Teaching and researching literary translation in the digital context: PEnPAL in trans as a case-study

The digital reinvention of literary studies within literary translation teaching and research informs the PEnPAL in Trans project. This inter-institutional venture joins higher education agents and researchers in Translation Studies, Literary Studies and Linguistics. Elaborating on the notions of process-oriented education and “social constructivism” (Kiraly, 2000), PEnPAL in Trans has developed a specific awareness of the literary translator’s “expert action” (Jones, 2011). Drawing on a project-based philosophy of translation training, it envisions the translated anthology as a collaborative format with potential in the digital environment. The database on English-Portuguese transfer problems under development combines the advantages of translation manuals and example-driven tools as translation memories. Thus, it will constitute a categorized database of examples from hard-to-translate texts together with their translation(s) and translation strategy(ies). This database will be accessible online, thereby providing a public tool on the English-Portuguese language pair

of the literary translator's "expert action" (Jones, 2011).Drawing on a project-based philosophy of translation training, it envisions the translated anthology as a collaborative format with potential in the digital environment.The database on English-Portuguese transfer problems under development combines the advantages of translation manuals and example-driven tools as translation memories.Thus, it will constitute a categorized database of examples from hard-to-translate texts together with their translation(s) and translation strategy(ies).This database will be accessible online, thereby providing a public tool on the English-Portuguese language pair.Keywords: applied literary translation; digital translation studies; collaborative translation; translated anthologies; data-base of translation problems.
he specificities of the literary text, especially the need to transpose polysemy, connotation and style, present particular challenges to the translator.To the common mind these are dissociated from recent technologies beyond the use of word processors, electronic dictionaries and online search engines for getting information.The influence of recent digital media and tools is mostly acknowledged as far as either research or specialized translation are concerned, and we generally think about IT translation tools and software as being in connection with technical or scientific translation.However, the digital revolution has definitely made its way to Translation Studies (see, among others, Baker, 1993;Laviosa, 2002;Rosa, 2003), practice (Craciunescu, Gerding-Salas & Stringer-O'Keeffe, 2004;PACTE, 2011) and teaching (McEnery & Wilson, 2001;Beeby, Inés & Sánchez-Gijón, 2009).And, just like Translation Studies and specialized translation, on the one hand, and literature and Literary Studies, on the other hand, literary translation should not shun the development of digital media, tools and skills.Recognition of this fact is what gave rise to the PEnPAL in Trans (Portuguese-English Platform for Anthologies of Literature in Translation) project.This project is used here as a case-study for considering some of the effects of the digital revolution upon the teaching and research of literary translation and for discussing the challenges posed by the construction of an online database on English-Portuguese transfer problems.

