Word Alignment Research Articles

A Novel Alignment-based Approach for PARSEVAL Measures

Abstract We propose a novel method for calculating PARSEVAL measures to evaluate constituent parsing results. Previous constituent parsing evaluation techniques were constrained by the requirement for consistent sentence boundaries and tokenization results, proving to be stringent and inconvenient. Our new approach handles constituent parsing results obtained from raw text, even when sentence boundaries and tokenization differ from the preprocessed gold sentence. Implementing this measure is our evaluation by alignment approach. The algorithm enables the alignment of tokens and sentences in the gold and system parse trees. Our proposed algorithm draws on the analogy of sentence and word alignment commonly used in machine translation (MT). To demonstrate the intricacy of calculations and clarify any integration of configurations, we explain the implementations in detailed pseudo-code and provide empirical proof for how sentence and word alignment can improve evaluation reliability.

Sign detection for cuneiform tablets

Abstract Among the many excavated cuneiform tablets, only a small portion has been analyzed by Assyriologists. Learning how to read cuneiform is a lengthy and challenging process that can take years to complete. This work aims to improve the automatic detection of cuneiform signs from 2D images of cuneiform tablets. The results can later be used for NLP tasks such as semantic annotation, word alignment and machine translation to assist Assyriologists in their research. We introduce the largest publicly available annotated dataset of cuneiform signs to date. It comprises of 52,102 signs from 315 fully annotated tablets, equating to 512 distinct images. In addition, we have preprocessed and refined four existing datasets, resulting in a comprehensive collection of 88,536 signs. Since some signs are not localized on fully annotated tablets, the total dataset encompasses 593 fully annotated cuneiform tablets, resulting in 654 images. Our efforts to expand this dataset are ongoing. Furthermore, we evaluate two state-of-the-art methods to establish benchmarks in the field. The first is a two-stage supervised sign detection approach that involves: (1) the identification of bounding boxes, and (2) the classification of each sign within these boxes. The second method employs an object detection model. Given the numerous classes and their varied distribution, the task of cuneiform sign detection poses a significant challenge in machine learning. This paper aims to lay a groundwork for future research, offering both a substantial dataset and initial methodologies for sign detection on cuneiform tablets.

Improving neural machine translation by word translations

Transformer-based neural machine translation (NMT) models have achieved state-of-the-art performance in the machine translation paradigm. These models learn the translation knowledge from the bilingual corpus through the attention mechanism automatically. This differs from the way human translators approach sentence translation, where prior knowledge plays a significant role. Inspired by this, a word translation augmentation (WTA) method is proposed to improve the Transformer-based NMT model. The main steps are as follows: Firstly, constructing the word alignment rules based on the training set. Next, generating the translation rules for source words according to the word alignment rules. Lastly, incorporating the potential translation candidates for each source word into the NMT model during the training and testing procedure. In addition, the WTA method introduces the idea of Mixup for translation candidates of a source word and employs two augmentation strategies to augment the encoder. The results of experiments on several translation tasks with high-resource and low-resource indicate the effectiveness of the proposed method compared with the corresponding strong baseline, and the improvement in BLEU score achieved ranges from 0.42 to 0.63.

Enhancing Low-Resource Relation Representations through Multi-View Decoupling

Recently, prompt-tuning with pre-trained language models (PLMs) has demonstrated the significantly enhancing ability of relation extraction (RE) tasks. However, in low-resource scenarios, where the available training data is scarce, previous prompt-based methods may still perform poorly for prompt-based representation learning due to a superficial understanding of the relation. To this end, we highlight the importance of learning high-quality relation representation in low-resource scenarios for RE, and propose a novel prompt-based relation representation method, named MVRE (Multi-View Relation Extraction), to better leverage the capacity of PLMs to improve the performance of RE within the low-resource prompt-tuning paradigm. Specifically, MVRE decouples each relation into different perspectives to encompass multi-view relation representations for maximizing the likelihood during relation inference. Furthermore, we also design a Global-Local loss and a Dynamic-Initialization method for better alignment of the multi-view relation-representing virtual words, containing the semantics of relation labels during the optimization learning process and initialization. Extensive experiments on three benchmark datasets show that our method can achieve state-of-the-art in low-resource settings.

Better in sync: Temporal dynamics explain multisensory word-action-object learning in early development.

