Distant 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.

A Benchmark Evaluation of Multilingual Large Language Models for Arabic Cross-Lingual Named-Entity Recognition

Multilingual large language models (MLLMs) have demonstrated remarkable performance across a wide range of cross-lingual Natural Language Processing (NLP) tasks. The emergence of MLLMs made it possible to achieve knowledge transfer from high-resource to low-resource languages. Several MLLMs have been released for cross-lingual transfer tasks. However, no systematic evaluation comparing all models for Arabic cross-lingual Named-Entity Recognition (NER) is available. This paper presents a benchmark evaluation to empirically investigate the performance of the state-of-the-art multilingual large language models for Arabic cross-lingual NER. Furthermore, we investigated the performance of different MLLMs adaptation methods to better model the Arabic language. An error analysis of the different adaptation methods is presented. Our experimental results indicate that GigaBERT outperforms other models for Arabic cross-lingual NER, while language-adaptive pre-training (LAPT) proves to be the most effective adaptation method across all datasets. Our findings highlight the importance of incorporating language-specific knowledge to enhance the performance in distant language pairs like English and Arabic.

Hyphal connection increases net resource utilization of saprotrophic fungal mycelia by improving local performance: A case study of the cord-forming basidiomycete, Phanerochaete velutina

To identify the functional advantages of hyphal connections in fungal mycelial networks, we evaluated the wood decaying abilities of the mycelia of a saprotrophic basidiomycete, Phanerochaete velutina, in soil dishes with wood baits. The dishes were surrounded by water, allowing the mycelia to access water. Dishes with two or three baits were placed adjacent to or distant from each other, representing the ease of creating hyphal bridges between the two dishes. After 8 months of incubation, the weight loss of the wood bait was significantly larger in the adjacent pairs of dishes compared to the distant pairs. Mycelia in the adjacent dishes had higher chances to access the water and thus increase water content of the wood, which was the primary factor related to wood weight loss. These results suggest that hyphal connections increase decay ability of P. velutina possibly due to water transfer across the mycelial network.

Unsupervised Multimodal Machine Translation for Low-resource Distant Language Pairs

Unsupervised machine translation (UMT) has recently attracted more attention from researchers, enabling models to translate when languages lack parallel corpora. However, the current works mainly consider close language pairs (e.g., English-German and English-French), and the effectiveness of visual content for distant language pairs has yet to be investigated. This article proposes an unsupervised multimodal machine translation model for low-resource distant language pairs. Specifically, we first employ adequate measures such as transliteration and re-ordering to bring distant language pairs closer together. We then use visual content to extend masked language modeling and generate visual masked language modeling for UMT. Finally, empirical experiments are conducted on our distant language pair dataset and the public Multi30k dataset. Experimental results demonstrate the superior performance of our model, with BLEU score improvements of 2.5 and 2.6 on translation for distant language pairs English-Uyghur and Chinese-Uyghur. Moreover, our model also brings remarkable results for close language pairs, improving 2.3 BLEU compared with the existing models in English-German.

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.

Improving linear orthogonal mapping based cross-lingual representation using ridge regression and graph centrality

Orthogonal linear mapping is a commonly used approach for generating cross-lingual embedding between two monolingual corpora that uses a word frequency-based seed dictionary alignment approach. While this approach is found to be effective for isomorphic language pairs, they do not perform well for distant language pairs with different sentence structures and morphological properties. For a distance language pair, the existing frequency-aligned orthogonal mapping methods suffer from two problems - (i)the frequency of source and target word are not comparable, and (ii)different word pairs in the seed dictionary may have different contribution. Motivated by the above two concerns, this paper proposes a novel centrality-aligned ridge regression-based orthogonal mapping. The proposed method uses centrality-based alignment for seed dictionary selection and ridge regression framework for incorporating influential weights of different word pairs in the seed dictionary. From various experimental observations over five language pairs (both isomorphic and distant languages), it is evident that the proposed method outperforms baseline methods in the Bilingual Dictionary Induction(BDI) task, Sentence Retrieval Task(SRT), and Machine Translation. Further, several analyses are also included to support the proposed method.

