DKSplit

⚠️ Security Notice: The only official repositories for this project are ABTdomain/dksplit (Python) and ABTdomain/dksplit-go (Go). We are aware of unauthorized clones that may distribute suspicious files. Please only download from our official repositories.

v0.2.2 — Retrained model with expanded brand and name coverage. ~7% accuracy improvement on real-world domains. API unchanged — just pip install --upgrade dksplit.

String segmentation using BiLSTM-CRF. Splits concatenated words into meaningful parts.

DKSplit is a lightweight model trained on millions of labeled samples covering domain names, brand names, tech terms, and multilingual phrases. It uses a BiLSTM-CRF architecture (9.47M parameters) exported to ONNX with INT8 quantization, delivering fast CPU inference in a 9 MB package.

Originally built for domain name analysis at DomainKits, but works well on any concatenated text — hashtags, URLs, identifiers, compound strings.

Install

pip install dksplit

Usage

import dksplit

dksplit.split("chatgptlogin")
# ['chatgpt', 'login']

dksplit.split("kubernetescluster")
# ['kubernetes', 'cluster']

dksplit.split("mercibeaucoup")
# ['merci', 'beaucoup']

dksplit.split_batch(["openaikey", "microsoftoffice", "bitcoinprice"])
# [['openai', 'key'], ['microsoft', 'office'], ['bitcoin', 'price']]

What's New in v0.2.2

Retrained model with significantly expanded brand and name coverage. The API is unchanged — just upgrade.

pip install --upgrade dksplit

Examples of improvements:

Input	v0.1.0	v0.2.2
cloudflarecdn	cloud flare cdn	cloudflare cdn
snowdenyes	snow den yes	snowden yes
robertdeniro	robert deniro	robert de niro

Benchmark

Dataset

1,000 newly registered .com domains randomly sampled from ABTdomain.com daily feed (February 8, 2026). No filtering or cherry-picking — a raw random sample of real-world domain registrations. GPT-5.2 segmentation is used as the reference answer.

The dataset and evaluation script are included in this repository:

benchmark/
├── sample_1000.csv        # 1,000 domains with GPT-5.2 reference labels
└── run_benchmark.py       # Evaluation script

pip install dksplit wordsegment wordninja
python benchmark/run_benchmark.py

Results

Accuracy (agreement with GPT-5.2 on 1,000 real domains):

Model	Accuracy
DKSplit v0.2.2	80.5%
WordSegment	59.1%
WordNinja	47.6%

DKSplit outperforms WordSegment by 21 percentage points and WordNinja by 33 percentage points.

Note: The remaining ~20% disagreement with GPT-5.2 largely comes from rare languages, invented words, and genuinely ambiguous cases (e.g., is christianalucas → christiana lucas or christian a lucas?). On standard English inputs, agreement is significantly higher.

Comparison

Input	DKSplit v0.2.2	WordSegment	WordNinja
chatgptlogin	chatgpt login	chat gpt login	chat gp t login
cloudflarecdn	cloudflare cdn	cloud flare cdn	cloud flare cd n
kubernetescluster	kubernetes cluster	ku bernet es cluster	ku berne tes cluster
instagramlogin	instagram login	insta gram login	insta gram login
ethereumwallet	ethereum wallet	e there um wallet	e there um wallet
spotifyplaylist	spotify playlist	spot if y playlist	spot if y playlist
lululemonoutlet	lululemon outlet	lululemon outlet	lulu lemon outlet
tensorflowlite	tensorflow lite	tensor flow lite	tensor flow lite
mercibeaucoup	merci beaucoup	merci beaucoup	mer ci beau coup
robertdeniro	robert de niro	robert deniro	robert deniro
snowdenyes	snowden yes	snowden yes	snow deny es
youtubedownloader	youtube downloader	youtube downloader	youtube down loader

How It Works

DKSplit treats segmentation as a sequence labeling task.

The training data includes:

• LLM-labeled domain name segmentations
• Brand names
• Personal name combinations
• Multilingual phrases (English, French, German, Spanish, and more)
• Tech product names and terminology

At inference, the BiLSTM runs as an INT8-quantized ONNX model and CRF decoding is performed in NumPy — no GPU required.

Features

• Brand-aware: Recognizes thousands of brands, tech products, and proper nouns
• Multilingual: Handles English, French, German, Spanish, and romanized text
• Lightweight: 9 MB model, minimal dependencies (numpy + onnxruntime)
• Offline: No API keys, no internet required

Limitations

• Characters: Only a-z and 0-9. Input is automatically lowercased.
• Max length: 64 characters.
• Script: Latin script only. Non-Latin scripts (汉字, かな, 한글, العربية) are not supported.
• Ambiguity: Some inputs are genuinely ambiguous. DKSplit optimizes for the most common interpretation.
• Rare languages: Accuracy is highest on English and major European languages.

Requirements

• Python >= 3.8
• numpy
• onnxruntime

Links

• Website: domainkits.com, ABTdomain.com
• GitHub: github.com/ABTdomain/dksplit
• Hugging Face: huggingface.co/ABTdomain/dksplit
• PyPI: pypi.org/project/dksplit
• Go: github.com/ABTdomain/dksplit-go
• Issues: GitHub Issues

License

Apache License 2.0 · Copyright 2026 ABTdomain

Please attribute as: DKsplit by ABTdomain