Natural Language Generation Tasks Research Articles

Use of SNOMED CT in Large Language Models: Scoping Review.

Large language models (LLMs) have substantially advanced natural language processing (NLP) capabilities but often struggle with knowledge-driven tasks in specialized domains such as biomedicine. Integrating biomedical knowledge sources such as SNOMED CT into LLMs may enhance their performance on biomedical tasks. However, the methodologies and effectiveness of incorporating SNOMED CT into LLMs have not been systematically reviewed. This scoping review aims to examine how SNOMED CT is integrated into LLMs, focusing on (1) the types and components of LLMs being integrated with SNOMED CT, (2) which contents of SNOMED CT are being integrated, and (3) whether this integration improves LLM performance on NLP tasks. Following the PRISMA-ScR (Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews) guidelines, we searched ACM Digital Library, ACL Anthology, IEEE Xplore, PubMed, and Embase for relevant studies published from 2018 to 2023. Studies were included if they incorporated SNOMED CT into LLM pipelines for natural language understanding or generation tasks. Data on LLM types, SNOMED CT integration methods, end tasks, and performance metrics were extracted and synthesized. The review included 37 studies. Bidirectional Encoder Representations from Transformers and its biomedical variants were the most commonly used LLMs. Three main approaches for integrating SNOMED CT were identified: (1) incorporating SNOMED CT into LLM inputs (28/37, 76%), primarily using concept descriptions to expand training corpora; (2) integrating SNOMED CT into additional fusion modules (5/37, 14%); and (3) using SNOMED CT as an external knowledge retriever during inference (5/37, 14%). The most frequent end task was medical concept normalization (15/37, 41%), followed by entity extraction or typing and classification. While most studies (17/19, 89%) reported performance improvements after SNOMED CT integration, only a small fraction (19/37, 51%) provided direct comparisons. The reported gains varied widely across different metrics and tasks, ranging from 0.87% to 131.66%. However, some studies showed either no improvement or a decline in certain performance metrics. This review demonstrates diverse approaches for integrating SNOMED CT into LLMs, with a focus on using concept descriptions to enhance biomedical language understanding and generation. While the results suggest potential benefits of SNOMED CT integration, the lack of standardized evaluation methods and comprehensive performance reporting hinders definitive conclusions about its effectiveness. Future research should prioritize consistent reporting of performance comparisons and explore more sophisticated methods for incorporating SNOMED CT's relational structure into LLMs. In addition, the biomedical NLP community should develop standardized evaluation frameworks to better assess the impact of ontology integration on LLM performance.

Real-Time Extraction of News Events Based on BERT Model

Abstract For the large number of news reports generated every day, in order to obtain effective information from these unstructured news text data more efficiently. In this paper, we study the real-time crawling of news data from news websites through crawling techniques and propose a BERT model-based approach to extract events from news long text. In this study, NetEase news website is selected as an example for real-time extraction to crawl the news data of this website. BERT model as a pre-trained model based on two-way encoded representation of transformer performs well on natural language understanding and natural language generation tasks. In this study, we will fine-tune the training based on BERT model on news corpus related dataset and perform sequence annotation through CRF layer to finally complete the event extraction task. In this paper, the DUEE dataset is chosen to train the model, and the experiments show that the overall performance of the BERT model is better than other network models. Finally, the model of this paper is further optimised, using the ALBERT and RoBERTa models improved on the basis of the BERT model, experiments were conducted, the results show that both models are improved compared to the BERT model, the ALBERT model algorithm performs the best, the model algorithm's F1 value is 1% higher than that of BERT. The results show that the performance is optimised.

ZeroNLG: Aligning and Autoencoding Domains for Zero-Shot Multimodal and Multilingual Natural Language Generation.

