Search

Medical imaging description and disease diagnosis are vitally important yet time-consuming. Automated diagnosis report generation (DRG) from medical imaging description can reduce clinicians' workload and improve their routine efficiency. To address this natural language generation task, fine-tuning a pre-trained large language model (LLM) is cost-effective and indispensable, and its success has been witnessed in many downstream applications. However, semantic inconsistency of sentence embeddings has been massively observed from undesirable repetitions or unnaturalness in text generation. To address the underlying issue of anisotropic distribution of token representation, in this study, a contrastive learning penalized cross-entropy (CLpCE) objective function is implemented to enhance the semantic consistency and accuracy of token representation by guiding the fine-tuning procedure towards a specific task. Furthermore, to improve the diversity of token generation in text summarization and to prevent sampling from unreliable tail of token distributions, a diversity contrastive search (DCS) decoding method is designed for restricting the report generation derived from a probable candidate set with maintained semantic coherence. Furthermore, a novel metric named the maximum of token repetition ratio (maxTRR) is proposed to estimate the token diversity and to help determine the candidate output. Based on the LLM of a generative pre-trained Transformer 2 (GPT-2) of Chinese version, the proposed CLpCE with DCS (CLpCEwDCS) decoding framework is validated on 30,000 desensitized text samples from the "Medical Imaging Diagnosis Report Generation" track of 2023 Global Artificial Intelligence Technology Innovation Competition. Using four kinds of metrics evaluated from n-gram word matching, semantic relevance, and content similarity as well as the maxTRR metric extensive experiments reveal that the proposed framework effectively maintains semantic coherence and accuracy (BLEU-1, 0.4937; BLEU-2, 0.4107; BLEU-3, 0.3461; BLEU-4, 0.2933; METEOR, 0.2612; ROUGE, 0.5182; CIDER, 1.4339) and improves text generation diversity and naturalness (maxTRR, 0.12). The phenomenon of dull or repetitive text generation is common when fine-tuning pre-trained LLMs for natural language processing applications. This study might shed some light on relieving this issue by developing comprehensive strategies to enhance semantic coherence, accuracy and diversity of sentence embeddings. [ABSTRACT FROM AUTHOR]

(1) Background: Information overload challenges decision-making in the Industry 4.0 era. While Natural Language Processing (NLP), especially Automatic Text Summarization (ATS), offers solutions, issues with factual accuracy persist. This research bridges cognitive neuroscience and NLP, aiming to improve model interpretability. (2) Methods: This research examined four fact extraction techniques: dependency relation, named entity recognition, part-of-speech tagging, and TF-IDF, in order to explore their correlation with human EEG signals. Representational Similarity Analysis (RSA) was applied to gauge the relationship between language models and brain activity. (3) Results: Named entity recognition showed the highest sensitivity to EEG signals, marking the most significant differentiation between factual and non-factual words with a score of −0.99. The dependency relation followed with −0.90, while part-of-speech tagging and TF-IDF resulted in 0.07 and −0.52, respectively. Deep language models such as GloVe, BERT, and GPT-2 exhibited noticeable influences on RSA scores, highlighting the nuanced interplay between brain activity and these models. (4) Conclusions: Our findings emphasize the crucial role of named entity recognition and dependency relations in fact extraction and demonstrate the independent effects of different models and TOIs on RSA scores. These insights aim to refine algorithms to reflect human text processing better, thereby enhancing ATS models' factual integrity. [ABSTRACT FROM AUTHOR]

As the popularity of large language models (LLMs) has risen over the course of the last year, led by GPT-3/4 and especially its productization as ChatGPT, we have witnessed the extensive application of LLMs to text summarization. However, LLMs do not intrinsically have the power to verify the correctness of the information they supply and generate. This research introduces a novel approach to abstractive summarization, aiming to address the limitations of LLMs in that they struggle to understand the truth. The proposed method leverages extracted knowledge graph information and structured semantics as a guide for summarization. Building upon BART, one of the state-of-the-art sequence-to-sequence pre-trained LLMs, multi-source transformer modules are developed as an encoder, which are capable of processing textual and graphical inputs. Decoding is performed based on this enriched encoding to enhance the summary quality. The Wiki-Sum dataset, derived from Wikipedia text dumps, is introduced for evaluation purposes. Comparative experiments with baseline models demonstrate the strengths of the proposed approach in generating informative and relevant summaries. We conclude by presenting our insights into utilizing LLMs with graph external information, which will become a powerful aid towards the goal of factually correct and verified LLMs. [ABSTRACT FROM AUTHOR]