Handling syntactic difference in Chinese-Vietnamese neural machine translation

As the typical distant language pair, Chinese and Vietnamese vary widely in syntactic structure, which significantly influences the performance of Chinese-Vietnamese machine translation. To address this problem, we present a simple approach with a pre-reordering model for closing syntactic gaps of the Chinese-Vietnamese language pair. Specifically, we first propose an algorithm for recognizing the modifier inverse, one of the most representative syntactic different in Chinese-Vietnamese language pair. Then we pre-train a pre-reordering model based on the former recognition algorithm and incorporate it into the attention-based translation framework for syntactic different reordering. We conduct empirical studies on Chinese-Vietnamese neural machine translation task, the results show that our approach achieves average improvement of 2.75 BLEU points in translation quality over the baseline model. In addition, the translation fluency can be significantly improved by over 2.44 RIBES points.

Superconducting state generated dynamically from distant pair source and drain

It has been well established that the origin of p-wave superconductivity is the balance between pair creation and annihilation, described by the spin-less fermionic Kitaev chain model. In this work, we study the dynamics of a composite system where the pair source and drain are spatially separated by a long distance. We show that this non-Hermitian system possesses a high-order exceptional point (EP) when only a source or drain is considered. The EP dynamics provide a clear picture: A pair source can fully fill the system with pairs, while a drain can completely empty the system. When the two coexist simultaneously, the dynamics depend on the distance and the relative phase between the pair creation and annihilation terms. Analytical analysis and numerical simulation results show that the superconducting state can be dynamically established at the resonant pair source and drain: from an initial empty state to a stationary state with the maximal pair order parameter. It provides an alternative way of understanding the mechanism of the nonequilibrium superconducting state.

Linguistic distance dynamically modulates the effects of bilingualism on executive performance in aging

AbstractTo better explain various neurocognitive consequences of bilingualism, recent investigations have adopted continuous measures of bilingual experience, as opposed to binary bi/monolingual distinctions. However, few studies have considered whether bilingualism's effects on cognition are modulated by the linguistic distance (LD) between L1 and L2, and none of the existing studies has examined cognitive consequences of LD in aging populations. Here, we investigated the modulatory role of LD on the relationship between bilingualism, executive performance, and cognitive reserve (CR) in a sample of senior bilinguals. Our results show a dynamic trajectory of LD effects, with more distant language pairs exerting maximum effects at initial stages of bilingual experience – and closer language pairs at advanced stages. Bilingualism-related CR effects emerged only in the individuals with closer language pairs, suggesting that the language control stage of bilingual experience may play a key role in CR accrual, as compared to the L2 learning stage.

Investigating Unsupervised Neural Machine Translation for Low-resource Language Pair English-Mizo via Lexically Enhanced Pre-trained Language Models

The vast majority of languages in the world at present are considered to be low-resource languages. Since the availability of large parallel data is crucial for the success of most modern machine translation approaches, improving machine translation for low-resource languages is a key challenge. Most unsupervised techniques for translation benefit closely related languages with monolingual data of substantial quantity. To facilitate research in this direction for the extremely low resource language pair English ( en ) and Mizo ( lus ), we have developed a parallel and monolingual corpus for the Mizo language from various news websites. We explore Unsupervised Neural Machine Translation (UNMT) based on the developed monolingual data. We observe that cross-lingual embedding (CLWE) initializations on subword segmented data during pre-training, based on both masked language modelling and sequence-to-sequence generation tasks, improve translation performance. We experiment with cross-lingual alignment and combined alignment and joint training for learning the cross-lingual embedding representations. We also report baseline performances and the impact of CLWE initialization using semi-supervised and supervised neural machine translation. Empirical results show that both CLWE initializations work well for the distant pair English-Mizo compared to the baselines.

Динамическое исследование кратной звезды ADS 9173 AB

Star ADS 9173=WDS 14135+5147=Hip 69483 is a complex system. The B component has a spectroscopic companion, whose orbit with a period of 4.9 years has been known since 1986. The Gaia telescope has detected a distant faint pair over 100" away from the bright AB pair. In our article, we study the movement in a bright pair based on long-term observations with the 26-inch refractor of the Pulkovo Observatory. The AB pair orbit with a period of 6306 years was calculated using the apparent motion parameters (AMP) method. The astrometric orbit of the component B was determined on the basis of the residuals of the homogeneous CCD observations up to 2023 with the 26-inch refractor. It is in agreement with the spectroscopic one. The remaining secondary residuals show a wave with a period of approximately 20 years, the reasons for which are discussed.