Natural Language Generation (NLG) accepts input data in the form of images, videos, or text and generates corresponding natural language text as output. Existing NLG methods mainly adopt a supervised approach and rely heavily on coupled data-to-text pairs. However, for many targeted scenarios and for non-English languages, sufficient quantities of labeled data are often not available. As a result, it is necessary to collect and label data-text pairs for training, which is both costly and time-consuming. To relax the dependency on labeled data of downstream tasks, we propose an intuitive and effective zero-shot learning framework, ZeroNLG, which can deal with multiple NLG tasks, including image-to-text (image captioning), video-to-text (video captioning), and text-to-text (neural machine translation), across English, Chinese, German, and French within a unified framework. ZeroNLG does not require any labeled downstream pairs for training. During training, ZeroNLG (i) projects different domains (across modalities and languages) to corresponding coordinates in a shared common latent space; (ii) bridges different domains by aligning their corresponding coordinates in this space; and (iii) builds an unsupervised multilingual auto-encoder to learn to generate text by reconstructing the input text given its coordinate in shared latent space. Consequently, during inference, based on the data-to-text pipeline, ZeroNLG can generate target sentences across different languages given the coordinate of input data in the common space. Within this unified framework, given visual (imaging or video) data as input, ZeroNLG can perform zero-shot visual captioning; given textual sentences as input, ZeroNLG can perform zero-shot machine translation. We present the results of extensive experiments on twelve NLG tasks, showing that, without using any labeled downstream pairs for training, ZeroNLG generates high-quality and "believable" outputs and significantly outperforms existing zero-shot methods.

Differentially Private Fine-tuning of Language Models

We give simpler, sparser, and faster algorithms for differentially private fine-tuning of large-scale pre-trained language models, which achieve the state-of-the-art privacy versus utility tradeoffs on many standard NLP tasks. We propose a meta-framework for this problem, inspired by the recent success of highly parameter-efficient methods for fine-tuning. Our experiments show that differentially private adaptations of these approaches outperform previous private algorithms in three important dimensions: utility, privacy, and the computational and memory cost of private training. On many commonly studied datasets, the utility of private models approaches that of non-private models. For example, on the MNLI dataset we achieve an accuracy of $87.8\%$ using RoBERTa-Large and $83.5\%$ using RoBERTa-Base with a privacy budget of $\epsilon = 6.7$. In comparison, absent privacy constraints, RoBERTa-Large achieves an accuracy of $90.2\%$. Our findings are similar for natural language generation tasks. Privately fine-tuning with DART, GPT-2-Small, GPT-2-Medium, GPT-2-Large, and GPT-2-XL achieve BLEU scores of 38.5, 42.0, 43.1, and 43.8 respectively (privacy budget of $\epsilon = 6.8,\delta=$ 1e-5) whereas the non-private baseline is $48.1$. All our experiments suggest that larger models are better suited for private fine-tuning: while they are well known to achieve superior accuracy non-privately, we find that they also better maintain their accuracy when privacy is introduced.

Harnessing the Power of LLMs in Practice: A Survey on ChatGPT and Beyond

This article presents a comprehensive and practical guide for practitioners and end-users working with Large Language Models (LLMs) in their downstream Natural Language Processing (NLP) tasks. We provide discussions and insights into the usage of LLMs from the perspectives of models, data, and downstream tasks. First, we offer an introduction and brief summary of current language models. Then, we discuss the influence of pre-training data, training data, and test data. Most importantly, we provide a detailed discussion about the use and non-use cases of large language models for various natural language processing tasks, such as knowledge-intensive tasks, traditional natural language understanding tasks, generation tasks, emergent abilities, and considerations for specific tasks. We present various use cases and non-use cases to illustrate the practical applications and limitations of LLMs in real-world scenarios. We also try to understand the importance of data and the specific challenges associated with each NLP task. Furthermore, we explore the impact of spurious biases on LLMs and delve into other essential considerations, such as efficiency, cost, and latency, to ensure a comprehensive understanding of deploying LLMs in practice. This comprehensive guide aims to provide researchers and practitioners with valuable insights and best practices for working with LLMs, thereby enabling the successful implementation of these models in a wide range of NLP tasks. A curated list of practical guide resources of LLMs, regularly updated, can be found at https://github.com/Mooler0410/LLMsPracticalGuide . An LLMs evolutionary tree, editable yet regularly updated, can be found at llmtree.ai .

