In recent years, the intersection of naturallanguageprocessing (NLP) and healthcare has emerged as a frontier for innovation, offering new pathways to enhance medical data analysis and healthcare automation. This pa...
详细信息
The safety management construction of the hydro-power units is necessary to prevent the quality failures. However, the traditional hydro-power units lack a unified safety management decision-making platform, making kn...
详细信息
Long-Form Question Answering (LFQA) represents a growing interest in Legal naturallanguageprocessing (Legal-NLP) as many individuals encounter legal disputes at some point in their lives, but lack of knowledge about...
详细信息
ISBN:
(纸本)9798400704369
Long-Form Question Answering (LFQA) represents a growing interest in Legal naturallanguageprocessing (Legal-NLP) as many individuals encounter legal disputes at some point in their lives, but lack of knowledge about how to negotiate these complex situations might put them at risk. The endeavor to generate detailed answers to contextually rich legal questions has faced challenges, primarily due to the limited availability of specialized datasets involving intensive manual effort or incapability of existing LFQA models to produce informative responses. Addressing this, our research introduces a semi-synthetic dataset, Legal-LFQA (L2FQA) created by exploiting a large language model (LLM) and utilizing contexts derived from existing legal datasets. Additionally, we hypothesize that integrating legal reasoning into the answer generation process of the LLMs will help bolster both the quality and interpretability of the produced responses. We systematically analyze the quality of L2FQA using human evaluation and naturallanguage inference based metrics. Next, we benchmark L2FQA on a wide range of general-purpose and domain-specific LLMs using fine-tuning and in-context learning (with zero, one and few shot) strategies. The efficacy of these techniques is gauged through several automated and human evaluations. Results indicate that incorporating legal reasoning into the answer generation process provides an avenue for improving the quality of responses in the context of Legal-LFQA task. By addressing the challenges faced in LFQA and emphasizing the potential of interpretability, this research contributes to the foundational work in enhancing question-answering systems within the legal domain.
Sincerely in part to the rise of high-performance computer systems and transformer models, naturallanguageprocessing has advanced. Also, a multitude of applications built on large language models continually improve...
详细信息
ISBN:
(纸本)9798400709784
Sincerely in part to the rise of high-performance computer systems and transformer models, naturallanguageprocessing has advanced. Also, a multitude of applications built on large language models continually improve people's cognitive abilities. Large language models continue to face difficulties when dealing with long context input. Many studies have suggested various specific strategies to address the challenge of extended context, however as of yet, no thorough summary of these studies exists. In this paper, we discuss the issues raised and the developments that have occurred in the long context application of large language models, and we attempt to suggest future directions for research and development.
Sentence representation is a major challenge in naturallanguageprocessing, especially in multilingual environments. Current approaches to sentence representation using Large language Models (LLMs) often require larg...
详细信息
ISBN:
(纸本)9798350359329;9798350359312
Sentence representation is a major challenge in naturallanguageprocessing, especially in multilingual environments. Current approaches to sentence representation using Large language Models (LLMs) often require large amounts of data for fine-tuning, and research has focused on English content. In addition, comparative datasets translated directly from English can contain many semantic and syntactic errors. To address these issues, we propose a new approach to enhance multilingual sentence embeddings using LLMs and knowledge graphs. We first present a dedicated designed prompt that exploits in-context learning of LLMs for sentence embedding without fine-tuning. We further introduce an innovative method that utilizes knowledge graphs, such as Wikidata, for generating diverse multilingual training data for contrastive finetuning. This approach significantly reduces the reliance on translated sentences and mitigates issues related to translation accuracy. Furthermore, we develop a unique multilingual contrastive learning loss function, which, when combined with QLora's efficient fine-tuning technique, enables LLMs to achieve state-of-the-art performance in Sentence Text Similarity (STS) tasks, even with limited computational resources.
