Search

Declerck, M., Koch, I., Dunabeitia, J., Grainger, J., & Denise, N. [(2019). What absent switch costs and mixing costs during bilingual language comprehension can tell us about language control? Jou...

Today, due to problems in improving computing performance, parallel programming continues to evolve. There are many different languages in which you can write parallel programs. One of them is the functional-threading parallel programming language Pifagor, which in turn is very specific and allows you to write a program with maximum parallelism, as well as it is designed to solve the portability problem of parallel programs. Tools and a library of functions continue to be developed for this language. This study is devoted to the development of elements of the mathematical library and the search for the most effective mathematical parallel algorithms. The following methods are considered and used in the work: sequential, recursive (left and right recursion), factorization, and pairwise comparisons. As a result of the study, a number of mathematical functions were developed, and a study was made of the possibility of using these functions in the development of programs for multiplying large-dimensional matrices. The work demonstrates the effectiveness of using the developed simple functions implemented by different methods in matrix multiplication programs. The prospects of further work in this direction are noted, having in mind the analysis of the possibility of using artificial intelligence methods to increase efficiency and facilitate the development of parallel programs with large-sized matrices.

This article discusses how students experience and evaluate the language practices they meet during the course of their university education. The study focuses on the students’ experiences and assessments of the use and mastery of English and the local language(s) in their education. Firstly, I describe language policies at higher education institutions in the Nordic countries, where parallel language use is a preferred strategy. I then present a thematic analysis of students’ comments, collected through a survey distributed to students at five universities in three of the Nordic countries. Whereas students value the use of both English and the local language(s), they ask for more attention directed at how the different languages should be managed throughout the course of their education. The student comments show that it can be challenging to succeed with parallel language use. Some of the difficulties students face are connected to their own challenges with understanding and using other languages than their first language, while others thematise lecturers’ lack of language competence. These challenges are discussed in light of Fricker’s (2007) concept of epistemic injustice. Lastly, I argue that language policies should to a greater extent include acquisition planning, in addition to status planning., Denne undersøkinga handlar om korleis studentar opplever og vurderer den språklege praksisen dei møter gjennom utdanningsløpet. Under­søkinga fokuserer på studentane sine opplevingar og vurderingar knytt til bruk og meistring av engelsk og dei lokale språka i undervisinga. Først skildrar eg språkpolitiske retningslinjer i dei nordiske landa, der parallell­språk har vore den føretrekte språkpolitikken i høgare utdanning. Så gjer eg ein tematisk analyse av studentkommentarar frå ei spørjeundersøking som blei gjennomført på fem universitet i tre av dei nordiske landa. Trass i at studentane i hovudsak er positive til å bruke både engelsk og dei lokale språka, etterspør dei meir fokus på handteringa av dei ulike språka gjennom utdanningsløpet. Studentkommentarane syner at det kan vere utfordrande å lukkast med parallellspråkpolitikken. Nokre vanskar som studentane fortel om, knyter dei til eigne utfordringar med å forstå og bruke andre språk enn førstespråket, medan andre vanskar vert knytt til opplevinga av at førelesarar manglar ferdigheiter i språka. Desse utfordr­ing­ane vert diskutert i lys av Fricker (2007) sitt omgrep epistemisk urett. Til slutt argumenterer eg for at språkpolitikken i større grad må formulere tydelege kompetansemålsettingar (‘acquisition planning’), i tillegg til statusplanlegging.

The bilingual language control literature generally assumes that cross-language interference resolution relies on inhibition of the non-target language. A similar approach has been taken in the bidialectal language control literature. However, there is little evidence along these lines for proactive language control, which entails a control process that is implemented as an anticipation of any cross-language interference. To further investigate the possibility of proactive inhibitory control, we examined the effect of language variety preparation time, by manipulating the cue-to-stimulus interval, on parallel language activation, by manipulating cognate status. If proactive language control relies on inhibition, one would expect less parallel language activation (i.e., a smaller cognate facilitation effect) with increased proactive inhibitory control (i.e., a long cue-to-stimulus interval). This was not the case with either bilinguals or bidialectals. So, the current study does not provide evidence for proactive inhibitory control during bilingual and bidialectal language production.