PEnPAL in Trans: a short presentation
PEnPAL in Trans is a project about applied literary translation that began in 2011.It constitutes an inter-institutional endeavour that provides support for literary translation teaching, and eventually literary translation practice.For this purpose, an online platform (http://penpalintranslation.com) and an accompanying blog (http://penpalintranslation.blogspot.pt)were created, which promoted collaborative work on a collectively built anthology of source texts.These texts have focused so far on narratives of displacement and cultural and interlingual exchange between Portuguese-speaking and English-American spaces.The didactic goals of the project and the creation of the online platform and blog have also demanded and promoted innova-T tive research in various academic fields, namely Translation Studies, Literary and American Studies, and Comparative and Computational Linguistics. 1hus, PEnPAL in Trans constitutes "a scene of encounters", which has been identified by Alan Liu (2008) as one of the major characteristics of Digital Humanities (henceforward DH).In fact, it depends on and promotes encounters between literary translation teachers and colleagues of the same or other universities (namely the University of Lisbon, the New University of Lisbon, the Catholic University of Lisbon and, until 2014, the Lusophone University), with authors and, hopefully, with publishers; furthermore, the project relies on and encourages (virtual) encounters between researchers from the same and other national and foreign universities and from various academic fields; and, last but not least, it is sustained through encounters of literary translation students with classmates, students from other classes and universities, researchers, authors and potentially the whole community.PEn-PAL in Trans therefore embodies the main dimensions of DH, which have been understood, more than once, as "both the research and the teaching" (Kirschenbaum, 2010;Bobley, 2011); it also cherishes a "co-creation" profile which, despite its tradition in translation (O 'Brien, 2011: 17), was taken several steps further within DH, since: [literacies] now move front and center inasmuch as the advent of Digital Humanities implies a reinterpretation of the humanities as a generative enterprise: one in which students and faculty alike are making things as they study and perform research, generating not just texts (in the form of analysis, commentary, narration, critique) but also images, interactions, cross-media corpora, software, and platforms.(Burdick et al., 2012: 10) Although little attention has been paid to the impact of the digital revolution on literary translation thus far, PEnPAL in Trans is not a stand-alone case.Collaborative literary translation has been encouraged on some sites, whether connected to the world of literary dissemination and journals (e.g.Poetry Translation Centre, poetrytranslation.org;Modern Poetry in Translation, mptmagazine.com)or to universities (e.g. the British Centre for Literary Translation at the University of East Anglia, bclt.org.uk; a distance-learning project at Universidade Aberta, Portugal, on the translation of a Mark Strand anthology, odisseia1.univ-ab.pt/cursos/Poetic_Strands).There are also useful sites with English-Portuguese comparative translation corpora-such as Compara (linguateca.pt/COMPARA)and COMET (comet.fflch.usp.br).However, we are not aware of other examples of translation education platforms that have the level of complexity we aim at, specifically: collaborative learning environments, aids to translation research and practice, dissemination of cultural products and tools to help competence in language transfer.

A collaborative anthology
Burdick et al. describe DH as "conspicuously collaborative and generative" (2012: 3).This characteristic, which is one of the reasons why the DH question has even disrupted some academic traditions (Andrade, 2015), was acknowledged and imported by PEnPAL in Trans.
Accordingly, the first goal of the project was the construction of a collaborative online anthology of texts related to the literary exchanges between Portugal and North-America in diasporic situations.These are not only texts written in English by a growing number of Portuguese-American authors, but also texts produced by North-American writers living in Portugal that focus on narratives of displacement and cultural and interlingual exchange.This collectively-built anthology has already been translated, also collectively, by more than 300 undergraduate and graduate students attending various higher education programs in Portugal, who have made use of online tools and PEnPAL's digital platform in particular.The anthologization and translation of source texts is now close to attaining its goal, namely the publication of the collective anthology Nem Cá nem Lá -Portugal e América do Norte entre Escritas (Neither Here nor There: Writings Across Portugal and North America), funded by Fundação Luso-Americana para o Desenvolvimento, by means of the Alberto de Lacerda Translation Award 2013.
The selection of texts for the anthology, which was a collaborative effort by the researchers and teachers involved in the project, took into consideration not only the diversity of its subject matter, but also the variety needed for learning purposes.It also includes source texts of various text typologies: short stories, novels, autobiographies, poems, drama and children's literature.This selection will no doubt broaden the target literary system, enhance its transatlantic dimension and contribute to the literary representations of our current diaspora.
Considering that Burdick et al. wonder whether "computational and digital environments [can] be designed to capture the fluidity of an intercultural dialogue between diasporic peoples" (2012: 9), we intend to tackle this question in two ways: by studying, teaching and practising the translation of dias-pora literature, and by designing a digital platform that accounts for this "intercultural dialogue".This platform will include information on authors, a discussion of literary and intercultural themes, and finally a database of collaborative input on the difficulties of translation raised both by the marked language of literature and by the superimposition of languages (Berman, 1985: 285) enhanced by Portuguese-American transits.

