Translation Pairs Research Articles

Enhancing bilingual lexicon induction via harnessing polysemous words

Bilingual Lexicon Induction (BLI) aims to induce word translation pairs across different languages. We find there is a large proportion of polysemous words in the dataset used by BLI, which impacts the quality of the induced translation pairs. In particular, the mainstream BLI models map two monolingual word embeddings (WEs) into a shared space and regard the closest cross-lingual words as translation pairs. However, when a word has multiple translations, it becomes challenging to ensure that the polysemous word is the closest neighbor to all its translations. Based on the analysis of polysemy, we propose a simple yet effective method by harnessing polysemous words to enhance BLI. We first detect polysemous words in the seed lexicons and filter their multiple translations by comparing their semantic similarities. Therefore, a refined seed lexicon mitigates the confusion caused by translations with divergent semantics for the BLI model. Moreover, in scenarios where only single translations are available in the seed lexicon, we propose to employ a Large Language Model (LLM) to acquire monolingual synonyms, thereby refining the structure of the seed lexicon more effectively. On benchmark BLI datasets, our method demonstrates superior performance compared to five state-of-the-art (SOTA) baseline systems across almost all experimental language pairs, encompassing both sides of three similar language pairs and three distant language pairs.

Bilingual phrase induction with local hard negative sampling

AbstractBilingual lexicon induction focuses on learning word translation pairs, also known as bitexts, from monolingual corpora by establishing a mapping between the source and target embedding spaces. Despite recent advancements, bilingual lexicon induction is limited to inducing bitexts consisting of individual words, lacking the ability to handle semantics‐rich phrases. To bridge this gap and support downstream cross‐lingual tasks, it is practical to develop a method for bilingual phrase induction that extracts bilingual phrase pairs from monolingual corpora without relying on cross‐lingual knowledge. In this paper, the authors propose a novel phrase embedding training method based on the skip‐gram structure. Specifically, a local hard negative sampling strategy that utilises negative samples of central tokens in sliding windows to enhance phrase embedding learning is introduced. The proposed method achieves competitive or superior performance compared to baseline approaches, with exceptional results recorded for distant languages. Additionally, we develop a phrase representation learning method that leverages multilingual pre‐trained language models. These mPLMs‐based representations can be combined with the above‐mentioned static phrase embeddings to further improve the accuracy of the bilingual phrase induction task. We manually construct a dataset of bilingual phrase pairs and integrate it with MUSE to facilitate the bilingual phrase induction task.

Do French speakers have an advantage in learning English vocabulary thanks to familiar suffixes?

Previous research has shown that languages from nearby families are easier to learn as second languages (L2) than languages from more distant families, attributing this difference to the presence of shared elements between the native language (L1) and L2. Building on this idea, we hypothesised that suffixes present in L1 might facilitate complex word acquisition in L2. To test this hypothesis, we recruited 76 late French-English bilinguals and tasked them with learning a set of 80 English-derived words containing suffixes that also exist in French (e.g., -able) or are unique to English (e.g., -ness). Consolidation of the learned words was assessed 1 week after the last learning session. The results showed a significant learning effect across the learning trials and consolidation, suggesting that the bilingual participants were able to acquire the derived words. However, contrary to our hypothesis, suffixes also existing in French did not give a significant advantage over English-unique suffixes. Further analysis revealed that this was due to variations in the consistency of familiar suffixes from L1. While some translation pairs shared the same suffix (e.g., amazement-étonnement), others had different suffixes (e.g., slippage-glissement). The type of translation pair with inconsistent suffix overlap (slippage-glissement) carried learning costs, preventing the bilingual participants from benefitting from the presence of familiar suffixes in L2 words. These findings suggest that shared information can be used effectively for L2 learning only if the mapping between L1 and L2 is consistent.

Enhancing Bilingual Lexicon Induction via Bi-directional Translation Pair Retrieving

Most Bilingual Lexicon Induction (BLI) methods retrieve word translation pairs by finding the closest target word for a given source word based on cross-lingual word embeddings (WEs). However, we find that solely retrieving translation from the source-to-target perspective leads to some false positive translation pairs, which significantly harm the precision of BLI. To address this problem, we propose a novel and effective method to improve translation pair retrieval in cross-lingual WEs. Specifically, we consider both source-side and target-side perspectives throughout the retrieval process to alleviate false positive word pairings that emanate from a single perspective. On a benchmark dataset of BLI, our proposed method achieves competitive performance compared to existing state-of-the-art (SOTA) methods. It demonstrates effectiveness and robustness across six experimental languages, including similar language pairs and distant language pairs, under both supervised and unsupervised settings.