The starting point of this paper is to the development of a relatively panoramic account of English and multilingualism in Thailand, illuminating the relationships among powers (trends), individuals and groups and their multilingual practices. Most previous studies draw upon arguments from the traditional monocentric model of nation-state multilingualism. Surprisingly, little is known about the polycentric forces influencing multilingual distributions. Methodologically, large-scale data sets are gathered through literature, documents, questionnaires, interviews and observing actual language behaviors. The ethnographic data gathered are analyzed through content analysis conventions. Challenging Smalley’s nationalist model, the central thesis of this paper is to present an alternative path, a polycentric model, to better understand Thailand’s multilingualism. There is a lack of adequate evidence within and across Thailand to show the existence of a universal and coherent hierarchy connecting different layers of language users. As such, this paper argues that Smalley’s universal hierarchy/dominance model has to compete with alternative models termed a ‘multiarchy’ or ‘multiarchies’ (a portmanteau of “multiple hierarchies”) where parallel language hierarchies co-exist with the orthodox one. Keywords: globalization; English; language use; multilingualism; Thailand

Cross-lingual text representations have gained popularity lately and act as the backbone of many tasks such as unsupervised machine translation and cross-lingual information retrieval, to name a few. However, evaluation of such representations is difficult in the domains beyond standard benchmarks due to the necessity of obtaining domain-specific parallel language data across different pairs of languages. In this paper, we propose an automatic metric for evaluating the quality of cross-lingual textual representations using images as a proxy in a paired image-text evaluation dataset. Experimentally, Backretrieval is shown to highly correlate with ground truth metrics on annotated datasets, and our analysis shows statistically significant improvements over baselines. Our experiments conclude with a case study on a recipe dataset without parallel cross-lingual data. We illustrate how to judge cross-lingual embedding quality with Backretrieval, and validate the outcome with a small human study., Comment: SIGIR 2021

EMI’s expansion in higher education has not been without growing pains. One unresolved problem lies in how to address English when it is not taught as a subject and is viewed as a means rather than an end; another involves how to maintain a local knowledge base and national identity when teaching through a non-local language at the tertiary level (Bowles and Murphy, English-medium instruction & the internationalization of Universities, Palgrave Macmillan, London 2020b; Coleman et al., TESOL Q 52:701–720, 2018; Pecorari and Malmstrom, TESOL Q 52:497–515 2018). To address the impact of “Englishization” that EMI might have on smaller European languages, the Nordic Council (2007). Deklaration om nordisk sprogpolitik 2006 [Declaration of a Nordic language policy]. Nordic Council of Ministers. Retrieved from: https://www.norden.org/da/publication/deklaration-om-nordisk-spraakpolitik) introduced a policy of “parallel language” use (or concurrent use of several languages), including strategies to prevent a single language (English) from devaluing or replacing local languages. The study outlined in this chapter takes a bottom-up view of the role EMI professors’ lived experiences and language ideologies can play in their enactment of parallel language policy and adoption of initiatives in a Nordic university. Framed on theories linking (monolingual) ideologies and policies, the findings highlight how the positioning of EMI professors aware of their students’ linguistic resources may shift in response to their views of pragmatic solutions to students’ language-related academic challenges. In the future, TESOL leaders must seek opportunities to align English and parallel language development in EMI and raise policymakers’ awareness of the role of multilingualism in internationalization overall.

This chapter describes the implementation and performance evaluation of Omni compiler, which is a reference implementation of the compiler for XcalableMP. For performance evaluation, this chapter also presents how to implement the HPC Challenge benchmarks, which is a benchmark suite for an HPC parallel language. The results show that the performance of XMP is comparable to that of MPI in many cases.

Scaling natural language processing (NLP) to low-resourced languages to improve machine translation (MT) performance remains enigmatic. This research contributes to the domain on a low-resource English-Twi translation based on filtered synthetic-parallel corpora. It is often perplexing to learn and understand what a good-quality corpus looks like in low-resource conditions, mainly where the target corpus is the only sample text of the parallel language. To improve the MT performance in such low-resource language pairs, we propose to expand the training data by injecting synthetic-parallel corpus obtained by translating a monolingual corpus from the target language based on bootstrapping with different parameter settings. Furthermore, we performed unsupervised measurements on each sentence pair engaging squared Mahalanobis distances, a filtering technique that predicts sentence parallelism. Additionally, we extensively use three different sentence-level similarity metrics after round-trip translation. Experimental results on a diverse amount of available parallel corpus demonstrate that injecting pseudoparallel corpus and extensive filtering with sentence-level similarity metrics significantly improves the original out-of-the-box MT systems for low-resource language pairs. Compared with existing improvements on the same original framework under the same structure, our approach exhibits tremendous developments in BLEU and TER scores.

