COMPARATIVE ANALYSIS OF BACK-TRANSLATION MODELS FOR NORMALIZATION MOBILE APP USER REVIEWS

This study delves into the translational challenges Arabic translators face in preserving the dynamism and multifaceted meanings of ‘come’ and ‘go’ phrasal verbs during translation using machine tools. Though existing literature has focused on linguistic and cultural equivalence, the ‘translational force’ of specific words is still under-researched. The aim of this study is to verify the accuracy of the output across tools, and to conclude how effective these applications were in the given language pair. The study adopts a mixed methods approach with a survey used to identify the most frequently used machine tools for translation by 27 learners of Translation Studies at Shaqra University, Saudi Arabia. Thereafter, a set of frequently used phrasal verbs in English that use the root verbs ‘come’ and ‘go’ were translated by a convenience sample of 27 final year students of translation as well as by using the popular and free translation tools in the English-Arabic language pair. Results indicated that Saudi students enrolled in translation studies used MateCAT, Google Translate, Reverso to a limited extent, but relied more on online dictionaries for translation to and from English. In addition to this, translation of English phrasal verbs was found to be most accurate in MateCat and there too, accuracy in back translation was found to be more accurate in the case of the phrasal verbs formed with ‘come’ at the root than with ‘go’ verb. Finally, it was found that the translation students preferred ‘soft tools’ like English movies for learning, recall, and use of English phrasal verbs in practice. This study is expected to provide a deeper understanding of the concept of "translational force" and its implications for translating dynamic words in the English-Arabic language pair in the realm of machine-based translation tools.

The impact of dental appearance, malocclusion and treatment for the same on psychological and functional well-being has drawn increasing attention over the past decade. Various psychometric instruments alongside normative indices have been used to predict orthodontic concerns. Evaluating the patients' experience during the orthodontic treatment can help us understand the true benefits and advantages of orthodontic therapy. The aim of the present study was to evaluate the change in the psychosocial impact of malocclusion using the Psychosocial Impact of Dental Aesthetics Questionnaire (PIDAQ) adapted for the Indian population after one year of fixed orthodontic treatment. This interventional study was conducted on 93 patients requiring fixed orthodontic treatment. Brazilian, Chinese, Spanish, Nepali and Moroccan versions of the PIDAQ have been published but the questionnaire is not available in Hindi. In the present study, the original PIDAQ was translated into Hindi language to adapt it for the Indian population and was validated by back translation and pretest. All the subjects answered the Hindi version of the questionnaire at pretreatment (T1) and at one year of orthodontic treatment (T2). Additionally, the Index of Orthodontic Treatment Need (IOTN) was applied to measure the severity and self-perception of malocclusion. The data were analysed using paired t-test, Wilcoxon signed rank test and Kruskal-Wallis test. Significant reduction was found in the total PIDAQ score and each factor's score (p<0.001) after one year of orthodontic treatment. There was a positive association of the psychosocial impact of malocclusion with the IOTN-AC (IOTN-Aesthetic Component). Adolescent females were found to be most concerned with their dentofacial appearance. Results showed significant improvement in the psychosocial impact of malocclusion with a reduction in the self-perceived needs of patients with orthodontic treatment. The psychometric instrument used may be recommended as an Oral Health Related Quality of Life (OHRQoL) assessment tool for the population in India for further research.