CPT: a pre-trained unbalanced transformer for both Chinese language understanding and generation

In this paper, we take the advantage of previous pre-trained models (PTMs) and propose a novel Chinese Pre-trained Unbalanced Transformer (CPT). Different from previous Chinese PTMs, CPT is designed to utilize the shared knowledge between natural language understanding (NLU) and natural language generation (NLG) to boost the performance. CPT consists of three parts: a shared encoder, an understanding decoder, and a generation decoder. Two specific decoders with a shared encoder are pre-trained with masked language modeling (MLM) and denoising auto-encoding (DAE) tasks, respectively. With the partially shared architecture and multi-task pre-training, CPT can (1) learn specific knowledge of both NLU or NLG tasks with two decoders and (2) be fine-tuned flexibly that fully exploits the potential of the model. Moreover, the unbalanced Transformer saves the computational and storage cost, which makes CPT competitive and greatly accelerates the inference of text generation. Experimental results on a wide range of Chinese NLU and NLG tasks show the effectiveness of CPT.

Contrastive Learning Penalized Cross-Entropy with Diversity Contrastive Search Decoding for Diagnostic Report Generation of Reduced Token Repetition

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.

ConsistentEE: A Consistent and Hardness-Guided Early Exiting Method for Accelerating Language Models Inference

Early Exiting is one of the most popular methods to achieve efficient inference. Current early exiting methods adopt the (weighted) sum of the cross entropy loss of all internal classifiers as the objective function during training, imposing all these classifiers to predict all instances correctly. However, during inference, as long as one internal classifier predicts an instance correctly, it can accelerate without losing accuracy. Thus, there is a notable gap between training and inference. We propose ConsistentEE, an early exiting method that is consistent in training and inference. ConsistentEE formulates the early exiting process as a reinforcement learning problem. A policy network is added to decide whether an instance should exit or continue. The training objective of ConsistentEE only requires each instance to be predicted correctly by one internal classifier. Additionally, we introduce the concept "Memorized Layer" to measure the hardness of an instance. We incorporate the memorized layer into reward function design, which allows "easy'' instances to focus more on acceleration while ``hard'' instances to focus more on accuracy. Experimental results show that our method outperforms other baselines on various natural language understanding and generation tasks using PLMs and LLMs as backbones respectively.

Beyond Attention: Breaking the Limits of Transformer Context Length with Recurrent Memory

A major limitation for the broader scope of problems solvable by transformers is the quadratic scaling of computational complexity with input size. In this study, we investigate the recurrent memory augmentation of pre-trained transformer models to extend input context length while linearly scaling compute. Our approach demonstrates the capability to store information in memory for sequences of up to an unprecedented two million tokens while maintaining high retrieval accuracy. Experiments with language modeling tasks show perplexity improvement as the number of processed input segments increases. These results underscore the effectiveness of our method, which has significant potential to enhance long-term dependency handling in natural language understanding and generation tasks, as well as enable large-scale context processing for memory-intensive applications.

A Review of Current Trends, Techniques, and Challenges in Large Language Models (LLMs)

Natural language processing (NLP) has significantly transformed in the last decade, especially in the field of language modeling. Large language models (LLMs) have achieved SOTA performances on natural language understanding (NLU) and natural language generation (NLG) tasks by learning language representation in self-supervised ways. This paper provides a comprehensive survey to capture the progression of advances in language models. In this paper, we examine the different aspects of language models, which started with a few million parameters but have reached the size of a trillion in a very short time. We also look at how these LLMs transitioned from task-specific to task-independent to task-and-language-independent architectures. This paper extensively discusses different pretraining objectives, benchmarks, and transfer learning methods used in LLMs. It also examines different finetuning and in-context learning techniques used in downstream tasks. Moreover, it explores how LLMs can perform well across many domains and datasets if sufficiently trained on a large and diverse dataset. Next, it discusses how, over time, the availability of cheap computational power and large datasets have improved LLM’s capabilities and raised new challenges. As part of our study, we also inspect LLMs from the perspective of scalability to see how their performance is affected by the model’s depth, width, and data size. Lastly, we provide an empirical comparison of existing trends and techniques and a comprehensive analysis of where the field of LLM currently stands.