作者:
Zhang, TongTang, AilingYan, RongCollege of Computer Science
Inner Mongolia University Inner Mongolia Key Laboratory of Mongolian Information Processing Technology National & Local Joint Engineering Research Center of Intelligent Information Processing Technology for Mongolian Hohhot010021 China
Short text classification is an important naturallanguageprocessing task due to the prevalence of short text on the internet and social media platforms. In this paper, we propose a novel graph-based short text class...
详细信息
Mining knowledge from the rapid growth geological documents is essential for the development of geoscience. languages such as Chinese, which consist of continuous characters, are difficult for computer processing and ...
详细信息
ISBN:
(纸本)9789819755004;9789819755011
Mining knowledge from the rapid growth geological documents is essential for the development of geoscience. languages such as Chinese, which consist of continuous characters, are difficult for computer processing and understanding, and word segmentation is a prerequisite in this situation. The quality of word segmentation has a crucial impact on the correct completion of downstream tasks, such as named entity recognition, relationship extraction and dependency analysis. Due to the specialization and complexity of the documents in the geological field, tools and models built in general domain usually have a poor performance in solving this problem. In this paper, we propose a new model named DualBERT which combines general model with domain model, and provides the ability to adapt to the new scenarios while maintaining the adaptability to the original scene. We introduce a words segmentation correction method based on the similarity with geologic terms, in order to enhance the ability to identify the boundaries of highly domain-specific geologic terms.
Large language Models (LLMs) are gaining popularity in the field of naturallanguageprocessing (NLP) due to their remarkable accuracy in various NLP tasks. LLMs designed for coding are trained on massive datasets, wh...
详细信息
Advancements in mobile technology makes it easier to communicate in real time, but at the cost of having a wider potential attack area for phishing. While there has been research in the field related to Email and SMS,...
详细信息
ISBN:
(纸本)9798350328837;9798350328844
Advancements in mobile technology makes it easier to communicate in real time, but at the cost of having a wider potential attack area for phishing. While there has been research in the field related to Email and SMS, Instant Messages lags behind. The widespread usage of instant messengers by individuals of all ages further motivates the addition of software security features in this context. This research aims to detect phishing in mobile instant messages by analysing the language of the message with the help of naturallanguageprocessing to detect keywords pointing towards phishing. We built the machine learning models using 3 different methods for feature extraction and 3 classification algorithms. Our tests showed that balancing the data with random oversampling increased the classifiers' performance, which were able to achieve an accuracy up to 99.2%.
Fake news detection (FND) is a critical task in naturallanguageprocessing (NLP) focused on identifying and mitigating the spread of misinformation. Large language models (LLMs) have recently shown remarkable abiliti...
详细信息
Fake news detection (FND) is a critical task in naturallanguageprocessing (NLP) focused on identifying and mitigating the spread of misinformation. Large language models (LLMs) have recently shown remarkable abilities in understanding semantics and performing logical inference. However, their tendency to generate hallucinations poses significant challenges in accurately detecting deceptive content, leading to suboptimal performance. In addition, existing FND methods often underutilize the extensive prior knowledge embedded within LLMs, resulting in less effective classification outcomes. To address these issues, we propose the CAPE-FND framework, context-aware prompt engineering, designed for enhancing FND tasks. This framework employs unique veracity-oriented context-aware constraints, background information, and analogical reasoning to mitigate LLM hallucinations and utilizes self-adaptive bootstrap prompting optimization to improve LLM predictions. It further refines initial LLM prompts through adaptive iterative optimization using a random search bootstrap algorithm, maximizing the efficacy of LLM prompting. Extensive zero-shot and few-shot experiments using GPT-3.5-turbo across multiple public datasets demonstrate the effectiveness and robustness of our CAPE-FND framework, even surpassing advanced GPT-4.0 and human performance in certain scenarios. To support further LLM-based FND, we have made our approach's code publicly available on GitHub (our CAPE-FND code: [Accessed on 2024.09]).
暂无评论