2.1 Use of MT as a bad model

Ball (Reference Ball1989) was one of the first authors to question the role of MT in the language class. To support this use Ball recommended the correction of errors appearing in computer-produced translations as the most immediately obvious language teaching application of MT for language students. In this sense the MT output is seen as a source of errors to be corrected by language students. More recently Somers (Reference Somers2003: 327) pointed out that this use can actually “bring out subtle aspects of language differences” and “reinforce learners’ appreciation of both L1 and L2 grammar and style”.

Anderson (Reference Anderson1995) and Richmond (Reference Richmond1994) made wide use of MT as a bad model. Anderson used a bidirectional English-Hebrew MT system to encourage Hebrew students of English to identify and correct the MT errors using native speakers’ intuition and/or L2 reference works depending on the translation direction. Into the L1 this activity can reinforce students’ awareness of differences between the two languages by showing them translation traps. However, its use into the L2 is controversial because it exposed the students to erroneous L2 text. Richmond opted for showing the students a model translation into the L2 and asking them to modify the L1 source text until appropriate target language texts were to be obtained.

2.2 Use of MT as a good model

This involves the use of translation memories, i.e. databases of aligned or semi-aligned text segments in multiple language pairs. These tools in combination with the use of online dictionaries and thesauri, parallel texts, concordancers and other corpus-based tools can help students sharpen their writing and translation skills into the L2, and its use is especially recommended with specialised languages where checking for appropriate terminology/phraseology is paramount. Authors such as DeCesaris (Reference DeCesaris1995) and Shei (Reference Shei2002a) recommend the use of translation memories in the language class as an autonomous resource to check for phraseological and grammatical correctness into the target language.

2.3 Vocational use: Translation quality assessment, pre-editing and post-editing

The use of translation memories and bi-texts as good models also constitutes one of the main uses of MT for language teaching and learning purposes. Accordingly, one of the main applications of the teaching of MT in the class is its use by professional translators who, apart from being proficient in two or more languages, need to know the intricacies of the translation art and be updated on the use of CAT (Computer-Aided Translation) tools such as translation memories or MT systems. Other common practice in this profession is translation quality assessment of human or MT output (MT evaluation).

Belam (Reference Belam2002) has investigated the use of MT evaluation by a group of final year undergraduates in Modern Languages in an introductory course to MT at the University of Exeter. The students were asked to design a project to evaluate the translation quality of raw MT output for a particular purpose, e.g. to evaluate terminology and dictionary tools in two different MT systems, one from Systran and another from Globalink, or to do a comparative evaluation of the translation of different types of text. According to Belam, the students who joined this course benefited from multiple transferable skills such as vocational practice, appreciation of the writing process and the value of imperfect communicationFootnote 1, together with increased language awareness.

Pre-editing involves the modification of the source text until appropriate MT output is obtained. It constitutes a rather artificial form of text manipulation that is more commonly used in the translation industry for disambiguation purposes. In language teaching French (Reference French1991) supported its use with relatively short texts, encouraging students to appreciate what might be potentially ambiguous in the source text in order to spot the type of linguistic difficulty the machine is facing. For this purpose French also suggests the manipulation of the MT system’s dictionary entries by the students.

In the same way Shei (Reference Shei2002b) also encouraged a group of Chinese students of English to write their compositions in English, enter them in an MT system and correct the English source text until the Chinese MT output sounded better to them. Then the students were asked to provide a list of MT flaws they observed. The only concern Shei had about this practice was that, in most cases, the MT output errors were due to insufficiencies in the MT system’s grammar and dictionaries and not to incorrect language input being fed into the machine.