Collaborative translation
It is important to document the methodology behind the didactic collaborative environment in which the whole project is founded.This methodology is divided into four main stages that may overlap: 1. Collaborative translation; 2. Questions and answers; 3. Proofreading; 4. Lessons learnt.
Firstly, in the "Collaborative translation" phase, students are asked to translate a text collaboratively, and to share their doubts and strategies/tactics on the project's blog (penpalintranslation.blogspot.pt).This way, all classes that are working on the texts for the anthology can benefit from such input and students can learn how to negotiate different translation options and to explain the rationale behind their options.One of the biggest difficulties that students face, in our experience, is to express the motives behind their choices, often adopting a non-professional and non-academic discourse: "because it sounds better" is the common reason given for choosing one option over another.This supports evidence that students and novice translators are less aware of translation problems than professional translators (Gerloff, 1988;Jääskeläinen, 1999), while reinforcing conclusions from work with think-aloud protocols suggesting that translators need to be trained to express what goes on in their minds while translating (Hansen, 2003).
Secondly, in the "Questions and answers" phase, authors give feedback to students and answer their questions, either in person, if they can come to class, or through the blog or e-mail.Students can therefore benefit from the unique experience of having their interpretation questions answered by a living author and also understand that the authors themselves are seldom aware of the translation problems that their texts pose.Such interaction has also proved to provide authors with new insights into their own semantic clusters and stylistic choices and has brought about opportunities for emerging Luso-American authors to showcase their work and present it before their heritage culture. 2hirdly, in the "Proofreading" phase, teachers work on the translation, providing other translation options and correcting errors.As mentioned above, in PEnPAL in Trans this proofreading process has comprised teacher to teacher interaction, with project members bringing into their classes revision issues raised by the interchange.
Fourthly, in the "Lessons learnt" phase, the parties involved, with the supervision of those more committed to the database design, have gathered to systematize the solutions found and the strategies used in order to feed our collaborative learning digital tool for English-Portuguese translation problems.

The translation database
Our endeavour in collaborative translation through digital support is aimed not only at creating the translation of a body of texts relevant to intercultural perceptions (section above), but also at structurally storing the translation problems encountered in the process of creating these translations, thus exploring the possibility of systematization and guided search in a database.
The latter objective is particularly relevant for literary translation from English into European Portuguese because of the absence of any structured contrastive stylistics between this language pair.Moreover, the digital resources available thus far -e.g.Linguateca's Compara and USP's 'Comet', mentioned above, or systems based on translation memories such as Linguee or Glosbe -are almost exclusively collections of concrete examples taken from translations, which as such only provide information when at least part of the translation that is being considered literally coincides with a previous example.Therefore, whenever the translator is attempting to translate something that has not yet been translated, or not yet translated satisfactorily, such resources are not very helpful.Given the rich combinatorics of language and the tendency for non-standardization, particularly marked in literary texts, the chances that a specific phrase has already been translated before are virtually non-existing.This is the reason why computer assisted translation is often deemed as not being helpful for literary translation.When a translator faces a problem there is a need to look at examples of similar cases, rather than consider cases that only match part of a literary text.In order to be able to make translation choices based on such similar cases, it is not enough to just have a list of examples: an explanation is needed to help in the decision process on whether the solution adopted in those examples is suitable for the case at hand.The database we have opted for in our project is therefore much closer to a translation manual than to a mere collection of translation examples.In this section, we describe the design and concept behind the database of translation problems that aims to tackle the abovementioned difficulties.We will focus mainly on two aspects of the database: on the one hand, the way it allows users to find similar examples, and on the other hand, how it helps the translator to elaborate a rationale for reaching a solution.The database can be found at the following address: http://alfclul.clul.ul.pt/penpal.