Chapel is a parallel language designed to support productive programming from laptops to supercomputers. Unlike typical general-purpose approaches to large-scale computing, Chapel supports a global namespace and a global view of control. These make programming at scale more productive and natural by avoiding the need to utilize a Single Program, Multiple Data (SPMD) approach. Chapel is being developed as portable, open-source software and has demonstrated its ability to match or beat C/C++/MPI/OpenMP performance using code that's as easy to read as Python or Swift. In this talk, I will introduce some of Chapel's key features for those who are new to the language. However, most of the talk will focus on unique aspects of compiling Chapel. I will illustrate the transformations and analyses we use to implement Chapel's global view of data and control. I will also describe some optimizations that make use of Chapel's feature set to improve performance without sacrificing ease-of-use. In doing so, I will argue that when designed well, a productive language can enhance a compiler's ability to achieve performance and scalability rather than thwarting it. I'll wrap up by touching on some upcoming challenges for the Chapel compiler, including code generation for GPUs, reducing compilation time, and supporting interactive programming.

Objective To describe the spatio-temporal dynamics and interactions during linguistic and memory tasks. Methods Event-related electrocorticographic (ECoG) spectral patterns obtained during cognitive tasks from 26 epilepsy patients (aged: 9–60 y) were analyzed in order to examine the spatio-temporal patterns of activation of cortical language areas. ECoGs (1024 Hz/channel) were recorded from 1567 subdural electrodes and 510 depth electrodes chronically implanted over or within the frontal, parietal, occipital and/or temporal lobes as part of their surgical work-up for intractable seizures. Six language/memory tasks were performed, which required responding verbally to auditory or visual word stimuli. Detailed analysis of electrode locations allowed combining results across patients. Results Transient increases in induced ECoG gamma power (70–100 Hz) were observed in response to hearing words (central superior temporal gyrus), reading text and naming pictures (occipital and fusiform cortex) and speaking (pre-central, post-central and sub-central cortex). Conclusions Between these activations there was widespread spatial divergence followed by convergence of gamma activity that reliably identified cortical areas associated with task-specific processes. Significance The combined dataset supports the concept of functionally-specific locally parallel language networks that are widely distributed, partially interacting in succession to serve the cognitive and behavioral demands of the tasks.

In a parallel language environment it is important that teaching takes account of both the languages students are expected to work in. Lectures in the mother tongue need to offer access to textbooks in English and encouragement to read. This paper describes a preliminary study for an investigation of the extent to which they actually do so. A corpus of lectures in English for mainly L1 English students (from BASE and MICASE) was examined for the types of reference to reading which occur, classified by their potential usefulness for access and encouragement. Such references were called ‘intertextual episodes’. Seven preliminary categories of intertextual episode were identified. In some disciplines the text is the topic of the lecture rather than a medium for information on the topic, and this category was not pursued further. In the remaining six the text was a medium for information about the topic. Three of them involved management, of texts by the lecturer her/himself, of student writing, or of student reading. The remaining three involved reference to the content of the text either introducing it to students, reporting its content, or, really the most interesting category, relativizing it and thus potentially encouraging critical reading. Straightforward reporting that certain content was in the text at a certain point was the most common type, followed by management of student reading. Relativization was relatively infrequent. The exercise has provided us with categories which can be used for an experimental phase where the effect of different types of reference can be tested, and for observation of the references actually used in L1 lectures in a parallel-language environment.

To improve productivity for developing parallel applications on high performance computing systems, the XcalableMP PGAS language has been proposed. XcalableMP supports both a typical parallelization under the “global-view memory model” which uses directives and a flexible parallelization under the “local-view memory model” which uses coarray features. The goal of the present paper is to clarify XcalableMP’s productivity and performance. To do so, we implement and evaluate the high performance computing challenge benchmark, namely, EP STREAM Triad, High Performance Linpack, Global fast Fourier transform, and RandomAccess on the K computer using up to 16,384 compute nodes and a generic cluster system using up to 128 compute nodes. We found that we could more easily implement the benchmarks using XcalableMP rather than using MPI. Moreover, most of the performance results using XcalableMP were almost the same as those using MPI.

