Romanian Style Chinese Modern Poetry Generation with Pre-Trained Model and Direct Preference Optimization

Pre-Trained Models (PTMs) can learn general knowledge representations and perform well in Natural Language Processing (NLP) tasks. For the Chinese language, several PTMs are developed, however, most existing methods concentrate on modern Chinese and are not ideal for processing classical Chinese due to the differences in grammars and semantics between these two forms. In this paper, in order to process two forms of Chinese uniformly, we propose a novel Classical and Modern Chinese pre-trained language model (CANCN-BERT), with the advantage of effectively processing both classical and modern Chinese, which is an extension of BERT. Form-aware pre-training tasks are elaborately designed to train our model, so as to better adapt it to classical and modern Chinese corpus. Moreover, we define a joint model, proposing dedicated optimization methods through different paths with the control of the switch mechanism. Our model merges characteristics of both classical and modern Chinese, which can adequately and efficiently enhance the representation ability for both forms. Extensive experiments show that our model outperforms baseline models on processing classical and modern Chinese and achieves significant and consistent improvements. Also, the results of ablation experiments demonstrate the effectiveness of each module.

Classical Chinese poetry has a history of thousands of years and is a precious cultural heritage of humankind. Compared with the modern Chinese corpus, it is irrecoverable and specially organized, making it difficult to be learned by existing pre-trained language models. Besides, with the thousands of years of development, many words in classical Chinese poetry have changed their meanings or been out of use today, which further limiting the capability of existing pre-trained models to learn the semantics of classical Chinese poetry. To address these challenges, we construct a large-scale sememe knowledge graph of classical Chinese Poetry (SKG-Poetry), which connects the vocabularies in classical Chinese poetry and modern Chinese. By extracting the sememe knowledge from classical Chinese poetry, our model PoetryBERT not only enlarges the irrecoverable pre-training corpus but also enriches the semantics of the vocabularies in classical Chinese poetry, which enables PoetryBERT to be successfully used in downstream tasks. Specifically, we evaluate our model in two tasks in the field of Chinese classical poetry, which are poetry theme classification and poetry-modern Chinese translation. Extensive experiments are conducted on the two tasks to show the effectiveness of sememe knowledge based pre-training model.

Reviewed by: Chinese Poetic Modernisms ed. by Paul Manfredi and Christopher Lupke C. T. Au Paul Manfredi and Christopher Lupke, editors. Chinese Poetic Modernisms. Leiden: Brill, 2019. 403 p. There is a saying that a good book has no ending, and Chinese Poetic Modernisms definitely falls under this category. This long-awaited ambitious project undertakes "the daunting task of examining Chinese poetic modernisms in different periods and locales" (7). Most of those ambitions are, in the end, fulfilled. The concept of Chinese modernisms has long been indistinctly present. In books focusing on modern Chinese literature, more often than not Taiwanese modernism, Hong Kong modernism, and Macanese modernism are treated separately from that of mainland China. The editors pledge to bring the concept to light through fourteen essays, divided into four sections that help delineate varied trajectories of modernist poetic trends. By following the book's logical structure, readers first become familiar with the major features of Chinese modernism that emerged mainly during the republican period. For example, Lan Dizhi's essay provides a comprehensive overview of the development of Chinese modernist poetry. The discussion revolves around the literary arts journal Xiandai, tracing its roots back to Li Jinfa and his contemporaries in the 1920s and tracking the trend's evolution through the 1940s (the Nine Leaves school, Jiuyepai) and identifying modernist characteristics such as "anti-lyricism," "anti-improvisation," and "anti-logical" (29). Modernist poets were eager to "advocate 'modern feeling' and 'pure poetry'" (32). Lan sheds light on two more characteristics embedded in Chinese modernisms from the outset. For one thing, modernist poets were under the influence of various Western sources, though in a complicated manner; for another, classical Chinese poetry played a significant role in their works. Contributors explore these intricate relationships. To name only a few Western influences: Li Jinfa was influenced by French Symbolism, Dai Wangshu by Spanish modernism (Lan 21-37), Feng Zhi by Rainer Maria Rilke (Geraldine Fiss 38-56), Yuan Kejia by T. S. Eliot (Yanhong Zhu 57-81), and Wen Yiduo's little poems by Shakespearean sonnets (Dian Li 82-106). Lan concludes that, as far as "acquisition of literary elements from both Chinese and [End Page 238] Western traditions" is concerned, these poets adopted a comparative literature approach which allowed them to absorb "the strengths and carefully eschewing the weaknesses of each tradition" (37). Lan's views, to various extents, are echoed by other contributors, not only in this section, but also elsewhere in the book. Expanding upon the apparent comprehensiveness of Lan's account of the first and third moments, the emergence of the Xiandai modernist group and the Nine Leaves school, Michelle Yeh's, Chen Fangming's (132-52), and Ruan Meihui's (153-80) discussions widen the geographic range and time frame, focusing on, but not limited to, postwar Taiwanese modernist poetry. Yeh identifies "six distinct moments when modernism is embraced as a literary paradigm" (108). While the formation of Le Moulin Poetry Society (Fengche Shishe) (1935-1936), which advocated Surrealism, is considered the second moment of modernism, the Modernist school (Xiandaipai) and the Epoch Poetry Society (Chuangshiji Shishe) arose in the 1950s as the fourth moment. Ruan's and Chen's essays, respectively, help elaborate these two moments. The fifth moment refers to the Modernist movement in Hong Kong in 1956, when Ma Lang published the journal Wenyi Xinchao (New Trends in Literature and Arts). Finally, the modernist poetry that emerged in post-Mao China is considered the sixth moment. Interestingly enough, Yeh's major concern does not lie in any one of these moments. By examining Xia Yu's poetry, Yeh seems to imply that there is one more moment–a seventh–in Chinese modernisms, which suggests that modernism is still an active force from the 1980s onward in Taiwan (111). At the outset, Chinese modernisms are an outgrowth of comparative studies. The fact that Dai Wangshu had translated Paul Van Tieghem's On Comparative Literature (1931) reminds us of the significance of the mutual relationships of different literatures. To varying degrees, most contributors in sections 1 and 2 have embraced the concept of "influence study." The emphasis in section 3, however, lies elsewhere; factual connection between...

