domain-classification

Classifying the content of domains

Classify the kind of content hosted by the domain using the domain name, and text and screenshot of the homepage.

Domain-adversarial neural network for emotion recognition based on facial expressions task. The 2 domains are “Women” and “Men” and the domain-adaptation challenge is training the network on one sex and applying it on the other.

An alternative, crawler-like browser that looks beneath the surface of the semantic web.

Open-source ML classifier for gambling domain detection. TFLite, Python, TensorFlow.

Production NestJS domain classifier (99.98% accuracy) Hybrid ML: Whitelist + Regex Heuristics + OpenRouter LLM fallback 918 domains → 700/sec → $0 cost → 100% TypeScript

IAB Content Taxonomy classification via LLM knowledge distillation -- TF-IDF + LinearSVC achieves 91.6% top-1 on 27 categories using LLM-corrected labels and keywords

Detecting malware-generated (DGA) domain names from lexical and character n-gram features with XGBoost, Random Forest, and Logistic Regression

Hybrid SBERT–LLM domain classification system using semantic retrieval and Qwen2.5 reranking to identify the top business domains from natural language queries.

Splits a parallel corpus into domains for stratified NMT training. Reads only the English side, so it works for any language pair with English on one side - the translation is carried through aligned. A pretrained classifier handles two thirds; an LLM-seeded fastText model decides the rest, with no hand-labelled training data.

Multi-task NLP system (shared XLM-RoBERTa encoder) for joint technology NER and multi-label domain classification from bilingual project descriptions. Trained end-to-end, deployed via ONNX + FastAPI.

Aplikasi Android untuk memantau akses domain terindikasi pornografi di perangkat sendiri - VpnService + AccessibilityService, offline-first (Hive), Flutter + Firebase. Artefak penelitian Tugas Akhir.

Host → category and service catalog for NetTrafficControl, a DNS-filtering iOS app. seeds/ is the source of truth, catalog.json is a deterministic build output, and every entry is licence-checked and validated. CC0 1.0.