Nouns are not always processed faster than verbs in bilingual speakers: effects of language distance

AbstractThe purpose of the current report is to study the effects of language distance on noun and verb processing in bilingual speakers. We recruited two groups of bilingual speakers: one group spoke two typologically distant languages (Cantonese and English) and the other group spoke two typologically similar languages (Mandarin and Cantonese). Participants named object and action pictures in their first language. We controlled psycholinguistic properties of words such as frequency, AoA, imageability, name agreement, visual complexity, familiarity, and participants’ bilingual language experiences. Our findings revealed a significant role for language distance. We observed a difference between noun and verb processing in the similar language pair (Mandarin–Cantonese) due to interference induced by language similarity. However, in the distant language pair (Cantonese–English), the difference disappeared because of the lack of cross-language interference. Our findings support that current and future models of bilingual language processing should take into account the effects of language distance.

Self-supervised pretraining improves the performance of classification of task functional magnetic resonance imaging.

Decoding brain activities is one of the most popular topics in neuroscience in recent years. And deep learning has shown high performance in fMRI data classification and regression, but its requirement for large amounts of data conflicts with the high cost of acquiring fMRI data. In this study, we propose an end-to-end temporal contrastive self-supervised learning algorithm, which learns internal spatiotemporal patterns within fMRI and allows the model to transfer learning to datasets of small size. For a given fMRI signal, we segmented it into three sections: the beginning, middle, and end. We then utilized contrastive learning by taking the end-middle (i.e., neighboring) pair as the positive pair, and the beginning-end (i.e., distant) pair as the negative pair. We pretrained the model on 5 out of 7 tasks from the Human Connectome Project (HCP) and applied it in a downstream classification of the remaining two tasks. The pretrained model converged on data from 12 subjects, while a randomly initialized model required 100 subjects. We then transferred the pretrained model to a dataset containing unpreprocessed whole-brain fMRI from 30 participants, achieving an accuracy of 80.2 ± 4.7%, while the randomly initialized model failed to converge. We further validated the model's performance on the Multiple Domain Task Dataset (MDTB), which contains fMRI data of 26 tasks from 24 participants. Thirteen tasks of fMRI were selected as inputs, and the results showed that the pre-trained model succeeded in classifying 11 of the 13 tasks. When using the 7 brain networks as input, variations of the performance were observed, with the visual network performed as well as whole brain inputs, while the limbic network almost failed in all 13 tasks. Our results demonstrated the potential of self-supervised learning for fMRI analysis with small datasets and unpreprocessed data, and for analysis of the correlation between regional fMRI activity and cognitive tasks.

Habitat suitability mapping of white stork Ciconia ciconia in one of its key European breeding areas

Due to rapid changes of the agricultural practice, the traditionally managed farmland became currently-one of the most endangered habitats in Europe. Species related to the rural landscape have been facing a great challenge of adaptation to quickly changing conditions. Therefore, recognizing and characterizing highly suitable areas for white storks is very important to plan conservation measures in the era of changes in agriculture and land use. Here, combining nest survey data (811 occupied nest), GIS-data on habitat availability and modelling approach, we characterize the current foraging conditions of the white stork Ciconia ciconia, an indicator of the agricultural habitat biodiversity and an umbrella species for organisms associated with agriculture landscape. The studied population breeds in the Warmińska Refuge Natura 2000 site (N Poland), one of the key breeding areas of this species in Europe. Breeding pair presence was predicted by the area of pasture, meadows, field ponds, and linear water within a 1 km buffer around the nest. The main hotspots of habitat suitability were located in the central and western parts of the study area, characterized by a more traditional model of agriculture. The distance to important foraging habitats does not influence white stork fecundity indicating that landscape composition is important when choosing the breeding site. The pairs with higher number of young were located further from other neighbouring pairs and at the areas with lower density of pairs in general. This is probably linked to the lower competition for food between more distant breeding pairs. Our results show that the traditional agricultural use in mosaics of pastures, meadows and aquatic habitats promotes white storks nesting.