How do control tokens affect natural language generation tasks like text simplification

Abstract Recent work on text simplification has focused on the use of control tokens to further the state-of-the-art. However, it is not easy to further improve without an in-depth comprehension of the mechanisms underlying control tokens. One unexplored factor is the tokenization strategy, which we also explore. In this paper, we (1) reimplemented AudienCe-CEntric Sentence Simplification, (2) explored the effects and interactions of varying control tokens, (3) tested the influences of different tokenization strategies, (4) demonstrated how separate control tokens affect performance and (5) proposed new methods to predict the value of control tokens. We show variations of performance in the four control tokens separately. We also uncover how the design of control tokens could influence performance and give some suggestions for designing control tokens. We show the newly proposed method with higher performance in both SARI (a common scoring metric in text simplificaiton) and BERTScore (a score derived from the BERT language model) and potential in real applications.

Enhancing diversity for logical table‐to‐text generation with mixture of experts

AbstractLogical table‐to‐text generation is a task within the realm of natural language generation (NLG) that aims to generate coherent and logically faithful sentences based on tables. Unlike conventional NLG tasks, this task demands not only surface‐level fluency but also a high degree of logic‐level fidelity in the generated outputs. Current table‐to‐text systems grapple with various quality issues, such as repetitive generation, insufficient reasoning and limited complexity. Therefore, we introduce LogicMoE, a dedicated Mixture‐of‐Experts (MoE) model tailored for logical table‐to‐text generation. The primary objective of LogicMoE is to enrich the diversity of generated sentences from both semantic and logical perspectives. In particular, each expert within the model serves as a specialized generator responsible for generating sentences of a specific logical type. Additionally, we propose and employ novel evaluation metrics to comprehensively assess the diversity of generated outputs. Our experimental results showcase LogicMoE's superiority with absolute improvements of 0.8 and 2.2 in BLEU‐3 over the strong baselines on LogicNLG and Logic2Text datasets, respectively, driving the state‐of‐the‐art performance to a new level. Furthermore, we highlight its inherent advantages in terms of diversity and controllability, signifying its potential to spearhead advancements in logical table‐to‐text generation applications.

CogNLG: Cognitive graph for KG‐to‐text generation

AbstractKnowledge graph (KG) has been fully considered in natural language generation (NLG) tasks. A KG can help models generate controllable text and achieve better performance. However, most existing related approaches still lack explainability and scalability in large‐scale knowledge reasoning. In this work, we propose a novel CogNLG framework for KG‐to‐text generation tasks. Our CogNLG is implemented based on the dual‐process theory in cognitive science. It consists of two systems: one system acts as the analytic system for knowledge extraction, and another is the perceptual system for text generation by using existing knowledge. During text generation, CogNLG provides a visible and explainable reasoning path. Our framework shows excellent performance on all datasets and achieves a BLEU score of 36.7, which increases by 6.7 compared to the best competitor.

Constraint-Based Adversarial Networks for Unsupervised Abstract Text Summarization

Abstract text summarization is a classic sequence-to-sequence natural language generation task. In order to improve the quality of unsupervised abstract text summarization in unsupervised mode, we propose two constraints for training text summarization model, embedding space constraint and information ratio constraint. We construct a generative adversarial network with two discriminators based on these two constraints (TC-SUM-GAN). We use unsupervised and supervised methods to train the model in the experiment. Experimental results show that the ROUGE-1 value of the unsupervised TC-SUM-GAN increases by [Formula: see text] points compared with the basic model and at least 1.96 points compared with other comparative models. The ROUGE scores of the supervised TC-SUM-GAN are also improved. TC-SUM-GAN achieves very competitive results for the metrics of ROUGE-1 and ROUGE-2. In addition, the abstracts generated by our model are closer to those generated manually.

