Different Types Of Errors Research Articles

Sequence-to-Sequence Models and Their Evaluation for Spoken Language Normalization of Slovenian

Sequence-to-sequence models have been applied to many challenging problems, including those in text and speech technologies. Normalization is one of them. It refers to transforming non-standard language forms into their standard counterparts. Non-standard language forms come from different written and spoken sources. This paper deals with one such source, namely speech from the less-resourced highly inflected Slovenian language. The paper explores speech corpora recently collected in public and private environments. We analyze the efficiencies of three sequence-to-sequence models for automatic normalization from literal transcriptions to standard forms. Experiments were performed using words, subwords, and characters as basic units for normalization. In the article, we demonstrate that the superiority of the approach is linked to the choice of the basic modeling unit. Statistical models prefer words, while neural network-based models prefer characters. The experimental results show that the best results are obtained with neural architectures based on characters. Long short-term memory and transformer architectures gave comparable results. We also present a novel analysis tool, which we use for in-depth error analysis of results obtained by character-based models. This analysis showed that systems with similar overall results can differ in the performance for different types of errors. Errors obtained with the transformer architecture are easier to correct in the post-editing process. This is an important insight, as creating speech corpora is a time-consuming and costly process. The analysis tool also incorporates two statistical significance tests: approximate randomization and bootstrap resampling. Both statistical tests confirm the improved results of neural network-based models compared to statistical ones.

A Simulation Study of the Effects of Additive, Multiplicative, Correlated, and Uncorrelated Errors on Principal Component Analysis

ABSTRACTMeasurement errors are ubiquitous in all experimental sciences. Depending on the particular experimental platform used to acquire data, different types of errors are introduced, amounting to an admixture of additive and multiplicative error components that can be uncorrelated or correlated. In this paper, we investigate the effect of different types of experimental error on the recovery of the subspace with principal component analysis (PCA) using numerical simulations. Specifically, we assessed how different error characteristics (variance, correlation, and correlation structure), loading structures, and data distributions influence the accuracy to estimate an error‐free (true) subspace from sampled data with PCA. Quality was assessed in terms of the mean squared reconstruction error and the congruence to the error‐free loadings, using the pseudorank and adjusting for rotational ambiguity. Analysis of variance reveals that the error variance, error correlation structure, and their interaction with the loading structure are the factors mostly affecting quality of loading estimation from sampled data. We advocate for the need to characterize and assess the nature of measurement error and the need to adapt formulations of PCA that can explicitly take into account error structures in the model fitting.

Physics simulation capabilities of LLMs

Large Language Models (LLMs) can solve some undergraduate-level to graduate-level physics textbook problems and are proficient at coding. Combining these two capabilities could one day enable AI systems to simulate and predict the physical world. We present an evaluation of state-of-the-art (SOTA) LLMs on PhD-level to research-level computational physics problems. We condition LLM generation on the use of well-documented and widely-used packages to elicit coding capabilities in the physics and astrophysics domains. We contribute ∼50 original and challenging problems in celestial mechanics (with REBOUND), stellar physics (with MESA), 1D fluid dynamics (with Dedalus) and non-linear dynamics (with SciPy). Since our problems do not admit unique solutions, we evaluate LLM performance on several soft metrics: counts of lines that contain different types of errors (coding, physics, necessity and sufficiency) as well as a more educational’ Pass-Fail metric focused on capturing the salient physical ingredients of the problem at hand. As expected, today's SOTA LLM (GPT4) zero-shot fails most of our problems, although about 40% of the solutions could plausibly get a passing grade. About 70%–90% of the code lines produced are necessary, sufficient and correct (coding & physics). Physics and coding errors are the most common, with some unnecessary or insufficient lines. We observe significant variations across problem class and difficulty. We identify several failure modes of GPT4 in the computational physics domain, such as poor physical units handling, poor code versioning, tendency to hallucinate plausible sub-modules, lack of physical justification for global run parameters (e.g., simulation time, or upper-lower bounds for parametric exploration) and inability to define steady-state or stopping conditions reliably. Our reconnaissance work provides a snapshot of current computational capabilities in classical physics and points to obvious improvement targets if AI systems are ever to reach a basic level of autonomy in physics simulation capabilities.