Functional Nanocomposites in the Development of Flexible Armor

The idea of flexible body armor has been around for many years. Initial development included shear thickening fluid (STF) as a backbone polymer to impregnate ballistic fibers such as Kevlar. At the core of the ballistic and spike resistance was the instantaneous rise in viscosity of STF during impact. Increase in viscosity was due to the hydroclustering of silica nanoparticles dispersed in polyethylene glycol (PEG) through a centrifuge and evaporation process. When STF composite was dry, hydroclustering was not possible due to absence of any fluidity in PEG. However, particles embedded within the polymer, covered the Kevlar fiber and offered some resistance to spike and ballistic penetration. The resistance was meagre and hence, the goal was to improve it further. This was achieved by creating chemical bonds between particles, and by strongly attaching particles to the fiber. PEG was replaced with silane (3-amino propyl trimethoxysilane), and a fixative cross-linker, Glutaraldehyde (Gluta), was added. Silane installed an amine functional group on the silica nanoparticle surface, and Gluta created strong bridges between distant pairs of amine groups. Amide functional groups present in Kevlar also interacted with Gluta and silane to form a secondary amine, allowing silica particles to attach to fiber. A network of amine bonding was also established across the particle-polymer-fiber system. In synthesizing the armor, silica nanoparticles were dispersed in a mixture of silane, ethanol, water, and Gluta, maintaining an appropriate ratio by weight, and using a sonication technique. Ethanol was used as a dispersion fluid and was evaporated later. Several layers of Kevlar fabric were then soaked with the admixture for about 24 h and dried in an oven. Armor composites were tested in a drop tower according to NIJ115 Standard using spikes. Kinetic energy at impact was calculated and normalized with the aerial density of the armor. NIJ tests revealed that normalized energy for 0-layer penetration increased from 10 J-cm2/g (STF composite) to 220 J-cm2/g for the new armor composite, indicating a 22-fold enhancement. SEM and FTIR studies confirmed that this high resistance to spike penetration was due to the formation of stronger C-N, C-H, and C=C-H stretches facilitated by the presence of silane and Gluta.

In Silico Stage-Matching of Human, Marmoset, Mouse, and Pig Embryos to Enhance Organ Development Through Interspecies Chimerism

Currently, there is a significant shortage of transplantable organs for patients in need. Interspecies chimerism and blastocyst complementation are alternatives for generating transplantable human organs in host animals such as pigs to meet this shortage. While successful interspecies chimerism and organ generation have been observed between evolutionarily close species such as rat and mouse, barriers still exist for more distant species pairs such as human–mouse, marmoset–mouse, human–pig, and others. One of the proposed barriers to chimerism is the difference in developmental stages between the donor cells and the host embryo at the time the cells are introduced into the host embryo. Hence, there is a logical effort to stage-match the donor cells with the host embryos for enhancing interspecies chimerism. In this study, we used an in silico approach to simultaneously stage-match the early developing embryos of four species, including human, marmoset, mouse, and pig based on transcriptome similarities. We used an unsupervised clustering algorithm to simultaneously stage-match all four species as well as Spearman’s correlation analyses to stage-match pairs of donor–host species. From our stage-matching analyses, we found that the four stages that best matched with each other are the human blastocyst (E6/E7), the gastrulating mouse embryo (E6–E6.75), the marmoset late inner cell mass, and the pig late blastocyst. We further demonstrated that human pluripotent stem cells best matched with the mouse post-implantation stages. We also performed ontology analysis of the genes upregulated and commonly expressed between donor–host species pairs at their best matched stages. The stage-matching results predicted by this study will inform in vivo and in vitro interspecies chimerism and blastocyst complementation studies and can be used to match donor cells with host embryos between multiple species pairs to enhance chimerism for organogenesis.

Cluster-in-Molecule Method Combined with the Domain-Based Local Pair Natural Orbital Approach for Electron Correlation Calculations of Periodic Systems.

The cluster-in-molecule (CIM) method was extended to systems with periodic boundary conditions (PBCs) in a previous work (PBC-CIM) [J. Chem. Theory Comput.2019, 15, 2933], which is able to compute the electronic structures of periodic systems at second-order Møller-Plesset perturbation theory (MP2) and coupled cluster singles and doubles (CCSD) levels. However, the high computational costs of CCSD with respect to the size of clusters limit the usage of PBC-CIM to crystals with small or medium unit cells. In this work, we further develop the PBC-CIM method by employing the domain-based local pair natural orbital (DLPNO) methods for the electron correlation calculations of clusters to reduce the computational costs. The combined approach allows CCSD with perturbative triples, denoted as CCSD(T), to be computationally available for accurate descriptions of periodic systems. The distant-pair correction is also implemented to improve the accuracy of PBC-CIM. As in the molecular cases, the distant pair correction significantly improves the accuracy of various PBC-CIM methods with few additional costs. The PBC-CIM-DLPNO-CCSD(T) approach has been applied to investigate the optimized lattice parameter of the cubic LiCl crystal and two adsorption problems (CO on the NaCl(100) surface and H2O on the h-BN surface). The results show that the CIM-DLPNO-CCSD(T) method offers accurate and efficient descriptions for the studied systems. Another application to the cohesive energy of the acetic acid crystal reveals that large basis sets are necessary for reliable calculations on the cohesive energies of molecular crystals.