Post-editing involves the correction of raw MT output into an acceptable text for a particular purpose. La Torre (Reference La Torre1999), Niño (Reference Niño2004), Belam (Reference Belam2003) and Kliffer (Reference Kliffer2005) reported on different uses of post-editing in the FL class. La Torre (Reference La Torre1999) offered an innovative web-based resource to upper intermediate and advanced students of Spanish as part of a module in “Introduction to Translation Theory” to familiarise the students with translation tools that are commonly used nowadays as well as to perform basic post-editing of raw MT output. The students were presented with guided tasks, for example in one of them they had to polish the raw output into English of a literary passage in Spanish. Niño (Reference Niño2004) provided a group of English advanced students of Spanish with a general English source text and a raw MT output into Spanish and prompted them to underline and correct the erroneous parts. This activity presupposed and fostered advanced language skills in the two languages and also translation skills. The focus was on the recognition of common language and translation errors and their correction in groups who suggested different alternatives for modification or correction. Belam (Reference Belam2003) argued that post-editing can serve as a complement to language tuition for it makes the students focus on analysing the source text, thus learning new vocabulary, expressions, grammar points and stylistic aspects. Kliffer (Reference Kliffer2005) made use of post-editing into the students’ L1 in an undergraduate class of French to English translation. The students showed a clear improvement of the MT output. Kliffer pointed out that the experience seemed to be more valuable for weaker students for whom post-editing a text seemed to be less stressful than doing the whole translation by themselves.

2.4 MT as a “CALL” tool

As long ago as 1985, Patrick Corness suggested the use of the ALPS Footnote 2 interactive MT system in an advanced German class. He pointed out that one of the advantages of this system was that the user interacted with the machine, choosing the correct translation for an ambiguous sentence into the target language. For Corness this practice was very useful in tackling language difficulties regarding grammar, context, denotation, connotation, register or cultural references. He also mentioned the system Autoterm Footnote 3 that could provide CALL exercises to practise these particular language difficulties. More modern and affordable MT systems such as PROMT,Footnote 4French Assistant Footnote 5 or Power Translator Pro Footnote 6 combine general and specialised dictionaries as lexical aids that can handle word or phrase units, and grammar aids such as verb conjugators that can serve as a constant reference for the students.

All in all, the use of free online MT is controversial because, although the linguistic quality assessment of its many forms of raw output constitutes a form of MT evaluation/vocational practice, its use as a good or bad model cannot be easily differentiated by FL students who often use these sites to find a quick solution to a language difficulty they do not know how to solve on their own. Since this is not commonly acknowledged and its use may not be well known and/or recommended among language tutors, in the next section we will look at some of the implications of the use of MT and free online MT for FL learning.

4.1 Examples of good practice may include the following:

1. 4.1.1 The use of free online MT output into the students’ L1 for summarising purposes at beginners’ level or for reading comprehension purposes in a language they do not understand, in order to get the gist of what the text is about.

2. 4.1.2 MT post-editing. Presenting the students with the source text alongside the MT output and relevant resources (parallel texts, dictionaries, grammars, etc.) for that particular text, and getting them to underline and correct the erroneous MT output for a particular purpose.

3. 4.1.3 Getting the students to enter terms and/or grammatical information into a commercial MT system’s dictionary. Lewis (Reference Lewis1997) reported on the efficiency of this activity in a group of students of German who found the entering of grammatical information into an MT dictionary a useful experience, especially for weak students for whom it provided a stimulus to investigate different areas of basic grammar such as gender, agreement or prepositional verbs.

4. 4.1.4 The use of parallel corpora with or without MT output. The author has used it in advanced specialised language lessons and it has proved to be beneficial for error identification and correction, to teach how to translate common problematic expressions, for style improvement and for phraseology/terminology. This use is particularly suitable for vocational training courses in translation that may well include an introduction to the use of translation memories. DeCesaris (Reference DeCesaris1995) states that these programs can be adapted for use in a classroom setting as a self-learning resource.

5. 4.1.5 MT evaluation. This presupposes that the students have some prior knowledge of MT, advanced language level and translation skills. Students can be asked to evaluate the MT output of an MT system for a particular purpose, to compare the translation quality of different MT systems, or else to compare different text types translated by the same MT system.