Neyman-Pearson Multi-class Classification via Cost-sensitive Learning

Most existing classification methods aim to minimize the overall misclassification error rate. However, in applications such as loan default prediction, different types of errors can have varying consequences. To address this asymmetry issue, two popular paradigms have been developed: the Neyman-Pearson (NP) paradigm and the cost-sensitive (CS) paradigm. Previous studies on the NP paradigm have primarily focused on the binary case, while the multi-class NP problem poses a greater challenge due to its unknown feasibility. In this work, we tackle the multi-class NP problem by establishing a connection with the CS problem via strong duality and propose two algorithms. We extend the concept of NP oracle inequalities, crucial in binary classifications, to NP oracle properties in the multi-class context. Our algorithms satisfy these NP oracle properties under certain conditions. Furthermore, we develop practical algorithms to assess the feasibility and strong duality in multi-class NP problems, which can offer practitioners the landscape of a multi-class NP problem with various target error levels. Simulations and real data studies validate the effectiveness of our algorithms. To our knowledge, this is the first study to address the multi-class NP problem with theoretical guarantees. The proposed algorithms have been implemented in the R package npcs, which is available on CRAN.

Effects of machine learning errors on human decision-making: manipulations of model accuracy, error types, and error importance

This study addressed the cognitive impacts of providing correct and incorrect machine learning (ML) outputs in support of an object detection task. The study consisted of five experiments that manipulated the accuracy and importance of mock ML outputs. In each of the experiments, participants were given the T and L task with T-shaped targets and L-shaped distractors. They were tasked with categorizing each image as target present or target absent. In Experiment 1, they performed this task without the aid of ML outputs. In Experiments 2–5, they were shown images with bounding boxes, representing the output of an ML model. The outputs could be correct (hits and correct rejections), or they could be erroneous (false alarms and misses). Experiment 2 manipulated the overall accuracy of these mock ML outputs. Experiment 3 manipulated the proportion of different types of errors. Experiments 4 and 5 manipulated the importance of specific types of stimuli or model errors, as well as the framing of the task in terms of human or model performance. These experiments showed that model misses were consistently harder for participants to detect than model false alarms. In general, as the model’s performance increased, human performance increased as well, but in many cases the participants were more likely to overlook model errors when the model had high accuracy overall. Warning participants to be on the lookout for specific types of model errors had very little impact on their performance. Overall, our results emphasize the importance of considering human cognition when determining what level of model performance and types of model errors are acceptable for a given task.

Machine Learning for Patient-Based Real-Time Quality Control (PBRTQC), Analytical and Preanalytical Error Detection in Clinical Laboratory.

The rapidly evolving field of machine learning (ML), along with artificial intelligence in a broad sense, is revolutionising many areas of healthcare, including laboratory medicine. The amalgamation of the fields of ML and patient-based real-time quality control (PBRTQC) processes could improve the traditional PBRTQC and error detection algorithms in the laboratory. This narrative review discusses published studies on using ML for the detection of systematic errors, non-systematic errors, and combinations of different types of errors in clinical laboratories. The studies discussed used ML for detecting bias, the requirement for re-calibration, samples contaminated with intravenous fluid or EDTA, delayed sample analysis, wrong-blood-in-tube errors, interference or a combination of different types of errors, by comparing the performance of ML models with human validators or traditional PBRTQC algorithms. Advantages, limitations, the creation of standardised ML models, ethical and regulatory aspects and potential future developments have also been discussed in brief.

RLclean: An unsupervised integrated data cleaning framework based on deep reinforcement learning

Data cleaning, a prerequisite to subsequent data analysis, has always been the focus of data science research. Datasets with errors can severely detract from the quality of downstream analytical results. Unfortunately, despite the proliferation of various data cleaning methods, it remains a time-consuming problem and frequently entails considerable labor expenses. In reality, errors are often heterogeneous and require different solutions. As a result, stand-alone methods often inadequate for addressing dirty data with multiple types of errors, while studies have demonstrated that combining such methods always require human intervention and the result remains unsatisfactory.In this paper, we propose an unsupervised integrated data cleaning framework, namely RLclean. Based on deep reinforcement learning, RLclean takes advantages of multiple data cleaning techniques, enabling it to effectively clean multiple types of errors and achieve satisfactory results. Additionally, it eliminates the need for costly human involvement, as the cleaning strategy is learned by data-driven, which further allows the framework to self-adapt to diverse domains. RLclean mainly consists of two parts: (i) an integrated error detection model that unites multiple techniques to detect different types of errors from multiple views; and (ii) an integrated data repair model that learns the optimal repair operations and repairs dirty data in an unsupervised manner. Extensive experiments on benchmark datasets have demonstrated the superiority of RLclean over state-of-the-art methods.