Visual word recognition of Chinese-Japanese bilinguals: limited role of phonology.

The investigation of how orthography and phonology influence lexical semantic access in visual word identification is a crucial area in psycholinguistics. Previous studies, focusing on alphabetic scripts in bilingual lexical recognition, have highlighted the facilitative role of phonological similarity. Yet, the impact of cross-language phonological similarity in bilinguals using non-alphabetic scripts remains unclear. In this study, we employed a lexical decision task to examine Chinese-Japanese bilinguals. Participants were presented with Chinese-Japanese cognate translation pairs, categorized into phonologically similar and dissimilar cognates. Analysis of event-related potentials (ERP) revealed no significant differences between phonologically similar and dissimilar contexts in the early time windows (90-170 ms and 170-270 ms). However, in the later time window (350-500 ms), significant differences were observed, with the phonologically dissimilar condition eliciting a larger negative wave. Contrary to findings in alphabetic script-based studies, our results suggest that in logographic script processing, the activation of phonology and semantics occurs simultaneously, and the influence of phonology is limited. This indicates a distinct cognitive processing mechanism in non-alphabetic language bilinguals, providing new insights into the dynamics of bilingual lexical recognition.

Automatically generated language learning exercises for Finno-Ugric languages

Morphologically rich languages always constitute a great challenge for language learners. The learner must be able to understand the information encoded in different word forms of the same root and to generate the correct word form to express certain syntactic functions and grammatical relations by conjugating a verb or declining a noun, an adjective or a pronoun. One way to improve one’s language skills is through exercises that focus on certain aspects of grammar. In this paper, a language learning application is presented that is intended to help learners of Finnish and Hungarian (with Hungarian and Finnish L1, respectively) acquire new vocabulary items, as well as practice some grammar aspects that according to surveys are considered difficult by learners of these languages with the other Finno-Ugric language being the learner’s native tongue, while alleviating the need to create these exercises manually. This application is a result of an on-going research project. In this research project, bilingual translation pairs and additional monolingual data were collected that can be utilized to build language learning exercises and an online bilingual dictionary with the help of automatic methods. Several linguistic patterns and rules were defined in order to automatically select example sentences that focus on a given part of the target language. These sentences were automatically annotated with the help of language processing tools. Due to the large size of the previously collected data sets, to date, only a subset of the analyzed sentences and the bilingual translation pairs has been manually evaluated. The results of this evaluation are discussed in this paper in order to estimate the precision of the methodology presented here. To ensure the precision of the information and the reliability of the application, only manually validated data sets are displayed. In this project, continuous data validation is planned, since it leads to more and more examples and vocabulary items that learners can benefit from.

Contrastive Adversarial Training for Multi-Modal Machine Translation

The multi-modal machine translation task is to improve translation quality with the help of additional visual input. It is expected to disambiguate or complement semantics while there are ambiguous words or incomplete expressions in the sentences. Existing methods have tried many ways to fuse visual information into text representations. However, only a minority of sentences need extra visual information as complementary. Without guidance, models tend to learn text-only translation from the major well-aligned translation pairs. In this article, we propose a contrastive adversarial training approach to enhance visual participation in semantic representation learning. By contrasting multi-modal input with the adversarial samples, the model learns to identify the most informed sample that is coupled with a congruent image and several visual objects extracted from it. This approach can prevent the visual information from being ignored and further fuse cross-modal information. We examine our method in three multi-modal language pairs. Experimental results show that our model is capable of improving translation accuracy. Further analysis shows that our model is more sensitive to visual information.

