v2.2.0 • Bayes + LSI + kNN + TF-IDF + Logistic Regression

Text classification for Ruby made simple

Bayesian, LSI, kNN, TF-IDF, and Logistic Regression algorithms for categorizing documents, detecting spam, analyzing sentiment, and building semantic search.

Get Started View on GitHub

Bayesian Classification

Train categories with examples, classify new text with probability scores.

Semantic Search

Find related documents using Latent Semantic Indexing and SVD.

k-Nearest Neighbors

Instance-based classification with interpretable neighbor results.

TF-IDF Vectorizer

Transform text into weighted feature vectors for ML and search.

Logistic Regression

Well-calibrated probabilities with interpretable feature weights.

Streaming Training

Train on multi-GB datasets with memory-efficient streaming and progress tracking.

See it in action

Train a Classifier

 require 'classifier'

classifier = Classifier::Bayes.new 'Spam', 'Ham'

# Train with keyword arguments
classifier.train(spam: "Buy cheap viagra now!!!")
classifier.train(ham: "Meeting tomorrow at 3pm")

# Or batch train multiple examples
classifier.train(
  spam: ["You won $1M!", "Free money!"],
  ham: ["Project update", "Lunch tomorrow?"]
)

Create categories and train with example text

Classify Text

 classifier.classify "Claim your free prize now"
# => "Spam"

classifier.classify "Quarterly review scheduled"
# => "Ham"

# Get log-probability scores (less negative = more likely)
classifier.classifications "Limited time offer"
# => {"Spam" => -10.5, "Ham" => -10.2}

Get the best category for new text

Semantic Search with LSI

 lsi = Classifier::LSI.new

# Add documents with categories
lsi.add("Ruby" => "Ruby programming language")
lsi.add("Python" => "Python snake reptile")
lsi.add("Ruby" => "Rails web framework")

lsi.search "web development with Ruby"
# => ["Rails web framework", "Ruby programming..."]

Find documents by meaning, not just keywords

k-Nearest Neighbors

 knn = Classifier::KNN.new(k: 3)

knn.add(
  spam: ["Free prize winner", "You won free money"],
  ham: ["Meeting at 3pm", "Review doc"]
)

result = knn.classify_with_neighbors "Claim your free prize"
result[:category]    # => "spam"
result[:confidence]  # => 0.67

Instance-based classification with interpretable results

TF-IDF Vectorizer

 tfidf = Classifier::TFIDF.new

tfidf.fit([
  "Ruby programming language",
  "Python programming language",
  "Dogs are great pets"
])

vector = tfidf.transform "Ruby programming"
# => {:rubi => 0.80, :program => 0.61}

Transform text into weighted feature vectors

Logistic Regression

 lr = Classifier::LogisticRegression.new([:spam, :ham])

lr.train(
  spam: ["Buy now!", "Free money!"],
  ham: ["Meeting at 3pm", "Project update"]
)

# Probabilities always sum to 1.0
lr.probabilities "Claim your prize"
# => {"spam" => 0.37, "ham" => 0.63}

Well-calibrated probabilities for confident decisions

Start in seconds

 $ gem install classifier

Or add to your Gemfile: gem 'classifier'

Tutorials

Step-by-step guides to build real projects with the classifier.

Guides

In-depth reference for Bayes, LSI, kNN, TF-IDF, Logistic Regression, and more.

Contribute

Join 50+ contributors on GitHub. Issues and PRs welcome!