There is growing interest in the anticancer effect of the diabetes drug, metformin, and over 100 clinical trials are underway worldwide. However it is still not determined as to whether metformin has significant direct effects on cancer cells or solely indirect effects via modulation of host metabolism. We recruited 41 non-diabetic patients with primary breast cancer to a neoadjuvant window trial. Patients received an escalating dose of metformin (500mg up to 1500mg) for 2 weeks with pre- and post-metformin pharmacodynamic assessments including, dynamic 18F-FDG PET-CT scans, serum metabolic markers, and tumour biopsies for whole transcriptome RNASeq, tumour metabolomics and immunohistochemistry. Assessment of tumour FDG kinetics using a classic 2-tissue compartment model with three rate constants showed a 1.3 fold change (FC) post-metformin in the composite 18F-FDG flux constant, Kflux (p = 0.041, on paired t-test). There was a non-statistically significant 1.1 fold increase in the variable, metabolic rate of glucose utilization (MRglu) (p = 0.141). Mass spectrometry analysis revealed a decrease in intra-tumoral levels of the short-chain acyl-carnitines, propionylcarnitine (FC -0.50, p = 0.039) and acetylycarnitine (FC -0.40, p = 0.046) consistent with inhibition of fatty acid oxidation (wilcoxon rank test). This effect on fatty acid oxidation was validated in pre-clinical in vitro and in vivo breast cancer cell line models. There was a statistically significant positive correlation between the actual difference in Kflux and intratumoral levels of acetylcarnitine (p = 0.012, Spearman's rank test). Small but consistent falls in the levels of serum glucose (p = 0.032), c-peptide (p = 0.001), insulin (p = 0.005), IGF1 (p = 0.048) and IGF2 (p = 0.040) were observed following metformin treatment (paired t-test) but there was no correlation with change in Kflux or short chain acyl-carnitine levels. Pathway annotation analysis using GeneCodis revealed several pathways associated with mitochondrial and fatty acid metabolism that were upregulated. Immunohistochemical intratumoral nuclear expression of pAMPK increased 1.5 fold (p = 0.037, paired t-test) but this did not correlate with change in Kflux or levels of short-chain acyl-carnitines. Peak serum metformin levels correlated with intratumoral metformin levels (p = 0.012, Spearman's rank test) but did not correlate with change in Kflux or short-chain acyl-carnitine levels. Our data shows that metformin treatment alters FDG kinetics, inhibits fatty acid oxidation and leads to altered gene expression in the mitochondrial (nuclear encoded) transcriptome but that these effects do not correlate with changes in host metabolism. This data provides strong evidence that metformin has a direct effect on breast cancer metabolism at clinical doses. Citation Format: Simon Lord, Daniel Liu, Syed Haider, Edoardo Gaude, Eugene Teoh, Patel Neel, Qifeng Zhang, Fergus Gleeson, Michael Wakelam, Christian Frezza, Francesca Buffa, Fenwick John, Adrian Harris. Integrating dynamic 18F-FDG PET-CT, tumor metabolomics and functional genomics to understand metformin's pharmacodynamic effects in breast cancer: results of a phase 0 clinical trial. [abstract]. In: Proceedings of the 107th Annual Meeting of the American Association for Cancer Research; 2016 Apr 16-20; New Orleans, LA. Philadelphia (PA): AACR; Cancer Res 2016;76(14 Suppl):Abstract nr LB-200.

We present a certified compiler from the simply-typed lambda calculus to assembly language. The compiler is certified in the sense that it comes with a machine-checked proof of semantics preservation, performed with the Coq proof assistant. The compiler and the terms of its several intermediate languages are given dependent types that guarantee that only well-typed programs are representable. Thus, type preservation for each compiler pass follows without any significant “proofs ” of the usual kind. Semantics preservation is proved based on denotational semantics assigned to the intermediate languages. We demonstrate how working with a type-preserving compiler enables type-directed proof search to discharge large parts of our proof obligations automatically. Categories and Subject Descriptors F.3.1 [Logics and meanings

Machine translations offer several features that simplify the translation process. These include the ability to automatically adjust grammar and vocabulary from the source language to the target language, store word banks, and tailor the target language to specific fields. However, machine translation outputs still contain errors. Therefore, it is essential to perform post-editing to ensure grammatical and lexical cohesion and alignment between the source and target languages. This research aims to observe the quality of Google Translate and the result of Post-editing machine translation by EFL learners in terms of grammatical cohesion. This study used a descriptive qualitative. This study applied purposive sampling to choose the data. Three abstracts from three different Translation and Interpreting classes were chosen for the data. Then, the data were analysed by using the theory of grammatical cohesion and translation quality assessment. The result of the study shows that the grammatical cohesions found are mostly reference and conjunction. This study also finds that the quality of post-editing is better than the result of Google Translation as machine translation. The samples of data provides that mostly the score of post-editing is better observed from translation acceptability and readability. This study is restricted to the translation result of three abstracts only and the students’ post-editing effort which means that the findings may not fully represent the overall quality of translations from a grammatical cohesion. Future research could delve deeper into this topic by expanding the data sources and incorporating machine translation tools other than Google Translate.

Requirements engineering is a process of constantly changing worlds of intentions, goals, and system models. Conventional semantics for goal specifications is synchronous. Semantics of conventional system modeling techniques is asynchronous. This semantic mismatch complicates requirements engineering. In this chapter, we propose a new method EXTREME that exploits similarities in semantics of goal specification and executable protocol models. In contrast with other executable modelling techniques, the semantics of protocol modelling is based on a data extended form of synchronous CSP-parallel composition. This synchronous composition provides advantages for relating goals and system models, reasoning on models, requirements management, and evolution.