Mutually improved response generation and dialogue summarization for multi-domain task-oriented dialogue systems

With the development of pre-trained language models (PrLM), the research of PrLM-based multi-domain task-oriented dialogue systems (TOD) has attracted growing attention and has achieved great progress. However, most current studies suffer from two problems. First, they model dialogue state tracking as an independent subtask supervised by slot-value pairs, resulting in poor adaptability when transferring to new task domains. Second, these studies ignore the fact that not all dialogue histories are valuable for the ensuing turns as they increase in length. To tackle these two issues, we propose a simple and novel framework to explore multi-domain TOD by jointly training response generation and dialogue summarization with PrLM as the backbone. Specifically, first, we use fluent text generated by the dialogue summarization model to replace formatted dialogue states, treating the dialogue state identification task as a natural language generation task, which allows dialogue state tracking to be easily extended to new task domains. Second, dialogue summarization removes redundant and useless information from the current dialogue process and is fed into the response decoder enabling the system to focus on crucial details in long dialogues. Furthermore, we employ a dialogue chunk detector to assist the dialogue summarization model and design a fusion mechanism to dynamically integrate helpful dialogue summarization into the response generation process. Experimental results show that the proposed model achieves state-of-the-art performance in both automatic and human evaluations on two public datasets and demonstrate that the joint framework can benefit both tasks from each other.

Practical and ethical challenges of large language models in education: A systematic scoping review

AbstractEducational technology innovations leveraging large language models (LLMs) have shown the potential to automate the laborious process of generating and analysing textual content. While various innovations have been developed to automate a range of educational tasks (eg, question generation, feedback provision, and essay grading), there are concerns regarding the practicality and ethicality of these innovations. Such concerns may hinder future research and the adoption of LLMs‐based innovations in authentic educational contexts. To address this, we conducted a systematic scoping review of 118 peer‐reviewed papers published since 2017 to pinpoint the current state of research on using LLMs to automate and support educational tasks. The findings revealed 53 use cases for LLMs in automating education tasks, categorised into nine main categories: profiling/labelling, detection, grading, teaching support, prediction, knowledge representation, feedback, content generation, and recommendation. Additionally, we also identified several practical and ethical challenges, including low technological readiness, lack of replicability and transparency and insufficient privacy and beneficence considerations. The findings were summarised into three recommendations for future studies, including updating existing innovations with state‐of‐the‐art models (eg, GPT‐3/4), embracing the initiative of open‐sourcing models/systems, and adopting a human‐centred approach throughout the developmental process. As the intersection of AI and education is continuously evolving, the findings of this study can serve as an essential reference point for researchers, allowing them to leverage the strengths, learn from the limitations, and uncover potential research opportunities enabled by ChatGPT and other generative AI models. Practitioner notesWhat is currently known about this topic Generating and analysing text‐based content are time‐consuming and laborious tasks. Large language models are capable of efficiently analysing an unprecedented amount of textual content and completing complex natural language processing and generation tasks. Large language models have been increasingly used to develop educational technologies that aim to automate the generation and analysis of textual content, such as automated question generation and essay scoring. What this paper adds A comprehensive list of different educational tasks that could potentially benefit from LLMs‐based innovations through automation. A structured assessment of the practicality and ethicality of existing LLMs‐based innovations from seven important aspects using established frameworks. Three recommendations that could potentially support future studies to develop LLMs‐based innovations that are practical and ethical to implement in authentic educational contexts. Implications for practice and/or policy Updating existing innovations with state‐of‐the‐art models may further reduce the amount of manual effort required for adapting existing models to different educational tasks. The reporting standards of empirical research that aims to develop educational technologies using large language models need to be improved. Adopting a human‐centred approach throughout the developmental process could contribute to resolving the practical and ethical challenges of large language models in education.

