ACM Transactions on Asian and Low-Resource Language Information Processing最新文献

With the successful application of deep learning, document summarization systems can produce more readable results. However, abstractive summarization still suffers from unfaithful outputs and factual errors, especially in named entities. Current approaches tend to employ external knowledge to improve model performance while neglecting the boundary information and the semantics of the entities. In this paper, we propose an entity-augmented method (EAM) to encourage the model to make full use of the entity boundary information and pay more attention to the critical entities. Experimental results on three Chinese and English summarization datasets show that our method outperforms several strong baselines and achieves state-of-the-art performance on the CLTS dataset. Our method can also improve the faithfulness of the summary and generalize well to different pre-trained language models. Moreover, we propose a method to evaluate the integrity of generated entities. Besides, we adapt the data augmentation method in the FactCC model according to the difference between Chinese and English in grammar and train a new evaluation model for factual consistency evaluation in Chinese summarization.

Several methodologies have recently been proposed to enhance the performance of low-resource Neural Machine Translation (NMT). However, these techniques have yet to be explored thoroughly in low-resource Thai and Myanmar languages. Therefore, we first applied augmentation techniques such as SwitchOut and Ciphertext Based Data Augmentation (CipherDAug) to improve NMT performance in these languages. We secondly enhanced the NMT performance by fine-tuning the pre-trained Multilingual Denoising BART model (mBART), where BART denotes Bidirectional and Auto-Regressive Transformer. We implemented three NMT systems: namely, Transformer+SwitchOut, Multi-source Transformer+CipherDAug, and fine-tuned mBART in the bidirectional translations of Thai-English-Myanmar language pairs from the ASEAN-MT corpus. Experimental results showed that Multi-source Transformer+CipherDAug significantly improved BLEU, ChrF, and TER scores over the first baseline Transformer and second baseline Edit-Based Transformer (EDITOR). The model achieved notable BLEU scores: 37.9 (English-to-Thai), 42.7 (Thai-to-English), 28.9 (English-to-Myanmar), 31.2 (Myanmar-to-English), 25.3 (Thai-to-Myanmar), and 25.5 (Myanmar-to-Thai). The fine-tuned mBART model also considerably outperformed the two baselines, except for the Myanmar-to-English pair. SwitchOut improved over the second baseline in all pairs and performed similarly to the first baseline in most cases. Lastly, we performed detailed analyses verifying that the CipherDAug and mBART models potentially facilitate improving low-resource NMT performance in Thai and Myanmar languages.

Manipuri is a low-resource, Tibeto-Burman tonal language spoken mainly in Manipur, a northeastern state of India. Tone identification is crucial to speech comprehension for tonal languages, where tone defines the word’s meaning. Automatic Speech Recognition for those languages can perform better by including tonal information from a powerful tone detection system. While significant research has been conducted on tonal languages like Mandarin, Thai, Cantonese and Vietnamese, a notable gap exists in exploring Manipuri within this context. To address this gap, this work expands our previously developed handcrafted speech corpus, ManiTo, which comprises of isolated Manipuri tonal contrast word pairs to study the tones of Manipuri. This extension includes contributions from twenty native speakers. Preliminary findings have confirmed that Manipuri has two unique tones, Falling and Level. The study then conducts a comprehensive acoustic feature analysis. Two sets of features based on Pitch contours, Jitter and Shimmer measurements are investigated to distinguish the two tones of Manipuri. Support Vector Machine, Long Short-Term Memory, Random Forest and k-Nearest Neighbors are the classifiers adopted to validate the selected feature sets. The results indicate that the second set of features consistently outperformed the first set, demonstrating higher accuracy, particularly when utilizing the Random Forest classifier, which provides valuable insights for further advancements in speech recognition technology for low-resource tonal language Manipuri.