Chinese traditional poetry is an important art form, and it is a meaningful task to use artificial intelligence algorithm to generate poetry. In the past, researchers have proposed different generation algorithms, but these algorithms have some limitations.This paper mainly tries to use BART and other pretrained models, proposes FS2TEXT and RR2TEXT to generate metrical poetry text and even specific style poetry text, and solves the problem that the user’s writing intention gradually reduces the relevance of the generated poetry text. Finally, in order to test the effect of the algorithm, this paper imitates Turing test and finds more than 600 testers to distinguish the works generated by the algorithm from the works of real human poets, and the results show that they can’t tell whether the work is from an algorithm or a human.This shows that the algorithm proposed in this paper works well.

In this study, predictor variables (age, gender, region and language family) affecting the scores of Turkish language learners are examined through multiple regression method. The study group consisted of 280 international students registered to Turkish Language Teaching Centers located at Gazi and Hacettepe Universities. The research data were obtained from the Turkish course completion exam papers and personal information forms. According to the results, the average scores of the students from the Afro-Asiatic, Indo-European, Bantu, Sino-Tibetan and Austronesian language families were lower than those from the Altai language family. Additionally, the writing scores of the students from the Afro-Asiatic and Austronesian language families; the speaking scores of the students from Afro-Asiatic, Indo-European language families; reading comprehension scores of the students from Afro-Asiatic, Indo-European, Bantu and Sino-Tibetan language families and grammar scores of the students from Sino-Tibetan and Austronesian language families were lower than the scores of the Altai language family. In addition, while the age variable was found to have a positive effect on speaking scores, it was observed that area and gender variables were not significant predicators of scores. Findings are discussed in the light of literature and suggestions for further research are provided.

Super-resolution based on deep convolutional networks is currently gaining much attention from both academia and industry. However, lack of proper evaluation measures makes it difficult to compare approaches, hampering progress in the field. Traditional measures, such as PSNR or SSIM, are known to poorly correlate with the human perception of image quality. Therefore, in existing works common practice is also to report Mean-Opinion-Score (MOS) — the results of human evaluation of super-resolved images. Unfortunately, the MOS values from different papers are not directly comparable, due to the varying number of raters, their subjectivity, etc. By this paper, we introduce Neural Side-By-Side — a new measure that allows super-resolution models to be compared automatically, effectively approximating human preferences. Namely, we collect a large dataset of aligned image pairs, which were produced by different super-resolution models. Then each pair is annotated by several raters, who were instructed to choose a more visually appealing image. Given the dataset and the labels, we trained a CNN model that obtains a pair of images and for each image predicts a probability of being more preferable than its counterpart. In this work, we show that Neural Side-By-Side generalizes across both new models and new data. Hence, it can serve as a natural approximation of human preferences, which can be used to compare models or tune hyperparameters without raters’ assistance. We open-source the dataset and the pretrained model <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1</sup> and expect that it will become a handy tool for researchers and practitioners.