The article presents a study of a five-year strategic project carried out at the University of Copenhagen from 2013 to 2018, with a focus on the role of languages in educational programs across the university. The project was based on needs analyses with students and program directors, and its main output was the development of diverse, locally appropriate models for integrating content and language. Whereas the university’s language policy of parallel language use prioritizes the use of English and Danish for academic purposes, the study underlines the need to include not only these two languages but also to build on other languages for subject specific purposes. The use of languages such as German, French, Latin, Arabic, and Chinese in study programs are motivated by traditional practices as well as by new orientations in the academic labor market. Their value is grounded in disciplinary logic, and they are further developed through bottom-up initiatives. However, the study shows that the formal inclusion of these languages in a university language policy also requires an explicit support from university management.

The nature of syntactic planning for language production may reflect language-specific processes, but an alternative is that syntactic planning is an example of more domain-general action planning processes. If so, language and non-linguistic action planning should have identifiable commonalities, consistent with an underlying shared system. Action and language research have had little contact, however, and such comparisons are therefore lacking. Here, we address this gap by taking advantage of a striking similarity between two phenomena in language and action production. One is known as syntactic priming-the tendency to re-use a recently produced sentence structure-and the second is hysteresis-the tendency to re-use a previously executed abstract action plan, such as a limb movement. We examined syntactic priming/hysteresis in parallel language and action tasks intermixed in a single experimental session. Our goals were to establish the feasibility of investigating language and action planning within the same participants and to inform debates on the language-specific vs. domain-general nature of planning systems. In both action and language tasks, target trials afforded two alternative orders of subcomponents in the participant's response: in the language task, a picture could be described with two different word orders, and in the action task, locations on a touch screen could be touched in two different orders. Prime trials preceding the target trial promoted one of two plans in the respective domain. Manipulations yielded higher rates of primed behavior in both tasks. In an exploratory cross-domain analysis, there was some evidence for stronger priming effects in some combinations of action and language priming conditions than others. These results establish a method for investigating the degree to which language planning is part of a domain-general action planning system.

Energy efficiency is one of the key aspects of modern software. Therefore, we present a tool for measuring the energy consumption of Erlang programs. Using this tool, we measured the energy consumption of different basic language elements, such as data structures, higher-order functions and parallel language constructs. Based on the results of our measurements, we present refactorings that may help to decrease the energy consumption of Erlang software. The refactorings are part of the well-known RefactorErl static analyser and transformer framework for Erlang.

Various partitioned global address space (PGAS) languages capable of providing global-view programming environments on multi-node computer systems have been proposed to improve programming productivity in high-performance computing. However, several PGAS languages often require a detailed description of the remote data access, similar to descriptions used in message passing interface one-sided communications. Some PGAS languages have limitations pertaining to remote data access and recommend their local-view programming models, rather than the global-view ones, due to performance-related reasons. In this study, we propose SMint, which is an application programming interface that provides a global-view programming model with a software distributed shared memory mSMS as the runtime. Using stencil computation as a typical processing method, the performance and programmability of SMint have been compared with those of XcalableMP and Unified Parallel C, which are well-known examples of PGAS languages based on the C language. It was found that SMint achieved the best performance under the ideal global-view programming model.

The main task of a university library is to serve the teaching, learning and research needs of staff and students in higher education. However, the lecturers, students and researchers need more than a collection of sources of information provided by librarians. Globalisation in general and the universities' internationalisation policy have led to a large international researcher and student population and thus increased requirements when it comes to language proficiency, intercultural competence and teaching and learning in a multicultural environment. As a result of internationalisation and the principle of widening participation, the student population has become more heterogeneous. In 2016, the participation rate for Norwegian students (age range 19-24) in higher education in Norway was 35.1%. The Centre for Academic Language and Practice is located in the University Library at OsloMet – Oslo Metropolitan University. The centre’s mission is to contribute to the quality of education and professional development by increasing the academic literacy of staff and students. Academic literacy is defined as receptive and productive skills to participate, learn, guide, teach and publish in an academic context. The Centre for Academic Language and Practice offers linguistic and pedagogical support and conducts research in the fields of academic literacy and pedagogy targeting students in higher education. This paper will look into the history and the portfolio of the Centre for Academic Language and Practice and illustrate why its competence and type of work belongs in a university library.