Defending against Backdoor Attacks in Natural Language Generation

The frustratingly fragile nature of neural network models make current natural language generation (NLG) systems prone to backdoor attacks and generate malicious sequences that could be sexist or offensive. Unfortunately, little effort has been invested to how backdoor attacks can affect current NLG models and how to defend against these attacks. In this work, by giving a formal definition of backdoor attack and defense, we investigate this problem on two important NLG tasks, machine translation and dialog generation. Tailored to the inherent nature of NLG models (e.g., producing a sequence of coherent words given contexts), we design defending strategies against attacks. We find that testing the backward probability of generating sources given targets yields effective defense performance against all different types of attacks, and is able to handle the one-to-many issue in many NLG tasks such as dialog generation. We hope that this work can raise the awareness of backdoor risks concealed in deep NLG systems and inspire more future work (both attack and defense) towards this direction.

Metadial: A Meta-learning Approach for Arabic Dialogue Generation

Dialogue generation is the automatic generation of a text response, given a user’s input. Dialogue generation for low-resource languages has been a challenging tasks for researchers. However, the advancements in deep learning models have made developing conversational agents that perform the tasks of dialogue generation not only possible, but also effective and helpful in many applications spanning a variety of domains. Nevertheless, work on conversational bots for low-resource languages such as the Arabic language is still limited due to various challenges, including the language structure, vocabulary, and the scarcity of its data resources. Meta-learning has been introduced before in the natural language processing (NLP) realm and showed significant improvements in many tasks; however, it has rarely been used in natural language generation (NLG) tasks and never in Arabic NLG. In this work, we propose a meta-learning approach for Arabic dialogue generation for fast adaptation on low-resource domains, namely, Arabic. We start by using existing pre-trained models; we then meta-learn the initial parameters on high-resource dataset before finetuning the parameters on the target tasks. We prove that the proposed model that employs meta-learning techniques improves generalization and enables fast adaptation of the transformer model on low-resource NLG tasks. We report gains in the BLEU-4 and improvements in Semantic textual Similarity (STS) metrics when compared to the existing state-of-the-art approach. We also do a further study on the effectiveness of the meta-learning algorithms on the response generation of the models.

Regularization-based pruning of irrelevant weights in deep neural architectures

Deep neural networks exploiting million parameters are currently the norm. This is a potential issue because of the great number of computations needed for training, and the possible loss of generalization performance of overparameterized networks. We propose in this paper a method for learning sparse neural topologies via a regularization approach that identifies nonrelevant weights in any type of layer (i.e., convolutional, fully connected, attention and embedding ones) and selectively shrinks their norm while performing a standard back-propagation update for relevant layers. This technique, which is an improvement of classical weight decay, is based on the definition of a regularization term that can be added to any loss function regardless of its form, resulting in a unified general framework exploitable in many different contexts. The actual elimination of parameters identified as irrelevant is handled by an iterative pruning algorithm.To explore the possibility of an interdisciplinary use of our proposed technique, we test it on six different image classification and natural language generation tasks, among which four are based on real datasets. We reach state-of-the-art performance in one out of four imaging tasks while obtaining results better than competitors for the others and one out of two of the considered language generation tasks, both in terms of compression and metrics.

Generic Mechanism for Reducing Repetitions in Encoder-decoder Models

Encoder-decoder models have been commonly used; they have achieved state-of-the-art results for many natural language generation tasks. However, according to the reports of previous studies, encoder-decoder models suffer from generating redundant repetitions. Thus, we herein propose a repetition reduction module (RRM) for encoder-decoder models that estimates the semantic difference of a source sentence before and after it is fed into the model to capture the consistency between the two sides. As an autoencoder, the proposed mechanism supervises the training of encoder-decoder models to reduce the number of repeatedly generated tokens. The evaluation results of the publicly available machine translation and response generation datasets demonstrate the effectiveness of our proposal.