Finding examples
In the database, example phrases are stored together with their translations as well as a discussion and classification of the problem in that phrase.The database only stores problematic sentences (instead of all translated sentences as in the case of a translation memory), and each entry revolves around a problem in that phrase.Since a single sentence can contain several unrelated translation problems, a database entry always relates to a single problem in each sentence, meaning that phrases with several translation problems are listed more than once.When attempting to translate a sentence with a new translation problem, the database can be used to locate entries that discuss a similar translation problem.
There is no obvious notion of what a similar example might be to any specific problematic phrase, nor is there a unique way in which users are likely to look for similar examples.That is why our database design contemplates several ways to reach cases that might be similar.The first and most traditional notion of similarity is an organization of translation problems comparable to that found in traditional translation manuals.Problems are classified into types of problems that are categorized and related to each other.Since the creation of a classification of problems is not a trivial matter (see section 4.3), the organization of translation problems in our database is richer than a traditional hierarchy, with several types of relations between various problems.The first relation is that each problem has a super type, hence creating a hierarchical structure.But on top of that, translation problems are also directly linked to other similar problems, or to the different terms that address the same problem, and besides browsing through a structured list of problems, you can directly search a problem by term(s).The second way to access a relevant example is by the built-in links between similar examples.These links are of two types: either simple direct links (which can be part of the description of the problem), or examples that have the same keyword.Keywords are loosely defined as any term that is deemed helpful in identifying relevant examples, which are neither names for translation problems nor strategies.Apart from being linked between examples, the user will also be able to look for keywords directly through a search function (in progress).In many cases, keywords will be taken from the translation units or from a representative semantic field.
The third way to access examples is by browsing for them by the strategy used in the translation of the example.3This is of course the less obvious way of access, since it implies making an educated guess of how best to translate the case at hand, so it already involves finding a way to reach a solution (see section 4.2).
To illustrate how these three ways of getting to relevant examples work, take the following example from our database: In this example, two translations are given for the same source text.The relevant problematic portion of the phrase is "a few things that glitter, decorations they call them", and the rest of the original text is present to make the (Christmas) context clear.Furthermore, even with the additional context, it is not necessarily clear from the text itself that the problem is due to the fact that at the time of writing there is a cultural difference as to how Christmas is celebrated in California in comparison with the Portuguese heritage culture.This clarification comes under "Description".The whole context presented in the example is loosely called a "snippet" in the database, whereas the bold part of the example is intended to be the translation unit where the problem resides (in and by itself a problematic notion). 4 The problem in this example is twofold: on the one hand, there is a "cultural lexical gap" (Janssen, 2012) as the reality named by "decorations" was still not captured by the Portuguese lexical equivalent, "enfeites", at the time of writing.On the other hand, in this literary excerpt, language is calling attention to itself, and in particular to how it fictionalizes the speech of a foreign tongue (the "they" referred to are American English speakers; the speech enunciator in the example is probably speaking Portuguese).The indication of type(s) of problems can represent both different types of problems at stake and near-synonymous terms for the same problem.By clicking on one of the problem types listed in the example -say, "metalanguage of culture specific terms" -the user will access the entry in the database of translation problems corresponding to the problem.That entry provides some data about the translation problem (see Figure 2): (1) a description of what that kind of problem involves; (2) relations to similar types of problems, and to superordinate categories of these -in this case, "metalanguage", which in turn will be linked to other types of metalanguage problems; (3) a list of all examples in the database that are of the type "metalanguage of culture specific terms".
It is important to stress that our database cannot be achieved through direct mark-up of actual corpora, as is done for instance in databases on translation errors such as MeLLANGE (Kübler, 2008), or data-driven translation websites such as Glosbe.Firstly, the marking functionality is mostly apt for lexical choices but fails with larger and diffuse translation units, and in cases where there is no direct correspondence between source and target segments.Secondly, merely providing the source text and the translation with a keyword is not always helpful to the user.It is necessary to have the possibility to complement each example with a short description of the problem at hand, and potentially provide various possible translations, with a short decult adherence by textual segmentation in translation memories.PEnPAL's database can provide grounds for research in terms of relevance and extension of translation units, and of what suffices to delimitate a problem (generally not just the coupled segments, but also description of co-and context).scription of their respective motivations and the consequences following a reflexive decision-led approach.