We investigated the temporal impact of multisensory settings on children's learning of word-object and action-object associations at 1- and 2-years of age. Specifically, we examined whether the temporal alignment of words and actions influenced the acquisition of novel word-action-object associations. We used a preferential looking and violation of expectation task in which infants and young children were first presented with two distinct word-object and action-object pairings either in a synchronous (overlapping in time) or sequential manner (one after the other). Findings revealed that 2-year-olds recognized both, action-object and word-object associations when they first saw the word-action-object combinations synchronously, but not sequentially, as evidenced by looking behavior. 1-year-olds did not show evidence for recognition for either of the word-object and action-object pairs, regardless of the initial temporal alignment of these cues. To control for individual differences, we explored factors that might influence associative learning based on parental reports of 1- and 2-year-olds development, however, developmental measures did not explain word-action-object associative learning in either group. We discuss that while young children may benefit from the temporal alignment of multisensory cues as it enables them to actively engage with the multisensory content in real-time, infants may have been overwhelmed by the complexity of this input.

Rethinking the Exploitation of Monolingual Data for Low-Resource Neural Machine Translation

Abstract The utilization of monolingual data has been shown to be a promising strategy for addressing low-resource machine translation problems. Previous studies have demonstrated the effectiveness of techniques such as back-translation and self-supervised objectives, including masked language modeling, causal language modeling, and denoise autoencoding, in improving the performance of machine translation models. However, the manner in which these methods contribute to the success of machine translation tasks and how they can be effectively combined remains an under-researched area. In this study, we carry out a systematic investigation of the effects of these techniques on linguistic properties through the use of probing tasks, including source language comprehension, bilingual word alignment, and translation fluency. We further evaluate the impact of pre-training, back-translation, and multi-task learning on bitexts of varying sizes. Our findings inform the design of more effective pipelines for leveraging monolingual data in extremely low-resource and low-resource machine translation tasks. Experiment results show consistent performance gains in seven translation directions, which provide further support for our conclusions and understanding of the role of monolingual data in machine translation.

TAXN: Translate Align Extract Normalize, a Multilingual Extraction Tool for Clinical Texts.

Several studies have shown that about 80% of the medical information in an electronic health record is only available through unstructured data. Resources such as medical terminologies in languages other than English are limited and restrain the NLP tools. We propose here to leverage English based resources in other languages using a combination of translation, word alignment, entity extraction and term normalization (TAXN). We implement this extraction pipeline in an open-source library called "medkit". We demonstrate the interest of this approach through a specific use-case: enriching a phenotypic dictionary for post-acute sequelae in COVID-19 (PASC). TAXN proved to be efficient to propose new synonyms of UMLS terms using a corpus of 70 articles in French with 356 terms enriched with at least one validated new synonym. This study was based on freely available deep-learning models.

On Applying Natural Language Processing Technology to Optimize Accuracy of English Interactive Translation Platform

Abstract This research establishes a phrase-based statistical machine translation platform composed of multiple functional modules, including data preprocessing, word alignment, translation model training, ordering model training, language model training, decoding, and parameter optimization. Special emphasis is placed on the quality of bilingual word alignment, utilizing both the IBM model and the statistical collocation model to enhance alignment accuracy. The study reveals that this new platform significantly enhances translation efficiency across diverse corpora, such as legal documents, news articles, and parliamentary records. Notably, the translation of parliamentary records demonstrates a remarkable reduction in average translation time by 27.53% and a decrease in the number of decoding iterations by 56.46% when compared to System1 using the Baseline system. Furthermore, a BLEU-based evaluation indicates that the Baseline system consistently surpasses the comparison system in translating various text types. This study underscores the practical effectiveness of the developed translation platform and its contribution to the advancement of natural language processing technology.

A Study of Business English Translation Skills Based on Parallel Corpus

Abstract In this paper, the support vector machine is used as the overall classification of the paragraph alignment method. The paragraph length relationship is taken as an essential feature of the business English paragraph alignment model, the CRFs discriminative model is used as the word alignment task model, and the maximum likelihood estimation algorithm is used for model parameter training. Aiming at the real-time alignment needs at the sentence or paragraph level in Business English, the word alignment algorithm based on semantic similarity calculation is proposed, and the evaluation indexes of English-Chinese EC, Chinese-English CE unidirectional alignment and bi-directional fusion alignment, and semantic extension alignment are compared. Adding traditional business English translation techniques, comparing the effects of conventional translation techniques on business English sentence and paragraph translation versus parallel corpus-based translation techniques. The dynamic chunking accuracy of two-way fusion alignment and semantic expansion alignment is 94.57% and 94.32%, respectively, and the accuracy of semantic expansion alignment is 6.85% and 8.17% higher than that of EC one-way alignment. Meanwhile, the number of errors in business English translation works based on parallel corpora is reduced, which can help translators overcome the interference of Chinese chapter structures and improve the quality of their translation works.

Decoupled Cross-Modal Phrase-Attention Network for Image-Sentence Matching.