The daily language in mainland China has experienced a shift from traditional Chinese language to modern mandarin Chinese at the beginning of the twentieth century. The Chinese poetry ‘revolution’started in the 1910s is considered as a turning point in the Chinese poetry evolution process due to the novel applications of the modern Chinese language. Many temporal poetic studies consider the poems written in traditional Chinese and modern Chinese as two different genres. The two genres are saliently different in rhyme, meter, theme, etc. We aim to detect the specific properties of the evolution process of Chinese poetry in terms of the word categories’ distribution patterns. For the purpose, a corpus with 438 randomly selected traditional and modern Chinese poems is built, and some quantitative language indicators (entropy, relative entropy, repeat rate) and some exploratory statistical analysis techniques applicable in corpus linguistics and quantitative linguistics (one-way ANOVA test, cluster analysis)1 are used to abstract and analyze language data from the corpus. It is concluded that the word categories are distributed significantly differently in traditional poetry and modern poetry. The sound reasons would be that (1) traditional Chinese poetry is more likely to focus on the application of some specific content word categories, for example, nouns, but not auxiliary words and (2) modern poems tend to choose more categories of words. From the perspective of word class distribution patterns, we suppose that the birth of modern Chinese poetry in the 1910s is a sharp change to Chinese poetry production.

We propose EXAMS -- a new benchmark dataset for cross-lingual and multilingual question answering for high school examinations. We collected more than 24,000 high-quality high school exam questions in 16 languages, covering 8 language families and 24 school subjects from Natural Sciences and Social Sciences, among others. EXAMS offers a fine-grained evaluation framework across multiple languages and subjects, which allows precise analysis and comparison of various models. We perform various experiments with existing top-performing multilingual pre-trained models and we show that EXAMS offers multiple challenges that require multilingual knowledge and reasoning in multiple domains. We hope that EXAMS will enable researchers to explore challenging reasoning and knowledge transfer methods and pre-trained models for school question answering in various languages which was not possible before. The data, code, pre-trained models, and evaluation are available at https://github.com/mhardalov/exams-qa.

Rhetoric is a vital element in modern poetry, and plays an essential role in improving its aesthetics. However, to date, it has not been considered in research on automatic poetry generation. In this paper, we propose a rhetorically controlled encoder-decoder for modern Chinese poetry generation. Our model relies on a continuous latent variable as a rhetoric controller to capture various rhetorical patterns in an encoder, and then incorporates rhetoric-based mixtures while generating modern Chinese poetry. For metaphor and personification, an automated evaluation shows that our model outperforms state-of-the-art baselines by a substantial margin, while human evaluation shows that our model generates better poems than baseline methods in terms of fluency, coherence, meaningfulness, and rhetorical aesthetics.

The creative possibilities and constraints of artificial intelligence in poetry generation are examined in this study, with an emphasis on how AI questions conventional ideas of authorship, creativity, and emotional expression. Three different poems, as well as an original thematic poem, were produced using ChatGPT-4 using thoughtfully crafted prompts that drew inspiration from the writings of Pablo Neruda and E.E. Cummings. Each poem was subjected to a thorough self-analysis that looked at thematic depth, literary devices, stylistic choices, and the degree to which AI can replicate human voice and emotion. While recognizing important limitations like the lack of lived experience, cultural context, and emotional authenticity in AI-generated content, the study also discusses graphological aspects and the point of view used in the poems. By providing insights into how machine -generated literature is changing our perception of literary authorship, the findings add to the continuing discussion about artificial intelligence and the creative arts. This study concludes that although AI cannot fully replace human creativity, it does offer significant chances for artistic expression and teamwork.

Literature has a strong cultural imprint and regional color, including poetry. Natural language itself is part of the poetry style. It is interesting to attempt to use one language to present poetry in another language style. Therefore, in this study, we propose a method to fine-tune a pre-trained model in a targeted manner to automatically generate French-style modern Chinese poetry and conduct a multi-faceted evaluation of the generated results. In a five-point scale based on human evaluation, judges assigned scores between 3.29 and 3.93 in seven dimensions, which reached 80.8–93.6% of the scores of the Chinese versions of real French poetry in these dimensions. In terms of the high-frequency poetic imagery, the consistency of the top 30–50 high-frequency poetic images between the poetry generated by the fine-tuned model and the French poetry reached 50–60%. In terms of the syntactic features, compared with the poems generated by the baseline model, the distribution frequencies of three special types of words that appear relatively frequently in French poetry increased by 12.95%, 15.81%, and 284.44% per 1000 Chinese characters in the poetry generated by the fine-tuned model. The human evaluation, poetic image distribution, and syntactic feature statistics show that the targeted fine-tuned model is helpful for the spread of language style. This fine-tuned model can successfully generate modern Chinese poetry in a French style.

