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.

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

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

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

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

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

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.

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.

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.