This paper presents ParaVis, a visualization library designed to aid programmers' understanding of their parallel programs and to help them identify bugs with parallelization. ParaVis is particularly targeted for programmers who are first learning parallel programming or learning a new parallel language. It provides easy-to-use C and C++ interfaces to create 2D animations of parallel computation that help programmers understand parallel data decomposition patterns. These visualizations are also helpful in illustrating errors in parallel programs. Additionally, because students often find visualization fun, the use of our library often results in students developing interesting extensions to problems, thus promoting a deeper understanding and richer experience with parallel computing. Currently we provide support and sample implementations for pthreads, OpenMP, CUDA, and sequential applications. To test its effectiveness for parallel computing education, we deployed ParaVis for lab assignments in both intermediate and upper level courses. We present example applications, and evaluate the use of the library across our undergraduate CS curriculum.

One of the most problematic perspectives of translation phenomenon is the cultural gap between the source language and the target language (Yang, 2010). This gap can be ideally filled through telecollaboration which provides internationally dispersed language learners in parallel language classes with cost-effective access to, and engagement with, peers who are expert speakers of the language under study (Belz, 2005). To investigate the effect of telecollaboration on the quality of translation of culture-bound texts, the current study was conducted on 64 Iranian undergraduate students of English translation at a university in Iran. Instruments used in the study consisted of three texts containing news excerpts from Voice of America (VOA). The study consisted of three phases: 1) assessing quality of translation of culture-bound texts, 2) random assignment of participants to two groups: one merely receiving cultural instruction while the other being linked to native English speakers through LinkedIn alongside receiving cultural instruction, and 3) assessing quality of translation of culture-bound texts immediately and two months following treatment. The results of mixed between-within subjects analysis of variance revealed the significant positive effect of telecollaboration on developing quality of translation of culture-bound texts and sustaining the attained knowledge. The pedagogical implications of the findings suggested incorporation of cultural components of source language society into translation courses and providing opportunities for translation students to be exposed to authentic and intensive source language culture through telecollaboration.

We investigated the spontaneous activation of phonologically related words in high and low proficient Hindi–English bilinguals during spoken word processing in an eye-tracking study. Participants listened to spoken words in L1/L2 and looked at a display (consisting of line drawings of phonological cohort of the translation equivalent of the spoken word and unrelated distractors). Both the groups were quick in orienting their attention towards the competitor with the onset of the spoken word. Furthermore, high proficient bilinguals showed higher and earlier activation of the competitor compared to low proficient bilinguals. Cross-language activations were higher in the L2–L1 direction for both the groups. The results strongly suggest language non-selective access of translation in Hindi–English bilinguals in both language directions. We discuss the results with regard to the predictions of the bilingual language processing models and the effect of language proficiency on conceptual access during li...

Fuzzy Logic System (FLS) is an efficient method to solve engineering problems. However, the training of a Fuzzy Logic System is a time-consuming task. Optimization Algorithm can be used to optimize the rule base of any FLS. Out of Type-1 FLS and Type-2 FLS, the type-2 found to be more effective to deal with noisy data. Due to their computational requirements Interval Type-2 (IT2) has been preferred over General Type-2. Whale Optimization Algorithm (WOA)has been introduced recently. The algorithm has been tested on different engineering problems and is found to be more effective. General Purpose Computing using graphics Processing Unit (GPGPU) is a new way to solve compute intensive problems on Graphics Processing Unit (GPU). CUDA-C is a parallel language that can be used to execute parallel code NVIDIA GPU. This paper integrates IT2 FLS, WOA and Processing Power of GPU. A toolbox is proposed that can be used to optimize the rule base in parallel. The toolbox provides both the implementations, i.e. serial and Parallel. FLS WOA Toolbox is design in such a way that user can pass parameter dynamically according to their need without interfacing with the code.