Aim/Purpose: The study aims to develop a multilingual healthcare assistance chatbot that provides real-time, accurate answers to a query related to health matters in multiple languages. Conversion of written responses into spoken words lets users have the medical information necessary for them without interrupting communication between patients and health services. The purpose of this system is to break the language barriers for healthcare users, making it easier for them to access vital medical advice and resources. Background: This research focuses on fine-tuned large language models (LLMs) for providing accurate, context-aware responses in multiple languages with speech-based output. The chatbot, built on pre-trained Hugging Face models and fine-tuned with healthcare datasets, demonstrates a comprehensive understanding of medical terminology, symptoms, and healthcare concepts across languages. Unlike many existing chatbots that offer limited medical knowledge or support only a single language, the proposed chatbot leverages fine-tuning on a specialized medical corpus to deliver more accurate, context-rich responses. Furthermore, it provides text-based and speech-based outputs, improving user engagement and accessibility compared to text-only models. Methodology: A multilingual healthcare assistance chatbot is proposed using the pre-trained model aboonaji/llama2finetune-v2 and the specialized medical dataset aboonaji/wiki_medical_terms_llam2_format from Hugging Face. Key steps in the methodology include cleaning and normalizing medical terms, symptoms, and treatment advice to ensure uniformity across multiple languages. The model is fine-tuned on this healthcare dataset, enabling accurate and context-sensitive responses. Text-to-speech (TTS) technology is integrated to provide natural-sounding, voice-based answers, enhancing accessibility. Multilingual capability is ensured through modules for smooth language transitions. The chatbot is deployed on an intuitive web or mobile platform, simplifying user interaction. Performance metrics, including response accuracy, linguistic consistency, and user satisfaction, continuously improve through feedback and periodic updates with evolving medical knowledge and language models. Contribution: This research adds value to the medical sector by maximizing access to healthcare information across heterogeneous linguistic groups. It uses advanced natural language processing techniques and text-to-speech technology, facilitating quick and efficient interaction between patients and health providers. This allows users to follow crucial medical advice and information in their preferred language, thus promoting greater patient understanding and engagement. The output of the accurate, context-sensitive responses to healthcare search terms given by the chatbot assists in bridging the gap between patients and medical resources to make informed decisions for better overall health literacy. This model works as an instrumental instrument in solving language barriers in healthcare by introducing inclusiveness and promoting a stronger case for equality in healthcare. Findings: Results indicate that the chatbot effectively addresses language gaps in healthcare by generating contextually accurate and relevant responses to medical queries with excellent quality and reliability. Performance metrics demonstrate a BLEU score between 0.8 and 0.9, a perplexity score of 80.45, and an average latency of 20 seconds, highlighting robust translation accuracy, coherent response generation, and reasonable interaction time. Text-to-speech integration enhances accessibility and user engagement, while high user satisfaction confirms its potential to improve health literacy and patient comprehension. Continuous feedback during testing has enabled iterative refinements, ensuring the chatbot remains a reliable and inclusive tool for medical information delivery. Recommendations for Practitioners: Clinical practitioners should also encourage the adoption of the multilingual healthcare assistance chatbot in their clinical settings to enhance engagement and communication with the patient. This will enable healthcare providers to effectively bridge the language gap to provide patients with the exact health information they wish to receive in the language. The practice should encourage the different patient populations to use the chatbot and assist them in seeking information confidently. Recommendation for Researchers: It would be great to challenge this multilingual medical assistance chatbot through further research, for example, testing in other languages and improving its natural language processing properties to provide users with accurate medical answers. This study should be followed by further studies measuring the extended effects of chatbots on health literacy and patient outcomes in different healthcare settings. Collaboration in design with healthcare professionals will provide insights into user needs and ensure the chatbot remains practical and meaningful. Additionally, artificial intelligence and machine learning could enhance the learning of the interactions from the interactions with the chatbot, thus enhancing its effectiveness over time. These efforts can significantly advance technology in healthcare communication and patient support. Impact on Society: A multilingual health assistance chatbot can greatly affect society by giving diverse populations easy access to essential health information. It bridges the language barrier gaps, enabling individuals of many linguistic backgrounds to gain self-confident medical advice and information, thus furthering health literacy and informed decision-making. That enables better healthcare because the patients will be able to understand what might be wrong with them and likely to comply with some prescriptions made for the treatment. In addition, the chatbot encourages egalitarian healthcare because it allows for the inclusion of marginalized groups of society to be treated equally. At the same time, there should still be an equal occurrence in the healthcare system. This means, in turn, that the chatbot does not only enhance individual results in health but also community health at large because it is sure to encourage proactive engagement with the services and healthcare resources. Future Research: Future research may focus on expanding the knowledge database for the chatbot by incorporating many languages and dialects. It could also work on perfecting the natural language processing to interpret complex medical-related queries better. The integration of more advanced techniques of artificial intelligence may also further enhance the learning abilities of chatbots from user interactions and sharpen their response across time cycles.

