace in chat logs.

from Leksara import Leksara
from Leksara.functions import to_lowercase, normalize_whitespace
from Leksara.patterns import MASK_PHONE_NUMBER

custom_pipeline = {
    'patterns': [MASK_PHONE_NUMBER],
    'functions': [to_lowercase, normalize_whitespace]
}

df['safe_message'] = Leksara(df['chat_message'], pipeline=custom_pipeline)
print(df[['chat_id', 'safe_message']])

Input Data (df):

chat_id	chat_message
101	Hi kak, pesanan saya INV/123 blm sampai. No HP saya 081234567890
102	Tolong dibantu ya sis, thanks

Output Data:

chat_id	safe_message
101	hi kak, pesanan saya inv/123 blm sampai. no hp saya [PHONE_NUMBER]
102	tolong dibantu ya sis, thanks

Goals & Objectives

• Provide an intuitive and adaptable cleaning tool for Indonesian text, focusing on domains like e-commerce.
• Enable Data Scientists and ML Engineers to clean and preprocess text with minimal effort.
• Allow for deep customization through configuration options and the use of custom dictionaries.

Success Metrics

• On-time Delivery: Targeted release by October 15, 2025.
• Processing Speed: Clean a 10,000-row Pandas Series in under 5 seconds.
• Cleaning Accuracy: Achieve over 95% accuracy for core cleaning functions.

Folder Structure

Below is the recommended folder structure for organizing the project:

[Leksara]/
├── pyproject.toml                  # packaging & deps (nltk, dll)
├── requirements.txt                # runtime deps (nltk, pandas, dll)
├── README.md                       # overview & usage
├── leksara/                        # package utama
│   ├── __init__.py                 # public API surface
│   ├── version.py                  # versi paket
│   ├── core/
│   │   ├── chain.py                # pipeline/CLI entry (sesuai pyproject scripts)
│   │   ├── logging.py              # util logging/benchmark
│   │   └── presets.py              # preset pipeline
│   ├── frames/
│   │   └── cartboard.py            # helpers untuk data frame
│   ├── functions/                  # modul granular
│   │   ├── __init__.py
│   │   ├── cleaner/
│   │   │   ├── __init__.py
│   │   │   └── basic.py            # remove_tags, case_normal, remove_stopwords, dll.
│   │   ├── patterns/
│   │   │   ├── __init__.py
│   │   │   └── pii.py              # masker PII (email/telepon, dll.)
│   │   └── review/
│   │       ├── __init__.py
│   │       └── advanced.py         # fungsi review lanjutan
│   ├── resources/                  # data pendukung (dibundel)
│   │   ├── acronyms.csv
│   │   ├── contractions.json
│   │   ├── slang_dict.json
│   │   └── stopwords/
│   │       └── id.txt              # stopwords Indonesia (tambahan/abbr)
│   ├── tests/
│   │   ├── test_chain.py
│   │   ├── test_cleaner_basic.py
│   │   ├── test_patterns_pii.py
│   │   └── test_review_advanced.py
│   └── utils/
│       ├── lang.py
│       ├── regexes.py
│       ├── text.py                 # text helpers
│       └── whitelist.py
└── notebooks/
    └── leksara_quickstart.ipynb    # quickstart & demo

Milestones

Sprint	Dates	Goal
1	Aug 18 – Aug 22	Project Kickoff, Discovery, Set up repository
2	Aug 22 – Aug 29	Build Core Cleaning Engine
3	Aug 29 – Sep 5	Develop Configurable Features
4	Sep 5 – Sep 12	Implement Advanced Customization
5	Sep 12 – Sep 19	Refine API
6	Sep 19 – Sep 26	Optimize System
7	Sep 26 – Oct 3	Finalize Documentation
8	Oct 3 – Oct 10	Prepare for Launch

Requirements

• Python 3.x
• Pandas

Install

pip install Leksara

Contributors

• Vivian & Zahra – Document Owners
• Salsa – UI/UX Designer
• Aufi, Althaf, Rhendy, Adit – Data Science Team
• Alya, Vivin – Data Analyst Team

For more details on the features and usage, refer to the official documentation linked above.

Links

• UI Design
• Product Design and Mockups