Current generative knowledge graph construction approaches usually fail to capture structural knowledge by simply flattening natural language into serialized texts or a specification language. However, large generative language model trained on structured data such as code has demonstrated impressive capability in understanding natural language for structural prediction and reasoning tasks. Intuitively, we address the task of generative knowledge graph construction with code language model: given a code-format natural language input, the target is to generate triples which can be represented as code completion tasks. Specifically, we develop schema-aware prompts that effectively utilize the semantic structure within the knowledge graph. As code inherently possesses structure, such as class and function definitions, it serves as a useful model for prior semantic structural knowledge. Furthermore, we employ a rationale-enhanced generation method to boost the performance. Rationales provide intermediate steps, thereby improving knowledge extraction abilities. Experimental results indicate that the proposed approach can obtain better performance on benchmark datasets compared with baselines.

Recent years have witnessed a surge of academic interest in knowledge-enhanced pre-trained language models (PLMs) that incorporate factual knowledge to enhance knowledge-driven applications. Nevertheless, existing studies primarily focus on shallow, static, and separately pre-trained entity embeddings, with few delving into the potential of deep contextualized knowledge representation for knowledge incorporation. Consequently, the performance gains of such models remain limited. In this paper, we introduce a simple yet effective knowledge-enhanced model, College (Contrastive Language-Knowledge Graph Pre-training), which leverages contrastive learning to incorporate factual knowledge into PLMs. This approach maintains the knowledge in its original graph structure to provide the most available information and circumvents the issue of heterogeneous embedding fusion. Experimental results demonstrate that our approach achieves more effective results on several knowledge-intensive tasks compared to previous state-of-the-art methods. Our code and trained models are available at https://github.com/Stacy027/COLLEGE.

In this research, we introduce two new machine learning regression methods: the Ensemble Average and the Pipelined Model. These methods aim to enhance traditional regression analysis for predictive tasks and have undergone thorough evaluation across three datasets: Kaggle House Price, Boston House Price, and California Housing, using various performance metrics. The results consistently show that our models outperform existing methods in terms of accuracy and reliability across all three datasets. The Pipelined Model, in particular, is notable for its ability to combine predictions from multiple models, leading to higher accuracy and impressive scalability. This scalability allows for their application in diverse fields like technology, finance, and healthcare. Furthermore, these models can be adapted for real-time and streaming data analysis, making them valuable for applications such as fraud detection, stock market prediction, and IoT sensor data analysis. Enhancements to the models also make them suitable for big data applications, ensuring their relevance for large datasets and distributed computing environments. It’s important to acknowledge some limitations of our models, including potential data biases, specific assumptions, increased complexity, and challenges related to interpretability when using them in practical scenarios. Nevertheless, these innovations advance predictive modeling, and our comprehensive evaluation underscores their potential to provide increased accuracy and reliability across a wide range of applications. The results indicate that the proposed models outperform existing models in terms of accuracy and robustness for all three datasets. The source code can be found at https://huggingface.co/DebajyotyBanik/Ensemble-Pipelined-Regression/tree/main.

Tamil text segmentation is a long-standing test in language comprehension that entails separating a record into adjacent pieces based on its semantic design. Each segment is important in its own way. The segments are organised according to the purpose of the content examination as text groups, sentences, phrases, words, characters or any other data unit. That process has been portioned using rapid tangled neural organisation in this research, which presents content segmentation methods based on deep learning in natural language processing (NLP). This study proposes a bidirectional long short-term memory (Bi-LSTM) neural network prototype in which fast recurrent neural network (FRNN) are used to learn Tamil text group embedding and phrases are fragmented using text-oriented data. As a result, this prototype is capable of handling variable measured setting data and gives a vast new dataset for naturally segmenting text in Tamil. In addition, we develop a segmentation prototype and show how well it sums up to unnoticeable regular content using this dataset as a base. With Bi-LSTM, the segmentation precision of FRNN is superior to that of other segmentation approaches; however, it is still inferior to that of certain other techniques. Every content is scaled to the required size in the proposed framework, which is immediately accessible for the preparation. This means, each word in a scaled Tamil text is employed to prepare neural organisation as fragmented content. The results reveal that the proposed framework produces high rates of segmentation for manually authored material that are nearly equivalent to segmentation-based plans.