The paper opens with the introduction which presents a background about the reasons behind the appearance of certain characteristics for Modern Poetry as a literary genre in Modern World. It discusses the circumstances and environment of the movements that absolutely have the great impact on developing this genre. The introduction also sheds light on the role of some Modern Poets who have participated , through their writings , in promoting Modern Poetry. Then , section one comes to define Modern Poetry , discussing Modernism and its influence upon Modern Poetry ; how the modern reader perceive Modern Poetry by the modern writers' themes and messages for the public. Some examples are also given in order to reinforce what has previously been discussed. As for section two , the main characteristics of Modern Poetry are analyzed as far as Modern Literature is concerned including poetry. This section revolves round how Modern poets focus on applying the elements of some features to their poetry appropriately in a common way ; so the reader, later, becomes conscious of the messages of the poems. Modern Poets identify these characteristics for a purpose. They study their reader's mind and the way he or she judges ideas , which is accordingly a good base for poets in order to write manageable poems to such readers. Mind plays a decisive role in writing poems for the technological evolution of the modern world and people's changing life system. This does not , necessarily, mean that Modern Poetry goes far beyond traditional techniques. Many Modern poets are still holding the traditional style and mixing the modern to the old in one poetic form. Emotions and imagination , on the other hand , have a position in some modern poems too. Sometimes , poets gather mind , heart , imagination , sensibility , and materialism in one place for a certain message in the poem. Some poems are selected for analysis in this section. Finally, the conclusion sums up the result of the study. Going through the characteristics of Modern Poetry may cause confusion because of complexity , however , it is a reliable source to tackle and discuss serious issues. In fact , Modern Poetry is found to be a good means to change society and heal many wounds that many modern societies suffer from. https://doi.org/10.24897/acn.64.68.258

Numerical solutions to partial differential equations (PDEs) are instrumental for material structural design where extensive data screening is needed. However, traditional numerical methods demand significant computational resources, highlighting the need for innovative optimization algorithms to streamline design exploration. Direct gradient-based optimization algorithms, while effective, rely on design initialization and require complex, problem-specific sensitivity derivations. The advent of machine learning offers a promising alternative to handling large parameter spaces. To further mitigate data dependency, researchers have developed physics-informed neural networks (PINNs) to learn directly from PDEs. However, the intrinsic continuity requirement of PINNs restricts their application in structural mechanics problems, especially for composite materials. Our work addresses this discontinuity issue by substituting the PDE residual with a weak formulation in the physics-informed training process. The proposed approach is exemplified in modeling digital materials, which are mathematical representations of complex composites that possess extreme structural discontinuity. This article also introduces an interactive process that integrates physics-informed loss with design objectives, eliminating the need for pretrained surrogate models or analytical sensitivity derivations. The results demonstrate that our approach can preserve the physical accuracy in data-free material surrogate modeling but also accelerates the direct optimization process without model pretraining.

Generating poetry using machine and deep learning techniques has been a challenging and exciting topic of research in recent years. It has significance in natural language processing and computational linguistics. This study introduces an innovative approach to generate high-quality Pashto poetry by leveraging two pre-trained transformer models, LaMini-Cerebras-590M and bloomz-560m. The models were trained on an extensive new and quality Pashto poetry dataset to learn the underlying complex patterns and structures. The trained models are then used to generate new Pashto poetry by providing them with a seed text or prompt. To evaluate the quality of the generated poetry, we conducted both subjective and objective evaluations, including human evaluation. The experimental results demonstrate that the proposed approach can generate Pashto poetry that is comparable in quality to human-generated poetry. The study provides a valuable contribution to the field of Pashto language and poetry generation and has potential applications in natural language processing and computational linguistics.