6. 4.1.6 A final example of good use of free online MT is with closely related languages. Language learners, translators and linguists interested in learning by comparison closely related languages such as Romance (Spanish/Portuguese or Spanish/Italian) or Slavic (Czech/Slovak, Czech/Polish, or Czech/Slovenian)languages, can learn the similarities, differences and idiosyncrasies between them. As expected, the output of free online MT systems tends to be of better linguistic quality if the source and target languages are closely related as compared to languages that differ substantially in lexical, morphological and syntactic structure. Exploring a multilingual system’s syntactic rules and/or dictionaries can provide us with privileged information regarding the intricacies of the different languages and, by introducing simple “simulated” examples in the L1, we can see how the translation is made and learn from any possible errors.

4.2 As examples of bad practice in using free online MT in the language class we can name the following:

1. 4.2.1 Using a free online MT system for L2 writing or for translating into the L2 without having a good command of the L2 and some previous knowledge about the kind of errors an MT system produces.

2. 4.2.2 Focusing on free online MT output as a product, not taking into consideration that it is not a very elaborate kind of MT system and without understanding the different stages of analysis, transfer/interlingua, generation and post-editing (Hutchins & Somers, 1992).

3. 4.2.3 With languages such as Japanese and Spanish which are not closely related, in MT and free online MT systems it is more likely that the quality of the translation will leave much to be desired due to striking lexico-semantic and structural differences between the two languages.

4. 4.2.4 Presenting low level students free online MT output with a high degree of various kinds of inaccuracies to be corrected, without a clear objective (i.e. minimal post-editing) or an indication of the prospective end user and the main purpose of the text.

5. 4.2.5 Selecting an inaccurate source text or a source text type not very suitable for MT, such as a horoscope full of idiomatic expressions or a complex literary extract.

6. 4.2.6 Not introducing MT properly to students, i.e. presenting a brief history, its main aims, advantages, disadvantages and main differences as compared to free online MT.

Fostering awareness of the potential and limitations of MT is fundamental for both language tutors and students and will hopefully help make a better use of this technology.

5.1 Learners’ perceptions on the use of MT for FL learning

The first survey (see Appendix A) was addressed to a group of sixteen advanced students of Spanish who undertook a ten-week introductory course in MT and MT post-editing. What follows are the responses to six open questions, aimed at evaluating the learners’ attitudes towards using raw MT output as input for FL written production, and intended to reveal some insights into the procedure followed by the students when post-editing into the FL.

To the question Which process have you followed in your post-editing, i.e. have you looked at the MT output before or after doing your post-editing? 87% of the students reported having looked at the MT output before or before and after their post-editing for reference and checking purposes, and the rest did not respond.

To the question Have you used any other language tool for your MT post-editing? If so, which one? 93% of the students reported having consulted other references such as outputs from different online MT systems, online dictionaries and glossaries, search engines, and parallel texts, and the rest did not respond.

To the question Are you going to use MT in the future? 69% of students reported that they would use MT in the future. Among the applications they mentioned were reading comprehension in other languages, as a quick lexical reference, with technical texts, and to obtain a quick draft to build on. 19% of them reported that they would not use MT in the future, the low percentage of correctness being the main reason mentioned, and 12% reported that their use would depend on the language pair and on the text type.

To the question Do you think MT is a useful language tool? 75% of students reported that MT is a helpful language tool. They argued that MT is quick, user-friendly, and it provides a starting draft. They also stated that it can be very useful with vocabulary, especially with terminology. 25% expressed the view that it may be useful as a comprehension tool to read texts in other languages.

To the question Has MT helped you to improve your level of Spanish? 81% of students expressed that MT had contributed to their FL improvement. They explained that polishing the raw MT output had promoted their error detection and correction skills, had made them reflect on their own errors, had fostered their comprehension skills, and it had also made them focus on language use and on accuracy. 19% of the students said that the lessons helped them improve their Spanish, not MT as such.

To the question Has MT given you more confidence in your FL written production? 75% of the students expressed that practising MT post-editing into the FL had boosted their confidence in FL written production. They argued that the experience had provided them with more self-evaluation strategies, with more language practice, with a little more confidence in their Spanish grammar and vocabulary, and with knowledge of the possibilities and limitations of MT. 25% said that it was the language practice with MT and the feedback from the teacher and from their classmates that gave them more confidence in their written production.

