Improved Sequence Labeling Algorithms and Their Applications for Translation Error Detection

DIRECT: Direct Decoding for Efficient and Aligned Sequence Labeling with Large Language Models

Sequence labeling is a fine-grained information extraction task, yet existing large language model-based approaches suffer from insufficient domain alignment and low inference efficiency. To address these issues, we propose DIRECT, a framework that addresses these issues through training-time optimization and inference-time rectification. Specifically, DIRECT performs Direct Preference Optimization (DPO) after supervised fine-tuning to strengthen task alignment with human preferences, and introduces a controlled decoding process that enforces fixed output formats and restricts predictions to candidate sets. To further improve efficiency, a template-filling mechanism requires the model to generate only label tokens while reusing prefixed content through the KV Cache, thus reducing redundant computation. Experimental results on eight datasets demonstrate that DIRECT achieves significant improvements in both performance and efficiency compared to existing methods.

BERT-based Automatic Error Correction System for Chinese Learners

Current automatic error correction methods for Chinese learners often focus on superficial word- or sentence-level processing and are affected by inconsistent annotation standards, resulting in limited generalization to real learner texts and causing misalignment or overcorrection. To improve grammatical correctness, semantic fidelity, and instructional relevance, this paper constructs a multi-granularity lexical representation method. Using characters as the basic unit, the method integrates three types of information: characters, words, and pinyin. These representations are learned through projection processing, concatenated into a unified embedding, and fed into a pre-trained Chinese BERT encoder. A dual-task head consisting of sequence labeling and lightweight generation is deployed on the shared BERT encoder for collaborative optimization. Key innovations include an alignment consistency loss to ensure character-level consistency, and the combination of word-segmentation augmentation and multi-reference soft-label training to reduce conflicts caused by segmentation differences. Experimental results show that the method achieves granular alignment consistency above 0.800, corrected-original sentence similarity of 0.888, and an overcorrection rate as low as 0.041. These results demonstrate improved robustness and explainability for automatic Chinese learner error correction.

Improving low-resource neural machine translation by semantic distance augmentation

Neural machine translation (NMT) has witnessed substantial advancements, leveraging its learning capability to deliver highly accurate translations. Nevertheless, the efficacy of NMT models is contingent upon the accessibility of extensive-scale, high-quality training data, and its performance suffers notably in the absence of such datasets. To tackle this challenge, we propose a semantic distance augmentation (SDA) method that integrates syntactic information from constituency parse trees into the NMT encoder to optimize self-attention. Specifically, the source language sentences in the training set are analyzed by constituency parse analysis and the semantic distance attention matrix is constructed. Then, a fusion strategy is designed to integrate this matrix into the self-attention weight, enhancing the representation of the source sentences. In addition, a SDA length-aware strategy is proposed to adaptively control the contribution of semantic distance in the attention computation. Empirical evaluations across multiple low-resource language pairs reveal that the SDA method achieves statistically significant improvements in translation quality over the strong baseline, without requiring additional training data or increasing model complexity.

Ensemble-Based Approach for Amazigh POS Tagging: Leveraging Multiple Models for Enhanced Performance in Low-Resource Language Processing

Part-of-Speech (POS) tagging is a foundational task in Natural Language Processing (NLP), yet it remains challenging for low-resource and morphologically rich languages such as Amazigh. This paper proposes a hybrid ensemble framework for Amazigh POS tagging that integrates three complementary models: a Bidirectional Long Short-Term Memory network (BiLSTM), a Conditional Random Field model (CRF), and a rule-based morphological analyzer (RBMA). Rather than treating prior results obtained on different corpora and tag inventories as directly comparable, the study evaluates all proposed components under a common 54-tag experimental setting based on the publicly available Amazigh Linguistic Dataset. Three ensemble strategies are examined: majority voting, validation-weighted voting, and logistic-regression stacking. An additional late-fusion ablation applies hard and soft RBMA constraints to CRF and Stacking outputs; hard masking degrades performance substantially, whereas soft masking is more robust but remains below unconstrained decoding. The best micro-level performance is obtained by the stacking ensemble, which reaches 98.51% Micro-F1/accuracy, whereas the boosting-like weighted ensemble obtains the strongest Macro-F1 among the ensemble variants, reaching 74.24%. These results show that hybrid ensemble methods can improve token-level accuracy in low-resource POS tagging, while also revealing a trade-off between frequent-tag accuracy and rare-tag robustness. The findings highlight the usefulness of combining neural, probabilistic, and rule-based information for Amazigh POS tagging, and point to class-balanced meta-learning and character/subword representations as important directions for improving rare and out-of-vocabulary categories.

Enhancing Fake News Detection in Low-Resource Linguistic Contexts using Translation-based NER and Lightweight NLI

A robust preprocessing pipeline incorporating Google Translation, spaCy-based NER with hyphenated-word normalization, and a future-event-aware fallback logic is proposed, achieving a 40 percent reduction in inference latency and closing the linguistic generalization gap left by previous works.

Zero-shot translation enhancement algorithm based on multilingual neural machine translation model