To say almost the same thing? A study on cross-linguistic variation in ancient texts and their translations

Abstract This article presents a study of several parallel corpora of historical languages and their translations. The aligned corpora are the result of a large crowdsourcing project, named Ugarit, aimed at supporting translation alignment for ancient and historical languages: the study of the resulting translation pairs allows us to observe cross-linguistic dynamics in a range of languages, some of which have never been systematically aligned before. The corpora considered are divided into two distinct groups: English translations of ancient languages, including Greek, Latin, Persian, and Coptic; and translations of ancient Greek into other languages, including Latin, English, Georgian, Italian, and Persian. We evaluated different ratios of word matching across each language pair (one-to-one, one-to-many, many-to-one, and many-to-many), and analyzed the resulting trends across the corpus. We propose some observations on how and why different types of alignment links are established in a given language pair, and what factors affect their creation beyond the control of the user: we propose two complementary hypotheses to explain the changes, one based on structural linguistic factors and the other based on cultural difference.

Are translation equivalents really equivalent? Evidence from concreteness effects in translation priming

Aims and Objectives: Translation equivalents intuitively seem to overlap in meaning. Moreover, the models of the bilingual lexicon often represent the meaning shared between two translations as a holistic node in the semantic network. However, research on semantic representation and processing questions this holistic approach. For instance, abstract words are assumed to be more language-dependent, while concrete words’ meanings are seen as more consistent cross-linguistically. The non-cognate translation priming paradigm offers an ideal methodological setting to study semantic overlap (proxied by concreteness) between translations. Priming effects between non-cognate translation equivalents are assumed to emerge due to spreading activation at the semantic level. Hence, a larger semantic overlap between translation prime-target pairs should lead to larger priming effects. Nevertheless, the evidence from previous translation priming studies investigating concreteness displays a blurry picture, potentially reflecting a shared limitation: their relatively small sample sizes. We overcame this problem by analysing the largest translation priming dataset to date. Methodology: Two hundred Spanish–English highly proficient bilinguals were tested in a bidirectional translation priming experiment employing 314 non-cognate translation equivalents differing in concreteness. Data and analysis: We analysed response times and error rates employing conservative (generalized) linear mixed-effects models. Findings: The results showed that concrete translation pairs elicited larger priming effects than abstract ones, evidencing differences in semantic representation between concrete and abstract words. Importantly, the influence of concreteness appeared only in the forward translation direction, suggesting language experience-related differences in meaning representation. Originality: The present study analysed the largest dataset in the translation priming literature to date, employing a conservative statistical approach to shed light on the effects of concreteness on translation priming. Implications: Our study spotlights the complexity and non-holistic nature of the bilingual semantic representation of concrete and abstract words. The present findings call for more research to help the current models of the bilingual lexicon implement more nuanced semantic representations.

Design and Analysis of a Novel Compliant Mechanism with RPR Degrees of Freedom

A novel compliant mechanism with RPR degrees of freedom (DOF) is proposed where R and P represent rotation and translation DOFs, respectively. The proposed compliant mechanism is obtained from dimension synthesizing a 2-RPU-UPR rigid parallel mechanism with the method of optimization of motion/force transfer characteristic. R, P and U represent rotation, translation and universal pairs, respectively. Firstly, inverse kinematics and Jacobian matrix are analyzed for the dimensional synthesis. Then, output transmission indexes of branches in the parallel mechanism are given. Dimensional synthesis is completed based on the normalized design parameter. And optimization of flexure joints based on constrained energy is carried out. Afterwards, the novel compliant mechanism is obtained by direct replacing method. Mechanical model of the compliant mechanism including static stiffness and input stiffness is built based on the pseudo-rigid body modeling method and virtual work principle. Finally, FEA simulation by Ansys Workbench is carried out to verify DOF, effectiveness of the dimension synthesis, and compliant model. Optimization of motion/force transfer characteristic is first applied for the design of compliant mechanisms to suppress drift of rotation axis in the paper.

Unsupervised English Intelligent Machine Translation in Wireless Network Environment

Researchers suggest unsupervised English machine translation to address the absence of parallel corpus in English translation. Unsupervised pretraining techniques, denoising autoencoders, back translation, and shared latent representation mechanisms are used to simulate the translation task using just monolingual corpora. This paper uses pseudo-parallel data to construct unsupervised neural machine translation (NMT) and dissimilar language pair analysis. This paper firstly analyzes the low performance of unsupervised translation on dissimilar language pairs from three aspects: bilingual word embedding quality, shared words, and word order. And artificial shared word replacement and preordering strategies are proposed to increase the shared words between dissimilar language pairs and reduce the difference in their syntactic structure, thereby improving the translation performance on dissimilar language pairs. The denoising autoencoder and shared latent representation mechanism in unsupervised English machine translation are only required in the early stage of training, and learning the shared latent representation limits the further improvement of performance in different directions. While training the denoising autoencoder by repeatedly altering the training data slows down the convergence of the model, this is especially true for divergent languages. This paper presents an unsupervised NMT model based on pseudo-parallel data to address this issue. It trains two standard supervised neural machine translation models using the pseudo-parallel corpus generated by the unsupervised neural machine translation system, which enhances translation performance and speeds convergence. Finally, the English intelligent translation model is deployed in the wireless network server, and users can access it through the wireless network.

