bm-eval-metrics 1.5.5

Python package providing easy-to-use evaluation metrics and utilities for Machine Learning

bm-eval-metrics

bm-eval-metrics is a Python package providing easy-to-use evaluation metrics and utilities for machine learning workflows.

Features

• Text cleaning and normalization
• Tokenization and stopword removal
• Lemmatization
• TF-IDF and Bag-of-Words vectorization
• Pipeline-based preprocessing
• Built on NLTK and pandas
• Scikit-learn style API

Installation

Install from PyPI:

pip install bm-eval-metrics

Quick Start

Basic Usage With Pipeline

from bm_eval_metrics import (
    TextCleaner,
    Tokenizer,
    Normalizer,
    StopwordFilter,
    Lemmatizer,
    Vectorizer,
    Pipeline,
)

# Sample documents
documents = [
    "This is an example document! It has punctuation and numbers: 123.",
    "Natural Language Processing is AMAZING!!!",
    "Preprocessing text is very important for NLP tasks.",
]

# Create preprocessing components
cleaner = TextCleaner(
    lowercase=True,
    remove_punctuation=True,
    remove_numbers=True,
    strip_whitespace=True,
)

tokenizer = Tokenizer(method="word")

normalizer = Normalizer(
    expand_contractions=True,
    fix_unicode=True,
)

stopword_filter = StopwordFilter(language="english")
lemmatizer = Lemmatizer(method="wordnet")

vectorizer = Vectorizer(
    method="tfidf",
    max_features=5000,
    ngram_range=(1, 2),
)

# Build pipeline
preprocessing_pipeline = Pipeline(
    [
        cleaner,
        normalizer,
        tokenizer,
        stopword_filter,
        lemmatizer,
        vectorizer,
    ]
)

# Run preprocessing
processed_data = preprocessing_pipeline.fit_transform(documents)

# Inspect output
print("Processed features shape:", processed_data.shape)
print("Sample vector:", processed_data[0])

Step-by-Step Processing Without Pipeline

from bm_eval_metrics import (
    TextCleaner,
    Tokenizer,
    StopwordFilter,
    Lemmatizer,
    Vectorizer,
)

docs = [
    "Machine learning is fun!",
    "Text preprocessing improves results.",
]

# Initialize tools
cleaner = TextCleaner(lowercase=True)
tokenizer = Tokenizer()
stopwords = StopwordFilter("english")
lemmatizer = Lemmatizer()
vectorizer = Vectorizer(method="bow")

# Process
cleaned = [cleaner.clean(d) for d in docs]
tokens = [tokenizer.tokenize(d) for d in cleaned]
filtered = [stopwords.remove(t) for t in tokens]
lemmatized = [lemmatizer.lemmatize(t) for t in filtered]

vectors = vectorizer.fit_transform(lemmatized)
print(vectors)

Components Overview

Component	Description
TextCleaner	Removes noise and formats text
Tokenizer	Splits text into tokens
Normalizer	Standardizes text
StopwordFilter	Removes common filler words
Lemmatizer	Converts words to base form
Vectorizer	Converts text to numeric features
Pipeline	Chains components into a workflow

Deep Learning Preparation Example

from bm_eval_metrics import (
    TextCleaner,
    Tokenizer,
    SequencePadder,
    VocabularyBuilder,
)

texts = [
    "Deep learning for NLP",
    "Transformers are powerful",
]

cleaner = TextCleaner(lowercase=True)
tokenizer = Tokenizer()
vocab = VocabularyBuilder(max_size=10000)
padder = SequencePadder(max_length=50)

# Clean
cleaned = [cleaner.clean(t) for t in texts]

# Tokenize
tokens = [tokenizer.tokenize(t) for t in cleaned]

# Build vocabulary
vocab.fit(tokens)

# Encode
encoded = [vocab.encode(t) for t in tokens]

# Pad
padded = padder.pad(encoded)

print(padded)

Requirements

• Python 3.11+
• nltk
• pandas
• scikit-learn

Install dependencies automatically with:

pip install bm-eval-metrics

Project Structure

bm-eval-metrics/
├── cleaning.py
├── tokenization.py
├── normalization.py
├── filtering.py
├── lemmatization.py
├── vectorization.py
├── pipeline.py
└── __init__.py

Contributing

Contributions are welcome.

1. Fork the repository.
2. Create a new branch.
3. Commit your changes.
4. Open a pull request.

License

This project is licensed under the MIT License.

Support

If you encounter issues or have feature requests, open an issue on GitHub.