Reaching a solution
As mentioned above, it is possible to see what examples have already been translated, but in the type of database that aims to provide hindsight for acquisition of transfer skills and even for professional translations, what is mostly needed is an explanation of why a specific example is problematic in the first place, which strategy was adopted for the translation of the example, what rationale played a role in the choice of the adopted translation, and which alternative options could be considered or were actually adopted by other translators.For this, in the PEnPAL database each example comes with the following information: the kind of problem raised by the example; a link to an explanation of that type of problem; one or more translations for the example; the strategies used in each translation, along with a link to an explanation of that strategy; finally, any additional information that is relevant for the specific example, its context, and the motivation behind the strategies adopted.All the information listed above helps to associate examples with a problem at hand.More often than not, none of the words in a phrase that the user is trying to translate will correspond with any of the examples in the database.Therefore, the most expedient way to find a solution is to browse through problems deemed to be similar, an exercise that will in itself develop an awareness of translation problems at stake, in turn enhancing the capacity to browse through the database.In a sense, once a selection of similar examples has been identified, finding a solution consists of identifying which of the strategies adopted in those examples is the most suitable for the task at hand.And in this selection process, the description of the rationale in those examples helps to build translation competence.Browsing through strategies, on the other hand, builds the students' cognitive agility in the face of the complexity of translation, and it is also a form of reaching groups of similar examples where the same strategy has been applied (Figure 3).Therefore, the database can be useful even when not in the process of translating an actual text.For both students and professionals of translation, it can be read and used like a translation manual, but much more exampledriven.For any interested individual (since it is meant to be open-source) and for related purposes of research or teaching (especially the acquisition of English as a second or foreign language), access through the categories of the navigation menu -problems, strategies, bibliography -or a keywordsearch will disclose information on contrastive features, whether structural, conventional or cultural, of the language pair, not found in other digital translation tools.

The problem of a taxonomy of problems
As described in the previous section, problems are ordered into an organized classification.The creation of a taxonomy is, however, not a linear process.
First of all, the normative translation manuals that could be of use are rather few, since, as they generally are dependent on the languages they contrast, they work with variable degrees of translatability.This is not to disregard the value of functionally structured classifications, dividing problems per area of linguistics or textual analysis, such as those found in introductory books to translation (e.g.Nord, 1991;Munday, 2012; or, specifically on literary issues in translation, Landers, 2001); in fact, the general typology we arrived at as a basic agreement for metalanguage between PEnPAL members (Figure 4, below) is loosely based on those sources.However, to have a systemic notion of translation problems that were typical of the English-Portuguese transfer, and how to describe and characterize them, we need to consult, on the one hand, comparisons between English and other Latin languages, and, on the other, mine through the case-studies for specific issues concerning English and Portuguese (e.g.Rosa, 2003;and Pinto, 2010) and the corpus-driven approaches that face the issues of non-alignment of source and target in problematic instances, such as Diana Santos's attempt at mapping Portuguese and English language models regarding tense and aspect (2004). 55 While Diana Santos's approach insists on the necessity of corpus linguistics delving in literary texts as potential problematic instances highlighting culture and performance-related aspects, the list of uses for a corpus she provides in a later study (2008: 51) makes clear that in most respects our database functions in the opposite direction of comparative corpora: corpus-driven systems are designed to flash out the most common translation patterns that often serve to legitimize a previous hypothesis.PEnPAL uses a database that consists of instances where literal translation does not work, and it is not the unanalyzed corpus that drives the database, but the motivations and possible solutions of the intricate literary cases.On another note related to what may be drawn from traditional contrastive manuals, on the one hand, and the existing digital tools for translation, especially translation memories, on the other, it is worth Another issue with any description and categorization is, of course, variation in terms and the theoretical fields in which this occurs.Terminology (not to mention academic consensus on what even constitutes a translation problem) 6 differs greatly according to the different backgrounds and interests of researchers.Moreover, the categorization and hierarchy depends on a structural approach that is difficult to reconcile with the fact that a translation problem will more often than not imply several superordinates of traditional grammar -after having created our own typology for PEnPAL, based on the agreement of researchers from different scholarly backgrounds (Figure 4), it became clear that it would be applicable only through a structured interface that could call up and link several layers/levels at once. 7 A final momentous difficulty is the fact that learners, students, or translators in need are rarely aware of the possible correct terms and classifications and subclassifications for what they need to solve, though the acquisition of metalanguage is certainly a requirement for a database like ours to furnish more than a collection of samples of translated phrases.
noting that the latter disregard the differential factor identified by Jakobson in the assertion that "any comparison between two languages implies an examination of their mutual translatability" (1971[1956]:168; see also Toury, 1986Toury, : 1119)), whereas the former rely perhaps too heavily on contrastive linguistics.This is an aspect to take into consideration if we wish our database to provide a transferable setup for other language pairs. 6Following Nord (1991: 151), the PEnPAL team adopted a pragmatic notion of "translation problem" as comprising anything the student flags up as problematic in the translation task. 7The scheme we arrived at as a consensual typology to mark translation problems in PEnPAL's database reflects the interdisciplinary constitution of the team: the interintra and context distinction being largely derived from Literary Studies (e. g.Genette's narratology), the classic categories of lexicon, semantics, morphology and syntax taken from the classical divisions of grammar in linguistics, and concerns with matricial aspects, typology and conventions drawn from functional approaches to Translation Studies.