Leveraging Multilingual News Websites for Building a Kurdish Parallel Corpus

Machine translation has been a major motivation of development in natural language processing. Despite the burgeoning achievements in creating more efficient machine translation systems, thanks to deep learning methods, parallel corpora have remained indispensable for progress in the field. In an attempt to create parallel corpora for the Kurdish language, in this article, we describe our approach in retrieving potentially alignable news articles from multi-language websites and manually align them across dialects and languages based on lexical similarity and transliteration of scripts. We present a corpus containing 12,327 translation pairs in the two major dialects of Kurdish, Sorani and Kurmanji. We also provide 1,797 and 650 translation pairs in English-Kurmanji and English-Sorani. The corpus is publicly available under the CC BY-NC-SA 4.0 license. 1

Image Visual Attention Mechanism-based Global and Local Semantic Information Fusion for Multi-modal English Machine Translation

<p>Machine translation is a hot research topic at present. Traditional machine translation methods are not effective because they require a large number of training samples. Image visual semantic information can improve the effect of the text machine translation model. Most of the existing works fuse the whole image visual semantic information into the translation model, but the image may contain different semantic objects. These different local semantic objects have different effects on the words prediction of the decoder. Therefore, this paper proposes a multi-modal machine translation model based on the image visual attention mechanism via global and local semantic information fusion. The global semantic information in the image and the local semantic information are fused into the text attention weight as the image attention. Thus, the alignment information between the hidden state of the decoder and the text of the source language is further enhanced. Experimental results on the English-German translation pair and the Indonesian-Chinese translation pair on the Multi30K dataset show that the proposed model has a better performance than the state-of-the-art multi-modal machine translation models, the BLEU values of English-German translation results and Indonesian-Chinese translation results exceed 43% and 29%, which proves the effectiveness of the proposed model.</p> <p> </p>

The L1 Semantic Retrieval of L2 Words: Evidence from Advanced L2 Learners’ Reaction Times

According to the first language (L1) lemma mediation hypothesis, second language learners, regardless of their level of second language (L2) proficiency, access the meaning of L2 words via their first language (Jiang, 2004). To test this hypothesis, a semantic judgment task was conducted on 30 advanced Arab speakers of English, in which they were presented with 86 pairs of English words and had to decide whether each pair was semantically related. Some semantically related pairs are classified as same translation pairs because their members share the same L1 translation, whereas others are semantically related but do not share the same L1 translation, hence they are classified as different translation pairs. Two instruments were used to record the reaction times and determine accuracy: DMDX and Gorilla. The results revealed that the highly proficient L2 speakers rated same translation pairs as semantically related significantly faster than their responses to different translation pairs. When compared with the 28 native speakers’ results, there was a significant difference in the reaction times of the two groups. This provides evidence that the underlying processes of L1 and L2 vocabulary acquisition is substantially different: L2 learners rely on their well-established conceptual system to access the meaning of L2 words.

Translation Alignment with Ugarit

Ugarit is a public web-based tool for manual annotation of parallel texts for generating word-level translation alignment. We aimed to develop a user-friendly interactive interface to visualize aligned texts and collect training data in the form of translation pairs to be used later, (i) for training an automatic translation alignment system for historical languages at the word/phrase level, (ii) as a gold standard to evaluate automatic alignment and machine translation systems. Ugarit is now widely used for learning new languages, especially historical languages, and as a reading environment for parallel texts. In the following sections, we present the related works and similar projects; then, we give an overview of the visualization techniques used to present the alignment results. Further, we explain how we could derive the translation graph from the aligned translation pairs. Finally, we discuss the usage limitations of Ugarit, possible improvements, and future development plans.