Errors in the English Article System Among EFL Learners: Evidence from the Kurdistan Region - Iraq

The identification of the sources and types of students’ errors in the use of the English article system is beneficial in assisting them to overcome such problems. This study aims to investigate the types and sources of article errors made by Kurdish EFL learners. The data was collected from 61 third-year university students who were studying English at the time of the present study. The types of article errors were analysed based on surface structure taxonomy (SST). Contrastive and error analyses were adopted to analyse the sources of article errors. The study findings suggest that the students committed different types of errors in the use of the English article system, namely, substitution, omission, and addition. Omission errors were the most frequent type of article errors, while substitution errors were the least frequent ones. The sources of errors are divided into two categories: inter-lingual and intra-lingual. Analysis of the student's written samples indicates that most of the errors in the articles are attributed to the target language itself. The implications of the study suggest that teachers should pay more attention to the acquisition of the English article system. Limitations and future research considerations are discussed in the article.

Error detection for radiotherapy planning validation based on deep learning networks.

Quality assurance (QA) of patient-specific treatment plans for intensity-modulated radiation therapy (IMRT) and volumetric modulated arc therapy (VMAT) necessitates prior validation. However, the standard methodology exhibits deficiencies and lacks sensitivity in the analysis of positional dose distribution data, leading to difficulties in accurately identifying reasons for plan verification failure. This issue complicates and impedes the efficiency of QA tasks. The primary aim of this research is to utilize deep learning algorithms for the extraction of 3D dose distribution maps and the creation of a predictive model for error classification across multiple machine models, treatment methodologies, and tumor locations. We devised five categories of validation plans (normal, gantry error, collimator error, couch error, and dose error), conforming to tolerance limits of different accuracy levels and employing 3D dose distribution data from a sample of 94 tumor patients. A CNN model was then constructed to predict the diverse error types, with predictions compared against the gamma pass rate (GPR) standard employing distinct thresholds (3%, 3mm; 3%, 2mm; 2%, 2mm) to evaluate the model's performance. Furthermore, we appraised the model's robustness by assessing its functionality across diverse accelerators. The accuracy, precision, recall, and F1 scores of CNN model performance were 0.907, 0.925, 0.907, and 0.908, respectively. Meanwhile, the performance on another device is 0.900, 0.918, 0.900, and 0.898. In addition, compared to the GPR method, the CNN model achieved better results in predicting different types of errors. When juxtaposed with the GPR methodology, the CNN model exhibits superior predictive capability for classification in the validation of the radiation therapy plan on different devices. By using this model, the plan validation failures can be detected more rapidly and efficiently, minimizing the time required for QA tasks and serving as a valuable adjunct to overcome the constraints of the GPR method.

Exploring the Performance of Continuous-Time Dynamic Link Prediction Algorithms

Dynamic Link Prediction (DLP) addresses the prediction of future links in evolving networks. However, accurately portraying the performance of DLP algorithms poses challenges that might impede progress in the field. Importantly, common evaluation pipelines usually calculate ranking or binary classification metrics, where the scores of observed interactions (positives) are compared with those of randomly generated ones (negatives). However, a single metric is not sufficient to fully capture the differences between DLP algorithms, and is prone to overly optimistic performance evaluation. Instead, an in-depth evaluation should reflect performance variations across different nodes, edges, and time segments. In this work, we contribute tools to perform such a comprehensive evaluation. (1) We propose Birth–Death diagrams, a simple but powerful visualization technique that illustrates the effect of time-based train–test splitting on the difficulty of DLP on a given dataset. (2) We describe an exhaustive taxonomy of negative sampling methods that can be used at evaluation time. (3) We carry out an empirical study of the effect of the different negative sampling strategies. Our comparison between heuristics and state-of-the-art memory-based methods on various real-world datasets confirms a strong effect of using different negative sampling strategies on the test area under the curve (AUC). Moreover, we conduct a visual exploration of the prediction, with additional insights on which different types of errors are prominent over time.