Collaborative construction
Despite the various advantages of a structured database of translation problems, one of the major drawbacks is that the creation of this type of database is very labour intensive.It does not only require providing a sentence and its translation, but it requires several additional manual steps: 1) identifying the type of problem the sentence presents and classifying that problem, 2) identifying the solution used for the problem in the translation and classifying that solution (not only for the chosen translation, but potentially for several alternative translations as well), and 3) describing the strategy(ies) employed, along with the rationale for its(their) adequacy to the given problem.And, on top of that, in the initial phases of constructing the database there will be new types of problems that were not yet contemplated in the problem taxonomy, for which it will be necessary to provide a description, and link it to the rest of the taxonomy.It is also worth mentioning that the corpus that feeds the PEnPAL database is particularly problematic in terms of finding correspondences and equivalents between source and target languages, as it must represent an additional third language -the heritage -creating heterolingualism (Meylaerts, 2006) in the source text but not necessarily so in the target, along with several other difficult representations of cultural and symbolic exchanges.Of course, this choice of corpus was also deliberate, as our research particularly wants to focus on this kind of translating difficulties, and additionally we deem that the complexity of these will allow a fair experiment on whether the goal of a database of transfer problems is feasible and helpful.
It is because of this labour intensive process that, at the time of writing, the structure of the database has been fully established, or at least only minor issues remain, but the number of entries in the database is still very modest.One of the ways in which the database is very useful is that it provides a very rich tool for students of Translation Studies.And, ideally, it would also be possible for students to add entries to the database.We have ample experience with students providing the type of description required for the database, but these descriptions were always provided in the form of a blog (penpalintranslation.blogspot.pt),with no internal structure and presenting each example discussed as just a piece of text.It is our aim for the wiki-style edition of content that we have been perfecting to reach a point where students can add examples and problem descriptions directly to the database.Apart from strict requirements on the usability of the interface, one of the major challenges in allowing this is how to assure the quality of the material in a collaborative environment.Currently, the problem taxonomy is still insufficiently complete to allow students to add content.Before we can reach the stage of transferring user knowledge to students there has got to be an investment on expert feeding of the database, which can also guide us in necessary improvements.
Thus far, the experiment has shown its potential in terms of research for project members who had to identify what needed to be collated as problems or strategies or keywords, in order to guide the decision-making of a translator in need, while at the same time seeking insights into the cognitive approach to contrastive linguistics.Certainly, thinking in terms of the applications of the digital to subjective and largely empirical matters of Humanities, especially when this involves archive, data treatment and modelling for interactive use, is an exercise that opens up the scope of teaching and research.

Conclusion
As we have shown in this paper, PenPAL in Trans attempts to bring novel uses of digital tools to the field of literary translation: it provides a platform with which literary texts can be translated collaboratively, where online tools are used as a means for translators to communicate their doubts and considerations amongst each other, and where teachers and authors can provide relevant feedback in the process.Hence, this online platform provides a digital record of the difficult-to-translate phrases encountered by the translators.Furthermore, the relevant parts of this digital record are then transferred into a structured database of real cases of translational difficulties, supplemented with translations as well as explanations about the motivation for the choice of translation in each particular case.
Although, given its labour-intensive nature, progress has been slow, we believe that the resulting database provides a unique way for translators to gather insight into the decision process of translational difficulties.This in turn provides a unique opportunity for translators to look for similar cases when facing a translation problem, and helps them take into consideration the decisions of others in those similar cases in order to decide which of the various translation strategies available is most adequate in his/her particular case.

Figure 1 .
Figure 1.Representation of examples in the PEnPAL database.

Figure 2 .
Figure 2. Representation of a translation problem type.

Figure
Figure 4. PEnPAL's scheme for a typology of translation problems.