Text summarization is a subtask of natural language processing referring to the automatic creation of a concise and fluent summary that captures the main ideas and topics from one or multiple documents. Earlier literature surveys focus on extractive approaches, which rank the top-n most important sentences in the input document and then combine them to form a summary. As argued in the literature, the summaries of these approaches do not have the same lexical flow or coherence as summaries that are manually produced by humans. Newer surveys elaborate abstractive approaches, which generate a summary with potentially new phrases and sentences compared to the input document. Generally speaking, contrary to the extractive approaches, the abstractive ones create summaries that are more similar to those produced by humans. However, these approaches still lack the contextual representation needed to form fluent summaries. Recent advancements in deep learning and pretrained language models led to the improvement of many natural language processing tasks, including abstractive summarization. Overall, these surveys do not present a comprehensive evaluation framework that assesses the aforementioned approaches. Taking the above into account, the contribution of this survey is fourfold: (i) we provide a comprehensive survey of the state-of-the-art approaches in text summarization; (ii) we conduct a comparative evaluation of these approaches, using well-known datasets from the related literature, as well as popular evaluation scores such as ROUGE-1, ROUGE-2, ROUGE-L, ROUGE-LSUM, BLEU-1, BLEU-2 and SACREBLEU; (iii) we report on insights gained on various aspects of the text summarization process, including existing approaches, datasets and evaluation methods, and we outline a set of open issues and future research directions; (iv) we upload the datasets and the code used in our experiments in a public repository, aiming to increase the reproducibility of this work and facilitate future research in the field. [ABSTRACT FROM AUTHOR]

Recent cross-lingual summarization research has pursued the use of a unified end-to-end model which has demonstrated a certain level of improvement in performance and effectiveness, but this approach stitches together multiple tasks and makes the computation more complex. Less work has focused on alignment relationships across languages, which has led to persistent problems of summary misordering and loss of key information. For this reason, we first simplify the multitasking by converting the translation task into an equal proportion of cross-lingual summary tasks so that the model can perform only cross-lingual summary tasks when generating cross-lingual summaries. In addition, we splice monolingual and cross-lingual summary sequences as an input so that the model can fully learn the core content of the corpus. Then, we propose a reinforced regularization method based on the model to improve its robustness, and build a targeted ABO mechanism to enhance the semantic relationship alignment and key information retention of the cross-lingual summaries. Ablation experiments are conducted on three datasets of different orders of magnitude to demonstrate the effective enhancement of the model by the optimization approach; they outperform the mainstream approaches on the cross-lingual summarization task and the monolingual summarization task for the full dataset. Finally, we validate the model's capabilities on a cross-lingual summary dataset of professional domains, and the results demonstrate its superior performance and ability to improve cross-lingual sequencing. [ABSTRACT FROM AUTHOR]

Named entity recognition (NER) is a subfield of natural language processing (NLP) that identifies and classifies entities from plain text, such as people, organizations, locations, and other types. NER is a fundamental task in information extraction, information retrieval, and text summarization, as it helps to organize the relevant information in a structured way. The current approaches to Chinese named entity recognition do not consider the category information of matched Chinese words, which limits their ability to capture the correlation between words. This makes Chinese NER more challenging than English NER, which already has well-defined word boundaries. To improve Chinese NER, it is necessary to develop new approaches that take into account category features of matched Chinese words, and the category information would help to effectively capture the relationship between words. This paper proposes a Prompt-based Word-level Information Injection BERT (PWII-BERT) to integrate prompt-guided lexicon information into a pre-trained language model. Specifically, we engineer a Word-level Information Injection Adapter (WIIA) through the original Transformer encoder and prompt-guided Transformer layers. Thus, the ability of PWII-BERT to explicitly obtain fine-grained character-to-word relevant information according to the category prompt is one of its key advantages. In experiments on four benchmark datasets, PWII-BERT outperforms the baselines, demonstrating the significance of fully utilizing the advantages of fusing the category information and lexicon feature to implement Chinese NER. [ABSTRACT FROM AUTHOR]

Large text documents are sometimes challenging to understand and time-consuming to extract vital information from. These issues are addressed by automatic text summarizing techniques, which condense lengthy texts while preserving their key information. Thus, the development of automatic summarization systems capable of fulfilling the ever-increasing demands of textual data becomes of utmost importance. It is even more vital with complex natural languages. This study explores five State-Of-The-Art (SOTA) Arabic deep Transformer-based Language Models (TLMs) in the task of text summarization by adapting various text summarization datasets dedicated to Arabic. A comparison against deep learning and machine learning-based baseline models has also been conducted. Experimental results reveal the superiority of TLMs, specifically the PEAGASUS family, against the baseline approaches, with an average F1-score of 90% on several benchmark datasets. [ABSTRACT FROM AUTHOR]