Efficient Multilingual Name Type Classification Using Convolutional Networks

TL;DR

The paper tackles multilingual proper-name classification by language and entity type, a task complicated by sparse name resources and transliteration across scripts. It introduces Onomas-CNN X, a CPU-optimized CNN with parallel branches, depthwise-separable convolutions, a learned pooling mix, and a two-stage hierarchical language clustering to reduce complexity. Empirical results show 92.1% accuracy on 104 languages with 2,813 names/second on a single CPU core, delivering about 46× faster throughput and energy savings versus fine-tuned XLM-RoBERTa while maintaining comparable accuracy. The work demonstrates that specialized, data-efficient architectures can rival large transformers for focused NLP tasks, enabling real-time deployment on commodity hardware and offering substantial practical benefits for production systems processing large volumes of names.

Abstract

We present a convolutional neural network approach for classifying proper names by language and entity type. Our model, Onomas-CNN X, combines parallel convolution branches with depthwise-separable operations and hierarchical classification to process names efficiently on CPU hardware. We evaluate the architecture on a large multilingual dataset covering 104 languages and four entity types (person, organization, location, other). Onomas-CNN X achieves 92.1% accuracy while processing 2,813 names per second on a single CPU core - 46 times faster than fine-tuned XLM-RoBERTa with comparable accuracy. The model reduces energy consumption by a factor of 46 compared to transformer baselines. Our experiments demonstrate that specialized CNN architectures remain competitive with large pre-trained models for focused NLP tasks when sufficient training data exists.