e13609 Background: Precision oncology revolutionized cancer treatment by identifying molecular biomarkers to guide personalized care. The ever-growing body of medical literature presents a challenge for oncologists researching targeted therapies. While recent studies investigated large language models (LLMs) to streamline this process, LLM reliance on general rather than medical knowledge limits clinical relevance and trustworthiness. To address these limitations, we developed a retrieval augmented generation (RAG) system that integrates PubMed clinical studies, trial databases and oncological guidelines with LLMs to support targeted treatment recommendations. The Molecular Tumor Board (MTB) at the Center of Personalized Medicine (ZPMTUM) guided and evaluated treatment options proposed by the LLM to assess their applicability for clinical decision support. Methods: We used 10 publicly accessible fictional patient cases with 7 tumor types and 59 distinct molecular alterations. Our LLM system MEREDITH (Medical Evidence Retrieval and Data Integration for Tailored Healthcare) consists of Google's Gemini Pro, enhanced with RAG and Chain-of-Thought (CoT) prompting. To establish a benchmark, clinical experts at ZPMTUM manually annotated the cases. Informed by MTB expert feedback, we iteratively improved our LLM system from a draft system relying on PubMed-indexed data to an enhanced system, which replicated expert annotation processes by incorporating oncology guidelines, drug availability and trial databases (ClinicalTrials.gov, QuickQueck.de). ZPMTUM assessed credibility and clinical relevance of manually annotated and LLM-generated recommendations. Patient-level data on (likely) pathogenic molecular alterations and recommended treatment options were summarized using median and interquartile range (IQR). Semantic similarity between LLM and clinician responses was assessed using cosine similarity of text vector embeddings; paired t-test evaluated significance. Results: The median of (likely) pathogenic molecular alterations per patient was 2.5 (IQR: 2-3). ZPMTUM identified a median of 2 treatment options per patient (IQR: 1-3), while the enhanced LLM identified a median of 4 (IQR: 3-5). MEREDITH proposed multiple relevant treatment suggestions, including therapies based on preclinical studies, and molecular interactions, for further assessment by the MTB. ZPMTUM prioritized the most suitable clinical option. The mean semantic textual similarity of LLM responses increased significantly from 0.69 in the draft system to 0.76 in the enhanced system (p &lt;0.001). Thus, feedback from ZPMTUM enhanced the model's ability to align its responses with clinician thought processes. Conclusions: Leveraging expert thought processes to instruct LLMs holds promise as a novel decision support tool for precision oncology.

Transformer models have great potential in Graph Representation Learning (GRL) for efficiently scaling the learning process on large datasets and solving many challenges presented in Graph Neural Networks, e.g., oversmoothing and suspended animation. To represent each node of a graph, Transformer models as input usually take a node together with the node context, i.e., a set of other nodes that serve as learning context for the target node. However, current GRL Transformer models mainly consider the graph topology when selecting the node context for each target node. In this work, we demonstrate the important role of node features in selecting the node context. Specifically, we propose a hybrid approach for selecting node context that considers both the graph topology and the semantic similarities between node features. Through the empirical evaluations, we show the advantages of our hybrid node context selection method for a downstream classification task on various datasets compared to selection methods that only consider graph topology or semantic similarities. The best classification accuracy improvements of our proposed hybrid methods over the baseline methods on each dataset range from 0.77% to 6.05%.

Hashing-based approximate nearest neighbor (ANN) search in huge databases has become popular due to its computational and memory efficiency. The popular hashing methods, e.g., Locality Sensitive Hashing and Spectral Hashing, construct hash functions based on random or principal projections. The resulting hashes are either not very accurate or are inefficient. Moreover, these methods are designed for a given metric similarity. On the contrary, semantic similarity is usually given in terms of pairwise labels of samples. There exist supervised hashing methods that can handle such semantic similarity, but they are prone to overfitting when labeled data are small or noisy. In this work, we propose a semi-supervised hashing (SSH) framework that minimizes empirical error over the labeled set and an information theoretic regularizer over both labeled and unlabeled sets. Based on this framework, we present three different semi-supervised hashing methods, including orthogonal hashing, nonorthogonal hashing, and sequential hashing. Particularly, the sequential hashing method generates robust codes in which each hash function is designed to correct the errors made by the previous ones. We further show that the sequential learning paradigm can be extended to unsupervised domains where no labeled pairs are available. Extensive experiments on four large datasets (up to 80 million samples) demonstrate the superior performance of the proposed SSH methods over state-of-the-art supervised and unsupervised hashing techniques.

