Neural Semi-Markov Conditional Random Fields for Robust Character-Based\n Part-of-Speech Tagging

2018/08/13 by Apostolos Kemos, Kemos, Apostolos, Heike Adel +3 · 1 citation
Computer Science · #Computation and Language (cs.CL) #FOS: Computer and information sciences #Natural Language Processing Techniques #Speech Recognition and Synthesis #Topic Modeling

paper · pdf · doi:10.48550/arxiv.1808.04208

openalex publication_date 2018/08/13 · openalex created_date 2022/08/04 · openalex updated_date 2026/07/28

Abstract

Character-level models of tokens have been shown to be effective at dealing\nwith within-token noise and out-of-vocabulary words. But these models still\nrely on correct token boundaries. In this paper, we propose a novel end-to-end\ncharacter-level model and demonstrate its effectiveness in multilingual\nsettings and when token boundaries are noisy. Our model is a semi-Markov\nconditional random field with neural networks for character and segment\nrepresentation. It requires no tokenizer. The model matches state-of-the-art\nbaselines for various languages and significantly outperforms them on a noisy\nEnglish version of a part-of-speech tagging benchmark dataset. Our code and the\nnoisy dataset are publicly available at http://cistern.cis.lmu.de/semiCRF.\n

Citations

• Character-Aware Neural Language Models
• Long Short-Term Memory
• Segmental Recurrent Neural Networks

Cited by

• A Survey on Recent Advances in Sequence Labeling from Deep Learning Models

Related

• MANTa: Efficient Gradient-Based Tokenization for Robust End-to-End Language Modeling
• Character Eyes: Seeing Language through Character-Level Taggers
• Bridging Pre-trained Language Models and Hand-crafted Features for Unsupervised POS Tagging
• Neural Token Segmentation for High Token-Internal Complexity
• How Robust Are Character-Based Word Embeddings in Tagging and MT Against Wrod Scramlbing or Randdm Nouse?