Making the most of comparable corpora in Neural Machine Translation: a case study

Comparable corpora can benefit the development of Neural Machine Translation models, in particular for under-resourced languages. We present a case study centred on the exploitation of a large comparable corpus for Basque-Spanish, created from independently-produced news by the Basque public broadcaster eitb, where we evaluate the impact of different techniques to exploit the original data, in order to complement parallel datasets for this language pair in both translation directions. Two efficient methods for parallel sentence mining are explored, which identified a common core of approximately half of the total number of aligned sentences, each one uniquely identifying valid parallel sentences not captured by the other method. Filtering the data via identification of length-difference outliers proved highly effective to improve the models, as was the use of tags to discriminate between comparable and parallel data in the training corpora. The use of backtranslated data is also evaluated in this work, with results indicating that alignment-based datasets remain the most beneficial, although complementary backtranslations should also be included to fully exploit the available comparable data. Overall, the results in this work demonstrate that this type of data needs to be carefully analysed prior to its use as training data for Neural Machine Translation, since issues such as information imbalance between source and target data can lead to unoptimal results for a given translation pair.

English Machine Translation Model Based on an Improved Self-Attention Technology

English machine translation is a natural language processing research direction that has important scientific research value and practical value in the current artificial intelligence boom. The variability of language, the limited ability to express semantic information, and the lack of parallel corpus resources all limit the usefulness and popularity of English machine translation in practical applications. The self-attention mechanism has received a lot of attention in English machine translation tasks because of its highly parallelizable computing ability, which reduces the model’s training time and allows it to capture the semantic relevance of all words in the context. The efficiency of the self-attention mechanism, however, differs from that of recurrent neural networks because it ignores the position and structure information between context words. The English machine translation model based on the self-attention mechanism uses sine and cosine position coding to represent the absolute position information of words in order to enable the model to use position information between words. This method, on the other hand, can reflect relative distance but does not provide directionality. As a result, a new model of English machine translation is proposed, which is based on the logarithmic position representation method and the self-attention mechanism. This model retains the distance and directional information between words, as well as the efficiency of the self-attention mechanism. Experiments show that the nonstrict phrase extraction method can effectively extract phrase translation pairs from the n-best word alignment results and that the extraction constraint strategy can improve translation quality even further. Nonstrict phrase extraction methods and n-best alignment results can significantly improve the quality of translation translations when compared to traditional phrase extraction methods based on single alignment.

Natural Language Processing and Computational Linguistics

Natural Language Processing and Computational Linguistics

We Don't Speak the Same Language: Interpreting Polarization through Machine Translation

Polarization among US political parties, media and elites is a widely studied topic. Prominent lines of prior research across multiple disciplines have observed and analyzed growing polarization in social media. In this paper, we present a new methodology that offers a fresh perspective on interpreting polarization through the lens of machine translation. With a novel proposition that two sub-communities are speaking in two different "languages", we demonstrate that modern machine translation methods can provide a simple yet powerful and interpretable framework to understand the differences between two (or more) large-scale social media discussion data sets at the granularity of words. Via a substantial corpus of 86.6 million comments by 6.5 million users on over 200,000 news videos hosted by YouTube channels of four prominent US news networks, we demonstrate that simple word-level and phrase-level translation pairs can reveal deep insights into the current political divide -- what is "black lives matter" to one can be "all lives matter" to the other.

Napisy filmowe jako źródło ekwiwalentów przekładowych w polsko-czeskim korpusie równoległym InterCorp

The article focuses on using film subtitles in the process of establishing Polish-Czech language pairs. Research material comes from the resources of the OpenSubtitles website with non-professional subtitling. The research material was analysed using KonText – the main search tool in InterCorp. As a starting point, the author analysed a dictionary entry for the Polish exclamation cholera in a traditional Polish translation dictionary like Polish-Czech Dictionary (Oliva). It presents translation equivalents in translation pairs such as cholera – sakra, hergot, do cholery – do hajzlu, idź do cholery – táhni ke všem čertům, do jasnej cholery – k sakru, do hajzlu, cholera mnie bierze – čerti mě berou. All the translation equivalent pairs have been analysed using the corpus analysis with three functionalities available in the KonText search engine: type of question (typ dotazu), content of the equivalent (obsahuje), and limiting search (omezit hledání). The conducted corpus analysis allowed for the verification of information in the traditional dictionary, but also to find newequivalents that have not been registered yet which could also be used in translation.