This study utilized a mixed research design, integrating both quantitative and qualitative methods to investigate the challenges of English email writing among bank employees and to explore the strategies they employ to manage these difficulties. For data collection, the study analyzed 172 English emails sent and received by 63 bank employees between May 1, 2023, and July 31, 2023. The purpose was to classify email types and examine writing issues along with strategies for improvement. The data were analyzed using descriptive statistics including frequency, percentage, mean, and standard deviation, along with comparative analysis through t-tests and F-tests. The findings showed that the most frequently received email types were Informational emails (33 items, 19.19%), followed by Request emails (25 items, 14.53%), Updating emails (18 items, 10.47%), Invitation emails (17 items, 9.88%), and Attachment emails (16 items, 9.30%). An analysis of grammar, spelling, and punctuation errors revealed that the most common mistakes occurred in the use of adjectives, verbs, and spelling. No statistically significant difference (α = .05) was found in English email writing styles between male and female employees. However, at the sub-variable level, a significant difference was identified in how employees addressed the challenges of writing emails, especially regarding the clarity of using online resources to enhance writing and make a favorable impression on recipients. This was supported by post hoc analysis using the Least Significant Difference (LSD) method. Both male and female employees, from various experience levels, often used translation tools such as Google Translate, DeepL, Wordvice AI, Alexa Translations, and Bing Microsoft Translator to reduce errors. Employees with less than five years of experience generally reported higher accessibility to these tools than those with over fifteen years of experience.

Intercultural Crisis Communication: Translation, Interpreting and Languages in Local Crises, Federico M. Federici and Christophe Declercqxii+2682020Bloomsbury AcademicLondon and New York9781350097056

Popular mobile applications receive millions of user reviews. These reviews contain relevant information, such as problem reports and improvement suggestions. The reviews information is a valuable knowledge source for software requirements engineering since the analysis of the reviews feedback helps to make strategic decisions in order to improve the app quality. However, due to the large volume of texts, the manual extraction of the relevant information is an impracticable task. In this paper, we investigate and compare textual representation models for app reviews classification. We discuss different aspects and approaches for the reviews representation, analyzing from the classic Bag-of-Words models to the most recent state-of-the-art Pre-trained Neural Language models. Our findings show that the classic Bag-of-Words model, combined with a careful analysis of text pre-processing techniques, is still a competitive model. However, pre-trained neural language models showed to be more advantageous since it obtains good classification performance, provides significant dimensionality reduction, and deals more adequately with semantic proximity between the reviews' texts, especially the multilingual neural language models.

A bilingual dictionary is a valuable linguistic resource which records, among other things, the differences in the segmentation of semantic space by the two languages and hence the difficulty in producing faithful translations between the two languages. Statistical analysis of nearly a hundred dictionaries has allowed us to determine how best to measure the semantic distance between languages from bilingual dictionaries. The distribution of the number of words in language A having n translations in language B, for n = 1, 2, 3, etc., was found to have a specific shape depending on the semantic distance between the two languages. A sample of only a thousand words was sufficient to obtain an estimate of semantic distance. We give a theoretical justification for this distance based on models of the historical evolution of monolingual and bilingual dictionaries. Among our linguistic findings, we discovered, for example, that French is semantically closer to Basque than to German. We envisage an application of our semantic distance measure in the choice of an intermediate language when performing indirect translation, i.e. translating from language A to language B via a third language C.

Semantic similarity information supports requirements tracing and helps to reveal important requirements quality defects such as redundancies and inconsistencies.Previous work has applied semantic similarity algorithms to requirements, however, we do not know enough about the performance of machine learning and deep learning models in that context.Therefore, in this work we create the largest dataset for analyzing the similarity of requirements so far through the use of Amazon Mechanical Turk, a crowd-sourcing marketplace for micro-tasks. Based on this dataset, we investigate and compare different types of algorithms for estimating semantic similarities of requirements, covering both relatively simple bag-of-words and machine learning models.In our experiments, a model which relies on averaging trained word and character embeddings as well as an approach based on character sequence occurrences and overlaps achieve the best performances on our requirements dataset.KeywordsRequirements engineeringSimilarity detectionMachine learning

In recent days, semantic web presents a key role in intelligent retrieval of information system that resolves vocabulary mismatch problem by query expansion process. However, achieving the scalable information retrieval (IR) in semantic web is a challenging issue in a large dataset. The semantic IR problem is addressed by an ontological-based semantic similarity measurement using natural language processing. The two novel algorithms namely syntactic correlation coefficient (SCC) and mapping-based K-nearest neighbour (M-KNN) for semantic similarity measurement is proposed which improves the accuracy of relevant result. The ontological constructs with word sense disambiguation (WSD) algorithm for document repository improves the conceptual relationships, reduces the ambiguities in ontology and improves scalability by intensely analysing the semantic relationship as well as dynamically reconstructing the ontology when numbers of documents are updated. Ranking is done with latent semantic analysis (LSA) after semantic similarity analysis, which improves the retrieved result and reduces the complexity in relevancy. The performance of the system is analysed with respect to different metrics such as processing time, F-measure (0.97), time complexity, precision (0.95), recall (0.98) and space complexity.