The mainstream of image and sentence matching studies currently focuses on fine-grained alignment of image regions and sentence words. However, these methods miss a crucial fact: the correspondence between images and sentences does not simply come from alignments between individual regions and words but from alignments between the phrases they form respectively. In this work, we propose a novel Decoupled Cross-modal Phrase-Attention network (DCPA) for image-sentence matching by modeling the relationships between textual phrases and visual phrases. Furthermore, we design a novel decoupled manner for training and inferencing, which is able to release the trade-off for bi-directional retrieval, where image-to-sentence matching is executed in textual semantic space and sentence-to-image matching is executed in visual semantic space. Extensive experimental results on Flickr30K and MS-COCO demonstrate that the proposed method outperforms state-of-the-art methods by a large margin, and can compete with some methods introducing external knowledge.

A Study of Computer-Assisted Automatic Translation Techniques and Practices for English Vocational Terminology

Abstract This paper mainly carries out a more in-depth study on the design and planning of computer-assisted translation models based on translation memory and demonstrates the new computer-assisted translation model formed in this way. The TMX specification model, combined with the actual needs, summarizes the corresponding terminology data model, introduces and summarizes the word splitting, similarity, and alignment techniques of the computer-assisted translation model, and verifies the effectiveness of the model through practical application. The study shows that the average code length of computer-aided translation is about 0.1~0.3 lower than that of human translation in specialized fields, and its performance and efficiency are better than that of human translation. The efficiency of computer-aided translation software in extracting terminology positively correlates with an increase in frequency. Computer-assisted translation plays a significant role in the professional field and provides empirical support for subsequent related research.

Exploration and Research on Chinese Culture in University English Translation Teaching in the Background of Internet+

Abstract Under the background of Internet+, for the problems of high comprehension difficulty and inaccurate translation in university English translation teaching, this paper proposes the skeleton knowledge translation model to assist university English translation teaching that integrates Chinese culture. Starting from the objectives of university English translation teaching, Chinese culture is explored in university English translation materials to complete the design of university English translation teaching that integrates Chinese culture. Propose the application of Internet technology in the above teaching design, utilize the IBM word alignment model and TF-IDF similarity to preprocess the university English translation textbooks, and on the basis of preprocessing, estimate the probability and decode the generated phrase pairs, and then complete the construction of the English translation model based on skeleton knowledge. The English translation model’s performance analysis is carried out using skeleton knowledge after determining the data set and evaluation criteria. The results show that for the whole 400 000 batch training, the training time of the English translation model based on skeleton knowledge is about one-fifth less than that of the unsupervised neural machine translation model, indicating that with the increase of the number of rounds, the performance of the English translation model based on skeleton knowledge is also better. This study serves as a guide for exploring and researching Chinese culture in university English translation teaching and promotes the development of university English translation teaching in colleges and universities.

Intelligent Translation System Aiding High-Quality Writing in English in the Age of the Internet

Abstract This study focuses on applying intelligent translation systems in assisting high-quality English writing in the Internet era. The study analyzes statistical machine translation techniques, especially N-gram language modeling and word alignment techniques, and their crucial role in translation quality improvement. In English writing, the intelligent translation system significantly improves the quality of students’ Writing through word block translation. By analyzing 320 students, the mean value of self-efficacy in writing skills of students in the high-quality writing group was 3.78, significantly higher than that of the low-quality group, which was 2.82. After the experiment, 0.57 students indicated that they would improve their English writing vocabulary with the aid of the Intelligent Translation System, which showed the potential of the Intelligent Translation System to enhance students’ interest in and autonomy in Writing. Average distribution analysis shows that word block usage positively correlates with writing performance, with an R Square value of 0.6726. The intelligent translation system improves students’ English writing and enhances their self-efficacy, which is of great significance to English teaching.

Exploring Translator Style Using Word Alignments

How the source text is rendered in the target language reflects a translator’s linguistics choices, which is informative of the translator’s style. Leveraging computational techniques, the present research seeks to explore translator style through word alignments derived from an entire corpus. The text material is two Chinese translations of Virginia Woolf’s novel, Jacob’s Room, by translators Pu Long and Wang Jiaxiang. Using a Transformer-based model, alignments are automatically extracted from parallel texts. A Support Vector Machine classifier is trained to test whether the alignments are indicative of the translators’ styles. Chi-square feature selection is then performed to identify the most distinguishing alignments for closer examination. Results indicate that Wang favours more explicit and literal translations, while Pu utilizes more concise, diverse, and idiomatic expressions. In addition, Wang’s translation is closer to the original, while Pu’s is more distant. This method enables us to provide a wide range of qualitative and quantitative evidence, and also observe differences not readily apparent when examining the target text alone.