Pan-brassinosteroid signaling revealed by functional analysis of NILR1 in land plants.

Brassinosteroid (BR) signaling has been identified from the ligand BRs sensed by the receptor Brassinosteroid Insensitive 1 (BRI1) to the final activation of Brassinozole Resistant 1/bri1 EMS-Suppressor 1 through a series of transduction events. Extensive studies have been conducted to characterize the role of BR signaling in various biological processes. Our previous study has shown that Excess Microsporocytes 1 (EMS1) and BRI1 control different aspects of plant growth and development via conserved intracellular signaling. Here, we reveal that another receptor, NILR1, can complement the bri1 mutant in the absence of BRs, indicating a pathway that resembles BR signaling activated by NILR1. Genetic analysis confirms the intracellular domains of NILR1, BRI1 and EMS1 have a common signal output. Furthermore, we demonstrate that NILR1 and BRI1 share the coreceptor BRI1 Associated Kinase 1 and substrate BSKs. Notably, the NILR1-mediated downstream pathway is conserved across land plants. In summary, we provide evidence for the signaling cascade of NILR1, suggesting pan-brassinosteroid signaling initiated by a group of distant receptor-ligand pairs in land plants.

Approach to Study pH-Dependent Protein Association Using Constant-pH Molecular Dynamics: Application to the Dimerization of β-Lactoglobulin.

Protein-protein association is often mediated by electrostatic interactions and modulated by pH. However, experimental and computational studies have often overlooked the effect of association on the protonation state of the protein. In this work, we present a methodological approach based on constant-pH molecular dynamics (MD), which aims to provide a detailed description of a pH-dependent protein-protein association, and apply it to the dimerization of β-lactoglobulin (BLG). A selection of analyses is performed using the data generated by constant-pH MD simulations of monomeric and dimeric forms of bovine BLG, in the pH range 3-8. First, we estimate free energies of dimerization using a computationally inexpensive approach based on the Wyman-Tanford linkage theory, calculated in a new way through the use of thermodynamically based splines. The individual free energy contribution of each titratable site is also calculated, allowing for identification of relevant residues. Second, the correlations between the proton occupancies of pairs of sites are calculated (using the Pearson coefficient), and extensive networks of correlated sites are observed at acidic pH values, sometimes involving distant pairs. In general, strongly correlated sites are also slow proton exchangers and contribute significantly to the pH-dependency of the dimerization free energy. Third, we use ionic density as a fingerprint of protein charge distribution and observe electrostatic complementarity between the monomer faces that form the dimer interface, more markedly at the isoionic point (where maximum dimerization occurs) than at other pH values, which might contribute to guide the association. Finally, the pH-dependent dimerization modes are inspected using PCA, among other analyses, and two states are identified: a relaxed state at pH 4-8 (with the typical alignment of the crystallographic structure) and a compact state at pH 3-4 (with a tighter association and rotated alignment). This work shows that an approach based on constant-pH MD simulations can produce rich detailed pictures of pH-dependent protein associations, as illustrated for BLG dimerization.

Phylogenetic relatedness and competition: a pot experiment with semiarid tree species

Background: Factors determining the coexistence of species have been recognized since Darwin, but empirical studies have brought the relationship between niche similarity and competition into question. Knowing the mechanisms of coexistence makes it possible to predict biological invasions and determine better species combinations for the restoration and recovery of degraded areas. The aim of this study was to test the competition-relatedness hypothesis for semiarid tree species. Thus, an experiment was carried out with a phylogenetic gradient of tree species, planted in pairs. The phylogenetically closest pair consisted of congeneric species of the same clade, the intermediate pair comprised congeneric species of distinct clades, and the more distant pair were species of the same family. Results: Our results show that the phylogenetically closest and the more distant pairs corroborated the competition-relatedness hypothesis, but the presence of one species with high competitive ability in the intermediate pair caused an unexpected pattern. Conclusion: Niche differences are important for coexistence, but it is necessary to know the competitive ability of each species and the specific associations that are more productive to improve the efficiency of restoration programs and to reveal the degree of aggressiveness in the case of invasive species.