Social media produces large amounts of contents every day. How to predict the potential influences of the contents from a social reply feedback perspective is a key issue that has not been explored. Thus, we propose a novel task named reply keyword prediction in social media, which aims to predict the keywords in the potential replies as many aspects as possible. One prerequisite challenge is that the accessible social media datasets labeling such keywords remain absent. To solve this issue, we propose a new dataset, to study the reply keyword prediction in Social Media. This task could be seen as a single-turn dialogue keyword prediction for open-domain dialogue system. However, existing methods for dialogue keyword prediction cannot be adopted directly, which have two main drawbacks. First, they do not provide an explicit mechanism to model topic complementarity between keywords which is crucial in social media to controllably model all aspects of replies. Second, the collocations of keywords are not explicitly modeled, which also makes it less controllable to optimize for fine-grained prediction since the context information is much less than that in dialogue. To address these issues, we propose a two-stage disentangled framework, which can optimize the complementarity and collocation explicitly in a disentangled fashion. In the first stage, we use a sequence-to-set paradigm via multi-label prediction and determinantal point processes, to generate a set of keyword seeds satisfying the complementarity. In the second stage, we adopt a set-to-sequence paradigm via seq2seq model with the keyword seeds guidance from the set, to generate the more-fine-grained keywords with collocation. Experiments show that this method can generate not only a more diverse set of keywords but also more relevant and consistent keywords. Furthermore, the keywords obtained based on this method can achieve better reply generation results in the retrieval-based system than others.

With its unique information-filtering function, text summarization technology has become a significant aspect of search engines and question-and-answer systems. However, existing models that include the copy mechanism often lack the ability to extract important fragments, resulting in generated content that suffers from thematic deviation and insufficient generalization. Specifically, Chinese automatic summarization using traditional generation methods often loses semantics because of its reliance on word lists. To address these issues, we proposed the novel BioCopy mechanism for the summarization task. By training the tags of predictive words and reducing the probability distribution range on the glossary, we enhanced the ability to generate continuous segments, which effectively solves the above problems. Additionally, we applied reinforced canonicality to the inputs to obtain better model results, making the model share the sub-network weight parameters and sparsing the model output to reduce the search space for model prediction. To further improve the model’s performance, we calculated the bilingual evaluation understudy (BLEU) score on the English dataset CNN/DailyMail to filter the thresholds and reduce the difficulty of word separation and the dependence of the output on the word list. We fully fine-tuned the model using the LCSTS dataset for the Chinese summarization task and conducted small-sample experiments using the CSL dataset. We also conducted ablation experiments on the Chinese dataset. The experimental results demonstrate that the optimized model can learn the semantic representation of the original text better than other models and performs well with small sample sizes.

Sign Language (SL) is the only means of communication for the hearing-impaired people. Normal people have difficulty understanding SL, resulting in a communication barrier between hearing impaired people and hearing community. However, the Sign Language Recognition System (SLRS) has helped to bridge the communication gap. Many SLRs are proposed for recognizing SL; however, a limited number of works are reported for Indian Sign Language (ISL). Most of the existing SLRS focus on global features other than the Region of Interest (ROI). Focusing more on the hand region and extracting local features from the ROI improves system accuracy. The attention mechanism is a widely used technique for emphasizing the ROI. However, only a few SLRS used the attention method. They employed the Convolution Block Attention Module (CBAM) and temporal attention but Spatial Attention (SA) is not utilized in previous SLRS. Therefore, a novel SA based SLRS named Spatial Attention-based Sign Language Recognition Module (SASLRM) is proposed to recognize ISL words for emergency situations. SASLRM recognizes ISL words by combining convolution features from a pretrained VGG-19 model and attention features from a SA module. The proposed model accomplished an average accuracy of 95.627% on the ISL dataset. The proposed SASLRM is further validated on LSA64, WLASL and Cambridge Hand Gesture Recognition (HGR) datasets where, the proposed model reached an accuracy of 97.84 %, 98.86% and 98.22’% respectively. The results indicate the effectiveness of the proposed SLRS in comparison with the existing SLRS.