Epub 2019 Mar 28. In the current study, we set out to investigate language control, which is the process that minimizes cross-language interference, during bilingual language comprehension. According to current theories of bilingual language comprehension, language-switch costs, which are a marker for reactive language control, should be observed. However, a closer look at the literature shows that this is not always the case. Furthermore, little to no evidence for language-mixing costs, which are a marker for proactive language control, has been observed in the bilingual language comprehension literature. This is in line with current theories of bilingual language comprehension, as they do not explicitly account for proactive language control. In the current study, we further investigated these two markers of language control and found no evidence for comprehension-based language-switch costs in six experiments, even though other types of switch costs were observed with the exact same setup (i.e., task-switch costs, stimulus modality-switch costs, and production-based language-switch costs). Furthermore, only one out of three experiments showed comprehension-based language-mixing costs, providing the first tentative evidence for proactive language control during bilingual language comprehension. The implications of the absence and occurrence of these costs are discussed in terms of processing speed and parallel language activation. (PsycINFO Database Record (c) 2019 APA, all rights reserved) This project has received funding from the European Union’s Horizon 2020 research and innovation programme under the Marie Sklodowska-Curie grant agreement No 706128. This research was also supported by grants ANR-11-LABX-0036 (BLRI), ANR-16-CONV-0002 (ILCB), and ANR-11-IDEX-0001-02 from the French National Research Council (ANR).

We have designed a parallel language called MULTI- ML for programming MULTI-BSP algorithms in ML. The MULTI- BSP model provides a tree-based view of nested components of hierarchical architectures. The structure and abstraction brought by MULTI-BSP allows for portable parallel programs with scalable performance predictions, without dealing with low-level details of architecture. The MULTI-ML language is essentially based on the concept of multi-functions which are recursive functions used to execute perform recursions on the nested sub¬components through the MULTI-BSP (hierarchical) architecture. In this paper we design a generic compilation scheme of the MULTI-ML language dedicated to parallel machines performing data exchanges using synchronous requests, derived from any ML compilation. This ensures that the implementation follows the semantics, allowing greater confidence in the safety of execution.

When developing applications on high-performance computing (HPC) cluster systems, Partitioned Global Address Space (PGAS) languages are used due to their high productivity and performance. However, in order to more efficiently develop such applications, it is also important to be able to combine a PGAS language with other languages instead of using a single PGAS language alone. We have designed an XcalableMP (XMP) PGAS language, and developed Omni Compiler as an XMP compiler. In this paper, we report on the development of linkage functions between XMP and {C, Fortran, or Python} for Omni Compiler. Furthermore, as a functional example of interworking between XMP and Python, we discuss the development of an application for the Graph Order/degree problem. Specifically, we paralleled all of the shortest paths among the vertices searches of the application using XMP. When the results of the application in XMP and the original Python were compared, we found that the performance of XMP was 21% faster than that of the original Python on a single CPU core. Moreover, when applying the application on an HPC cluster system with 1,280 CPU cores of 64 compute nodes, we could achieve a 921 times better performance than that on a single CPU core.

In the article a technology is considered which aims at creating architecture-independent parallel programmes based on the data-driven functional paradigm. A proposed toolkit provides the translation, execution, debugging, optimisation and verification of programmes. A programme in a data-driven functional parallel language is translated into the data-flow graph (which describes the data dependencies of an implemented algorithm) of the programme. On the basis of this representation, the control-flow graph (which defines the organisation of computations) is generated. Both graphs allow to carry out various optimising transformations. The resulting data-flow graph is also used for the formal verification of the programme. A computation process is considered as a cooperation of the control-flow graph and the data-flow graph. The execution of data-driven functional parallel programmes is carried out by a special interpreter (event machine), which consist of a number of event processors controlled by a special manager.

Within the last decade, many universities in non-Anglophone Europe have developed new language policies in order to sharpen their international profile. Since 2008, the University of Copenhagen has been guided by a parallel language strategy aiming at a balance between English and the national language Danish. In addition, the university has recently decided to develop a new strategy which introduces several other languages as potentially relevant for students in their academic programs and when they target the Danish or the international labor market. At an institutional level this is seen as a way of promoting quality of programs and securing the university’s position in the international market. However, neither all languages nor all four language skills are equally important for all purposes. Therefore a careful needs analysis is an important first step in order to decide which languages and language skills to include in the actual planning. At the University of Copenhagen the needs analyses involve students as well as instructors and members of study boards. These different participants in the academic community are often motivated by different perspectives on teaching and learning issues, e.g., on curricular priorities and traditions, but also hold individual preferences. Based on a language-in-education perspective, this paper will focus on the different organizational levels involved when developing and implementing a new language strategy and the possible alignments and mismatches between these levels. As an example it will look at the role of the French and German languages as mentioned by different stakeholders and compare this with the situation for English.