Skip to content

typo

Conservative typo correction engine.

The typo object provides Levenshtein-based spelling correction with explicit vocabulary management.

Unlike many text normalization systems, BASA requires users to opt in to typo correction.

Import

from basa import typo

Loading a Vocabulary

Before using typo correction, add words to the vocabulary.

typo.add_to_vocab({
    "makan",
    "minum",
    "berangkat"
})

Correcting Individual Words

typo.correct(
    "mkan"
)

Output:

'makan'

Correcting Sentences

typo.correct_text(
    "saya mkan dulu"
)

Output:

'saya makan dulu'

Custom Maximum Distance

typo.correct(
    "mkan",
    max_dist=2
)

The default behavior is intentionally conservative to avoid overcorrection.

Protected Vocabulary

BASA automatically protects normalized slang outputs from typo correction.

Example:

gk
↓
tidak

The word "tidak" is never treated as a typo candidate, preventing interactions between normalization stages.

Vocabulary Management

Add new words:

typo.add_to_vocab({
    "tensorflow",
    "lightgbm",
    "xgboost"
})

This is particularly useful for:

  • Technical datasets
  • Financial documents
  • Domain-specific NLP pipelines

Design Philosophy

The typo engine follows several principles:

  • Explicit user control
  • Conservative corrections
  • Domain adaptability
  • Predictable behavior
  • Minimal surprises

The long-term roadmap includes:

  • Frequency-based ranking
  • BK-tree acceleration
  • Context-aware corrections
  • Regional language support