Navigating the Environmental Education and Climate Change Curriculum in Türki̇ye through the Prism of Climate Change Literate Citizenship

In the present study a theoretical framework (i.e., Climate Change Literate Citizenship) that combines climate change literacy and climate citizenship was developed first. Then, criteria that correspond to each of the components of the proposed theoretical framework were developed and used to analyze the learning objectives in the “Environmental Education and Climate Change” curriculum. Data analyses utilized in the study presented a snapshot of the descriptive landscape of the curriculum in terms of the distribution of the objectives within the curriculum. Manifest and latent analyses realized through MAXQDA 2022 software were used to illustrate world cloud and word trends and alignment of the curriculum objectives with respect to the criteria developed for the Climate Change Literate Citizenship framework. Findings showed that there was an uneven distribution of the learning objectives within the Climate Change Literate Citizenship framework. Most of the learning objectives aimed to support climate change literacy, where objectives pertaining to the promotion of climate citizenship were relatively less in number. A similar uneven distribution was observed within the components of climate change literacy (i.e., functional, cultural, critical) and climate citizenship (i.e., personally responsible, participatory, justice oriented) as well. That is, there was not any objectives related to the cultural component of climate change literacy; most of the objectives aimed to develop functional component followed by the critical component. Among the components of climate citizenship, personal responsibility component and participatory component were addressed equally but there was only one learning objective that is in line with justice-oriented climate citizenship. Thus, our findings as a whole suggest that the curriculum adopts a fragmented rather than a holistic approach to climate change education. Nonetheless, further studies are required to make more robust conclusions about the curriculum and its practice.

A fuzzy model for NMT word alignment using quasi-perfect matching

A fuzzy model for NMT word alignment using quasi-perfect matching

Analysis of Fuzzy Semantic Translation in English Language and Literature Based on Bayesian Nonparametric Model

Abstract The development of machine learning provides a new path for translating fuzzy semantics into English and literature. In this paper, we propose translation points for fuzzy semantics based on the fuzziness of language by studying the linguistic features in English literature. The hierarchical PY process under the Bayesian nonparametric model is adopted to achieve word alignment in English translation, and the clustering process of English words is controlled by introducing discount and intensity parameters. Based on this, the parameter estimation of the word probability model is completed using Gibbs sampling, and the Moses decoder extracts the translation features. Based on monolingual word classification, this model reduces 6.85% in the average time consumed per sentence compared to the Beam algorithm and 9.57% compared to the A* algorithm.

GERNERMED++: Semantic annotation in German medical NLP through transfer-learning, translation and word alignment

We present a statistical model, GERNERMED++, for German medical natural language processing trained for named entity recognition (NER) as an open, publicly available model. We demonstrate the effectiveness of combining multiple techniques in order to achieve strong results in entity recognition performance by the means of transfer-learning on pre-trained deep language models (LM), word-alignment and neural machine translation, outperforming a pre-existing baseline model on several datasets. Due to the sparse situation of open, public medical entity recognition models for German texts, this work offers benefits to the German research community on medical NLP as a baseline model. The work serves as a refined successor to our first GERNERMED model. Similar to our previous work, our trained model is publicly available to other researchers. The sample code and the statistical model is available at: https://github.com/frankkramer-lab/GERNERMED-pp.

WAD-X: Improving Zero-shot Cross-lingual Transfer via Adapter-based Word Alignment

Multilingual pre-trained language models (mPLMs) have achieved remarkable performance on zero-shot cross-lingual transfer learning. However, most mPLMs implicitly encourage cross-lingual alignment in pre-training stage, making it hard to capture accurate word alignment across languages. In this paper, we propose Word-align ADapters for Cross-lingual transfer (WAD-X) to explicitly align word representations of mPLMs via language-specific subspace. Taking a mPLM as the backbone model, WAD-X constructs subspace for each source-target language pair via adapters. The adapters use statistical alignment as the prior knowledge to guide word-level aligning in the corresponding bilingual semantic subspace. We evaluate our model across a set of target languages on three zero-shot cross-lingual transfer tasks: part-of-speech tagging (POS), dependency parsing (DP), and sentiment analysis (SA). Experimental results demonstrate that our proposed model improves zero-shot cross-lingual transfer on three benchmarks, with improvements of 2.19, 2.50, and 1.61 points in POS, DP, and SA tasks over strong baselines.

Word Alignment of Chinese Poetry Parallel Corpus based on Word Embedding Technology

The research of word alignment technology provides the basis for many fields in natural language processing, such as speech recognition, bilingual dictionary writing, corpus construction and information retrieval. Based on the existing research foundation of parallel corpus for poetry sentence alignment, this paper studies how to align English-Chinese words. Firstly, the parallel corpus of Chinese poetry is preprocessed and translated into words, and the English poetry word vector is trained based on word embedding method, and the optimal corresponding word is judged by similarity calculation, and finally getting word alignment from parallel corpus sentence alignment.