Emoji Extractor

Extract, detect, and count emoji from text — fast and accurate. Fully supports multi-codepoint sequences (skin tones, ZWJ sequences, flags).

Uses a trie-based greedy longest-match engine (pure Python, zero dependencies) that is 27× faster than regex for single strings and 115× faster when processing large datasets with automatic multiprocessing.

Installation

pip install emoji_extractor

Quick Start

Use the top-level convenience functions for simple tasks:

from emoji_extractor import count_emoji, detect_emoji

# Check if a string contains emoji
detect_emoji("Hello 👋")   # True
detect_emoji("No emoji")   # False

# Count emoji in a single string — returns a Counter
counts = count_emoji("I love 🍎 and 🍌🍌")
print(counts)
# Counter({'🍌': 2, '🍎': 1})

Single Strings vs Bulk Processing

The package provides two tiers of counting methods:

count_emoji(string) — Single string

Scans one string and returns a Counter. Fast enough for real-time use (~9µs per line).

from emoji_extractor import Extractor

ext = Extractor()
ext.count_emoji("Great job 🎉🎉🎉")
# Counter({'🎉': 3})

count_all_emoji(iterable) — Bulk processing

Processes a list (or any iterable) of strings. For inputs with 1,000+ lines, work is automatically distributed across multiple CPU cores for significantly faster throughput.

tweets = ["Love this 🍎", "So funny 😂😂", "Hello world", ...]

# Automatically parallelised for large inputs
totals = ext.count_all_emoji(tweets)
print(totals.most_common(5))
# [('😂', 2813), ('❤', 1150), ('😍', 974), ...]

Advanced Usage

Version Selection

By default, the package uses the latest Unicode Emoji data (currently 17.0). To extract emoji as defined in a specific historical version:

from emoji_extractor import Extractor

ext_14 = Extractor(version='14.0')
ext_15 = Extractor(version='15.0')

# 🩷 Pink heart was introduced in 15.0
ext_14.detect_emoji("🩷")  # False
ext_15.detect_emoji("🩷")  # True

Available versions: 4.0, 5.0, 11.0, 12.0, 12.1, 13.0, 14.0, 15.0, 15.1, 16.0, 17.0, 18.0.

Tone-Modifiable Emoji

Count emoji that support skin tone modifiers, plus their unmodified base forms:

ext = Extractor()
ext.count_tme("High five ✋🏽")
# Counter({'✋🏽': 1})

ext.count_tones("Waves 👋🏻👋🏿")
# Counter({'🏻': 1, '🏿': 1})

Controlling Parallelism

# Use fewer workers (default: min(cpu_count, 8))
ext = Extractor(n_workers=4)

# Disable multiprocessing entirely
ext = Extractor(n_workers=1)

# Clean up worker processes when done
ext.close()

Details & Features

• Accurate Counting: Uses a greedy longest-match trie to correctly handle multi-codepoint emoji, including ZWJ sequences like 👩‍🦰 and flag sequences like 🇬🇧.
• Fast: 27× faster than regex for single strings. 115× faster with parallelism for bulk data.
• Zero Dependencies: Pure Python — no external packages required.
• Historical Accuracy: Supports strict adherence to older Unicode specifications, avoiding false positives on newer emoji.
• Always Up to Date: Automatically checks for new Unicode releases via GitHub Actions and updates itself.

How It Works Under the Hood

The package relies on official Unicode data parsed from emoji-test.txt. For each supported version, the data/ folder contains:

• emoji_sequences.json: All emoji strings, sorted longest-first. Used to build a nested-dict trie for greedy matching.
• tme_sequences.json: Tone-modifiable emoji sequences.
• possible_emoji.json: A set of all characters that could be part of an emoji (used by detect_emoji() for fast presence checking).

The trie scanner walks through text character by character, always matching the longest possible emoji sequence at each position. This naturally handles cases where a shorter emoji is a prefix of a longer one (e.g., 👩 vs 👩‍🦰).

Note: Some emoji include a variation selector (U+FE0F), but some platforms strip it while still rendering the emoji. The trie captures both forms.

Changelog

17.0.2

• Engine: Regex replaced with pure-Python trie (27× faster single, 115× bulk with multiprocessing)
• Data: big_regex.txt / tme_regex.txt → emoji_sequences.json / tme_sequences.json
• check_first parameter is now a no-op (accepted for compatibility)
• count_all_* methods auto-parallelise for large inputs
• Added n_workers parameter and close() method to Extractor
• Removed Extractor.big_regex and Extractor.tme (raise helpful error if accessed)

Other Work

If you want to do more than detecting, extracting, and counting emoji, this Python package may be useful.

Contact

Feel free to email me about any of this stuff.