Further comments by the students indicated that the experience had provided them with more language awareness, more accuracy, and more fluency in their FL written production. They described this activity as challenging, easy to use, and non-threatening. Many also stated that after the experience they had developed a better attitude towards MT.

The results of this survey are useful because they reflect the views of a group of advanced students of Spanish who have received previous training in translation and who have been introduced to MT and in particular to MT post-editing. On the other hand, a wider spectrum of students ranging from Beginners’ to Advanced level may have provided us with a broader picture of various uses of free online MT depending on the student’s command of the target language.

In any case, we believe that students of all levels would agree that reading comprehension and lexical reference are probably the most popular uses of free online MT among language learners, without mentioning the unauthorised use of free online MT output for composition or translation assignments in the language class. As advanced students those questioned realised the convenience of free online MT (quick, easy to use, non-threatening), but also noticed the drawbacks of this online reference tool (low correctness, dependence on language pair and text type) and have therefore realised that, given their language knowledge and preparation, and having compared the raw MT output with reliable translations, it is probably less time- consuming to try writing a text or translating it into the target language on their own, using online references such as the ones mentioned above to check for accuracy and correctness. A native speaker of the target language will need to be consulted about more complex issues such as adequacy, language connotations and other sociolinguistic and cultural issues.

In sum, judging by the students’ answers, the MT language learning experience was useful for the language practice that it involved, because by detecting and correcting errors in the MT output they put into practice all their previous language, translation and MT knowledge and became aware of the applications and limitations of MT from a vocational point of view.

5.2 Language tutors’ perceptions on the use of MT for FL teaching

In order to learn the opinions of language tutors on the use of MT for FL teaching and learning we carried out a survey (see Appendix B) to find out about their attitudes and views towards translation and MT as pedagogical tools in the language class.

Thirty language tutors of foreign languages at university level were recruited to participate in the survey by means of the EUROCALL discussion list and by sending emails to the heads of several modern languages departments and language centres around the world. 53% of the participants were between 30 and 40 years old, 23% were between 50 and 60 years old, 17% were between 40 and 50 years old, and 7% were less than 30 years old. A majority of them (60%) were female. They were native speakers of different languages: Spanish, English, French, Italian, German, Polish, Dutch, Japanese, and Finnish. Most of them had taught their L1 as a FL at all levels (beginners, intermediate, advanced, higher, and improvers).

MT was familiar to 93% of the participants. Most of them (60%) had learned about MT on the Internet. 60% of the participants had used online MT systems, 20% of them also had used commercial MT systems.

With regard to frequency of use, 60% of the participants used MT several times a year and 17% had never used it. Concerning their main motivation for using MT, 33% used MT for curiosity, 20% used it as a first draft, and 17% used it for reading texts in other languages. Other uses specified were: to teach students about the linguistic limitations of MT output, to teach language ambiguities in MT, to communicate via chat or email with speakers of other languages, in a computer-assisted English course, and for research.

Regarding the use of translation for pedagogical purposes, 70% of the participants had used translation as a teaching/learning tool, mostly from L1 into L2 (30%), and also in the two language directions (27%), generally with intermediate and advanced or higher levels (43%), and with general texts (50%). The main objectives stated for the use of translation as a pedagogical tool were the following: for reading comprehension purposes, for written production in the L2, for revision purposes, for translation practice, for accuracy training, and for assessment.

Regarding the use of MT in the language class, only 23% of the participants had used MT in their lessons both from L1 into L2, and from L2 into L1, from intermediate level onwards, and with both general and specialised texts. The main objectives stated for the use of MT in the language class were the following: for written production practice in the L2, for revision, for assessment, for reading comprehension in the L2, and for translation theory practice. All the participants who used MT in their lessons think it is easy to use and useful, and most of them would include hands-on practice on MT in a language syllabus.

From the participants who had not used MT in their lessons 30% would use it, 30% would not use it, and 37% did not respond. The participants who would use MT in the language class would use it mostly at advanced levels of the language, both from L1 into L2 and from L2 into L1, mainly with online MT systems, and using general source texts. The main objectives specified for the use of MT as a teaching/learning tool in the language class were the following: for post-editing and pre-editing purposes, as a translation draft, to practise contrastive analysis, for translation practice, to raise consciousness about the complexity of translation, for reading comprehension purposes, for revision, and to make the students aware of the main limitations of MT.