Reading takes it all? – the role of language competence and subject knowledge in legal translation

ABSTRACT In this paper the role of language competence and thematic competence in translation and post-editing was studied. The sample included 49 translation students who translated part of a copyright agreement from English into Hungarian under two different conditions, from-scratch human translation and post-editing. Our findings indicated that there was no difference between from-scratch human translation and post-editing in terms of quality and speed, and that time spent on task was only loosely connected to achievement. We also found that thematic (legal) knowledge, as assessed by our multiple-choice test, did not correlate with from-scratch human translation/post-editing performance. Perhaps the most important finding of the study is that source language (SL) reading plays a key role in translation students’ performance, whether it is traditional from-scratch human translation or post-editing. It was also highlighted that SL reading was even more important in post-editing than in from-scratch human translation, and that weaknesses in reading competence may lead to different types of errors in the two working conditions.

Understanding the operators’ cloud change errors based on cognitive abilities and personality traits: An investigation integrated with quantitative and qualitative methods

While cloud services make industrial data convenient, they also expose it to cloud incidents. Operators' error in cloud change activities is a leading factor for cloud incidents, which have received relatively less attention in cloud security research. This study conducted a two-stage research process using an integrated approach to explore the stable individual factors related to cloud change errors. First, in the qualitative research, content analysis based on interviews and historical documents was conducted to extract the operator's cognitive abilities and personality traits and develop hypotheses. Five cognitive abilities and six personality traits were extracted. Second, quantitative research based on an experiment was conducted to test relationships between operators' different types of cloud change errors and 1) cognitive ability and 2) personality traits, respectively. Results of error type comparisons suggested that operators generated more uncorrected errors than corrected errors and more operational errors than omission errors in cloud change activities. The multivariate Poisson regression analysis suggested that cognitive abilities of sustained attention, divided attention, and long-term memory negatively predicted the number of operators' total errors, uncorrected errors, and operational errors. Regarding personality traits, with the increase in resilience capacity and carefulness and the decrease in self-esteem, the number of different types of errors reduced, except for omission errors. Working memory and risk-taking propensity were also significant predictors of the number of uncorrected errors with negative and positive coefficients, respectively. Logical reasoning, emotional stability, and sense of responsibility were not observed as predictors of cloud change errors. The present findings have several implications for the industry and cloud providers to enhance industrial cloud data security regarding human cognitive abilities and personality traits.

Errors in visual search: Are they stochastic or deterministic?

In any visual search task in the lab or in the world, observers will make errors. Those errors can be categorized as “deterministic”: If you miss this target in this display once, you will definitely miss it again. Alternatively, errors can be “stochastic”, occurring randomly with some probability from trial to trial. Researchers and practitioners have sought to reduce errors in visual search, but different types of errors might require different techniques for mitigation. To empirically categorize errors in a simple search task, our observers searched for the letter “T” among “L” distractors, with each display presented twice. When the letters were clearly visible (white letters on a gray background), the errors were almost completely stochastic (Exp 1). An error made on the first appearance of a display did not predict that an error would be made on the second appearance. When the visibility of the letters was manipulated (letters of different gray levels on a noisy background), the errors became a mix of stochastic and deterministic. Unsurprisingly, lower contrast targets produced more deterministic errors. (Exp 2). Using the stimuli of Exp 2, we tested whether errors could be reduced using cues that guided attention around the display but knew nothing about the content of that display (Exp3a, b). This had no effect, but cueing all item locations did succeed in reducing deterministic errors (Exp3c).

Preceding word information for predicting speech errors in English as foreign language speech

Speech errors, including disfluency errors (e.g., filled pauses (“uh”, “um”), repetition (“I me-mean right now.”)), and mispronunciation of speech segments (e.g., “think” as /sɪŋk/) are natural occurrences in speech production and they can affect speech fluency and proficiency. Detecting these errors is important, especially in assessing second language (L2) learners. Non-native speakers often produce speech errors, even in read speech, due to increased cognitive load when simultaneously producing the current word and processing the upcoming word. By analyzing two L2 speech corpora having different types of errors (disfluency and mispronunciation), the study aims to assess the effectiveness of incorporating preceding word information in speech error prediction. The results indicate that longer syllable duration in the word preceding the mispronunciation error correlates with error production. The study extends its investigation to identify acoustic and lexical features influenced by the presence of speech errors. The study found that speech errors increase duration of speech. The study also found that words with more syllables may cause L2 speakers to produce speech errors.