The last part of the language tutors’ survey consisted of three questions of a more open nature. Here we report on the answers collected:

To the question How do you suggest that MT can be used in the language class? Participants suggested the following: to understand the general meaning of a FL text at beginner’s level, to practise post-editing in a specialised language course, to practise error detection and correction, to reflect on the L2 (on how analysis together with evaluation of the MT output can foster the improvement of the quality of the final translation), to produce first drafts until students can learn to translate into the L2, and to compare the MT output with the students’ own output and a sample target text in order to show the potential and limitations of MT.

To the question Which advantages do you think MT has for FL teaching/learning purposes? these were the main responses from the participants: it entails error detection and correction practice; it involves written production practice; it is user friendly, and it is fast. Other advantages expressed by the teachers were: the teacher can control the input text; it generates first drafts to work on; it easily creates ready-made texts for practice; it involves professional training; it fosters contrastive analysis and raises consciousness of the complexity of translation; it explores the role of culture in MT, and it makes students aware of the potential and pitfalls of MT.

Finally, to the question Which limitations do you think MT has for FL teaching/learning purposes? these were the main responses from the participants: low quality; text type constraint (not suitable for creative texts); it requires training; and it may be a bad model (especially with beginners and intermediate level students). Other limitations put forward by the participants were that it is too form-focused, that it may lead to over-correction, that it is unreliable, that it implies a lot of work and does not always provide effective results, and that it does not provide a very good translation as yet.

Judging by the results of the language tutors’ survey we learnt that the majority of them knew about MT and especially about free online MT but only a few of them actually dared to try out the quality of its output in order to check what MT errors look like and up to what point they would be able to distinguish them from students’ output. Their linguistic curiosity, though, did not prevent them from using these systems for reading comprehension purposes and as a first draft in other languages.

The majority of the respondents had used free online MT from L1 into L2, presumably for language testing purposes, to check the accuracy of the MT system. Nevertheless, only a minority had actually used MT in their language classes with both general and specialised texts and from intermediate level onwards. We assume that the language tutors selected source texts according to the students’ interests and language level, and in accordance with the language content seen previously in class, hence they stated revision, written production, translation practice and assessment as main practices in the language class. The tutors, we assume, also selected the MT output according to the students’ normal FL written production, including similar types of mistakes. For this group MT seems to be a reality that is worth investigating. In their answers they mention translation theory practice so there is probably a correlation between language tutors with a translation knowledge/background and sympathy with MT from a linguistic point of view. Thus, the majority of respondents who would be ready to explore MT in the language class and integrate it into the curriculum mention key words and activities directly related to translation as a vocational activity, such as pre-editing, post-editing, contrastive analysis, and learning outcomes such as broadening their translation knowledge into the L2 and exploring the potential and limitations of MT.

Despite agreeing that MT output constitutes a ready-made product and that the free online MT technology is easy to control and to use from the teaching point of view the majority of the respondents showed some concern about the following issues:

1. (a) the general low quality output that may confuse lower level or weak students even more,

2. (b) the text type constraint that may restrict too much the adequacy of source texts and MT outputs to be used in class,

3. (c) the required training, for in order to explore the full potential of MT in language learning it is advisable to have advanced knowledge of the target language, translation skills and MT knowledge, and

4. (d) the overwhelming focus on form as opposed to the communicative approach, still one of the methods most commonly used in the language class.

The majority of them, however, hope that maybe in the future, when the quality of the MT output is better, they can start thinking about incorporating this technology into the language class. These comments seem to reveal a lack of understanding of the usefulness of MT as it stands nowadays. What we ignore is whether the idea of an MT system able to produce near-to-perfect output is what they had in mind, something that would signal the end of traditional language learning and could be the start of an MT CALL era where language and translation skills are learnt in conjunction from scratch with clear vocational purposes and with facilities to expand this knowledge to other related languages.