Obtaining Lower Query Complexities Through Lightweight Zeroth-Order Proximal Gradient Algorithms.

Zeroth-order (ZO) optimization is one key technique for machine learning problems where gradient calculation is expensive or impossible. Several variance, reduced ZO proximal algorithms have been proposed to speed up ZO optimization for nonsmooth problems, and all of them opted for the coordinated ZO estimator against the random ZO estimator when approximating the true gradient, since the former is more accurate. While the random ZO estimator introduces a larger error and makes convergence analysis more challenging compared to coordinated ZO estimator, it requires only O(1) computation, which is significantly less than O(d) computation of the coordinated ZO estimator, with d being dimension of the problem space. To take advantage of the computationally efficient nature of the random ZO estimator, we first propose a ZO objective decrease (ZOOD) property that can incorporate two different types of errors in the upper bound of convergence rate. Next, we propose two generic reduction frameworks for ZO optimization, which can automatically derive the convergence results for convex and nonconvex problems, respectively, as long as the convergence rate for the inner solver satisfies the ZOOD property. With the application of two reduction frameworks on our proposed ZOR-ProxSVRG and ZOR-ProxSAGA, two variance-reduced ZO proximal algorithms with fully random ZO estimators, we improve the state-of-the-art function query complexities from Omindn1/2ε2,dε3 to O˜n+dε2 under d>n12 for nonconvex problems, and from Odε2 to O˜nlog1ε+dε for convex problems. Finally, we conduct experiments to verify the superiority of our proposed methods.

Advancing real-time error correction of flood forecasting based on the hydrologic similarity theory and machine learning techniques

Real-time flood forecasting is one of the most pivotal measures for flood management, and real-time error correction is a critical step to guarantee the reliability of forecasting results. However, it is still challenging to develop a robust error correction technique due to the limited cognitions of catchment mechanisms and multi-source errors across hydrological modeling. In this study, we proposed a hydrologic similarity-based correction (HSBC) framework, which hybridizes hydrological modeling and multiple machine learning algorithms to advance the error correction of real-time flood forecasting. This framework can quickly and accurately retrieve similar historical simulation errors for different types of real-time floods by integrating clustering, supervised classification, and similarity retrieval methods. The simulation errors “carried” by similar historical floods are extracted to update the real-time forecasting results. Here, combining the Xin'anjiang model-based forecasting platform with k-means, K-nearest neighbor (KNN), and embedding based subsequences matching (EBSM) method, we constructed the HSBC framework and applied it to China's Dufengkeng Basin. Three schemes, including “non-corrected” (scheme 1), “auto-regressive (AR) corrected” (scheme 2), and “HSBC corrected” (scheme 3), were built for comparison purpose. The results indicated the following: 1) the proposed framework can successfully retrieval similar simulation errors with a considerable retrieval accuracy (2.79) and time consumption (228.18 s). 2) four evaluation metrics indicated that the HSBC-based scheme 3 performed much better than the AR-based scheme 2 in terms of both the whole flood process and the peak discharge; 3) the proposed framework overcame the shortcoming of the AR model that have poor correction for the flood peaks and can provide more significant correction for the floods with bad forecasting performance. Overall, the HSBC framework demonstrates the advancement of benefiting the real-time error correction from hydrologic similarity theory and provides a novel methodological alternative for flood control and water management in wider areas.

Error Culture, Auditors’ Error Communication, and the Performance of the Auditee: A Study Among German Local Public Sector Auditors

In the course of financial and value for money audits, public sector auditors are facing different types of errors: accounting and economic errors, respectively. This study examines the relations between error culture in public sector audit organizations, auditors’ communication of accounting and economic errors, and performance of the auditee. The analyses of survey data from German local public sector auditors show that a strong error culture within the audit organization positively affect the auditors’ communication of errors to the auditee, regardless of the error type. Additionally, a strong error culture positively mediates the performance of the audited institution through the auditors’ communication of economic errors. This implies that it is important for public sector audit organizations to build a strong error culture. In addition to its practical contrition, the study provides novel theoretical insights as it demonstrates that the error management of one organization (audit entity) matters for the performance of another organization (audited entity).

GeneGPT: augmenting large language models with domain tools for improved access to biomedical information.

While large language models (LLMs) have been successfully applied to various tasks, they still face challenges with hallucinations. Augmenting LLMs with domain-specific tools such as database utilities can facilitate easier and more precise access to specialized knowledge. In this article, we present GeneGPT, a novel method for teaching LLMs to use the Web APIs of the National Center for Biotechnology Information (NCBI) for answering genomics questions. Specifically, we prompt Codex to solve the GeneTuring tests with NCBI Web APIs by in-context learning and an augmented decoding algorithm that can detect and execute API calls. Experimental results show that GeneGPT achieves state-of-the-art performance on eight tasks in the GeneTuring benchmark with an average score of 0.83, largely surpassing retrieval-augmented LLMs such as the new Bing (0.44), biomedical LLMs such as BioMedLM (0.08) and BioGPT (0.04), as well as GPT-3 (0.16) and ChatGPT (0.12). Our further analyses suggest that: First, API demonstrations have good cross-task generalizability and are more useful than documentations for in-context learning; second, GeneGPT can generalize to longer chains of API calls and answer multi-hop questions in GeneHop, a novel dataset introduced in this work; finally, different types of errors are enriched in different tasks, providing valuable insights for future improvements. The GeneGPT code and data are publicly available at https://github.com/ncbi/GeneGPT.

Rate of penetration prediction with uncertainty assessment: Case study of a middle-east oil field

To increase drilling operation efficiency, attempts have been made to increase the rate of bit penetration into the formations, which is a function of controllable and uncontrollable parameters. Several mathematical models are available for penetration rate estimation. However, these models are not applicable in all circumstances. The most used model for rate of penetration is the one proposed by Bourgoyne and Young and the model has shown reliable results in different oil and gas provinces. However, the model performance significantly depends on the quantity and quality of input data. When the input data are subjected to different types of errors, some extent of uncertainty may propagate to the model. In this work, a Monte Carlo approach was followed to model the uncertainty in the predicted rates of penetrations. For this purpose, input data from an Iranian oil field were acquired and processed to check their quality. After fitting a probability distribution to each input data, the level of uncertainty in the predicted rates of penetration was quantified. Results showed that with an uncertain space of input data, a probabilistic rate of penetration was predicted. For a specific depth point, i.e. 8158.6 ft, rate of penetration range was narrowed to 1.16–1.23 ft/h with a certainty of 10 % and a mean value of 1.23 ft/h. However, for higher certainty levels a wider range of rate of penetration was obtained. Therefore, at each depth point, the certainty bands related to the predicted rate of penetration should also be stated as the model output.

Application of Corpus Analysis Software and Error Analysis Theory on Non-English Majors’ Chinese-English Translation Features

This research focuses on non-English major students’ Chinese-English (C-E) translation features. It aims at annotating and summarizing the features and error types in students’ C-E translation by applying to corpus analysis software Readability Analyzer. This study first collects students’ C-E translation texts from a computer-based exam, then forms a self-constructed corpus and a reference corpus. The translation materials in the target corpus and the reference corpus are first analyzed and compared by Readability Analyzer to figure out the features of students C-E translation texts from the perspectives of sentence length, word length, percentage of difficult words, Flesch Reading Ease, Dale-Chall Score, Fry Readability Grade Level. After that, different types of errors in students C-E translation are labeled and analyzed based on statistical analysis and Error Analysis (EA) theory. According to Readability Analyzer, students C-E translation texts have the following features. From passage statistics, it shows that translation texts that get lower scores tend to use shorter sentences and easier words. In addition, from readability scores, Flesch scores, Dale-Chall Scores shows and Fry Readability Grade Level are taken into consideration. Flesch scores shows that compared with the reference corpus, students’ translation which get lower scores are harder to understand since there are more language errors than higher-score translation. Dale-Chall Scores indicate that translation materials which score 8-9 are very close in the difficulty of word-using with the reference corpus. Fry Readability Grade Level shows that texts of lower scores are simpler and more accurate than texts of higher scores. According to Error Analysis (EA) Theory, students’ errors are further classified and discussed on lexical level, sentence level and textual level. The conclusions of this research are expected to make contributions to the improvement of C-E translation teaching and help college students, especially non-English majors, to enhance their C-E translation competence.