Written by Letizia Merello

Post-editing: going beyond the cliché

Post-editing of automatic translation, often snubbed because deemed a low-quality solution, instead in many cases is a chance to cut costs and time while still gaining outstanding results.

This approach isn’t suitable to all types of texts: generally speaking, you should go for linear texts that aren’t particularly conversational, with numerous repetitions. How about a practical example? Automatic translation works better on a technical text rather than a marketing text, since the latter requires revision of the original text in the target language, which would make the automatic translation practically useless. Therefore it is recommended you perform an analysis before starting, or pre-editing, to see if automatic translation could actually be useful, and above all, to detect and predict potential problems, so the translator knows what aspects to concentrate his or her attention on during stage of review.
If we’re talking about post-editing, the most common error is thinking that all the translation engines are generic and often imprecise like Google Translate. In reality, many companies have been taking advantage of personalized translation engine creation for some time now, since they have a large quantity of data available to train them with, and even diversify them based on topics or type of text.

Based on the text and the quality of the translation engine, the post-editing job can be minimal, limited just to the essential corrections to get a formally correct text, or be more thorough, complete with changes to the style for example. To this regard, some of the other variables in play are the target audience and naturally, the specifications provided by the client. It is also good to keep in mind that the automatic translation doesn’t guarantee satisfying results in all language combinations. Generally speaking, the quality is good for Latin and Nordic languages, while the Asian languages and those that are read from right to left are harder to treat.
Intelligent use of automatic translation, together with targeted post-editing jobs can therefore give surprisingly good results in terms of productivity, without having to trade off the indispensable ingredient for an impeccable final text: a human touch.

Written by ASTW

Automatic translation works!

Are translators afraid of automatic translation? Oddly, you have two completely opposite viewpoints: one that holds that automatic translation is useless because “it’s faster for me to translate from the start than to do post-editing”; or that sooner or later automatic translation will jeopardize translators’ jobs, because “it will reach a level of quality that will make our profession almost superfluous”. In my opinion, these two points of view aren’t just arguable, but are entirely unfounded.

It’s true, machine translation (MT) has major limits in texts for which creativity and, taking it a step farther, transcreation, are essential. Nevertheless, it shows major potential for the technical texts, and in general, those with scientific content. But in the status quo, the technology and knowledge we have yield results far from representing a real threat to the profession of the translator.

Ride the tiger!

It’s undeniable: some translators are guilty of Luddism. First they feared the CAT tools and now machine translation. New technology is a challenge, not a threat that we must defend ourselves from. Here you can see our experience in using machine translation in patent translations, especially availing of Kantan MT technology.

Written by ASTW

Is scientific misconduct inevitable?

«Modesty befits the scientist, but not the ideas that inhabit him and which he is under the obligation of upholding.» These are the words of Jacque Monod, French molecular biologist and Nobel Prize winner in the preface to his essay “Chance and Necessity”. But what is “modesty” when we’re talking about scientific research? And what did Monod mean by this term? Perhaps that as the researcher works he or she must zero in on the subject he is studying without harboring any manipulative, devouring, narcissistic intent from the start? Rather, that his outlook should be that of one who is aware, first and foremost, that his observation may be fallacious—ontologically speaking, that is, deeply rooted in man’s intelligence and his instrumental investigations?

If this were so, the researcher would be someone that studies the particular sphere of investigation while keeping an essential distance, almost using a sort of discretion when later reporting what information that sphere provided him. This person should also feel “sacred horror” in transgressing from the scientific method; a super ego that he introjected during his schooling— the result of an ethical education rather than an academic one— should keep him from behaving in unethical ways.

He should report about that sphere, be an intermediary and a sounding board, nothing more. This “echo” should not contain any reverberations from his ego, but just that little information necessary to divulging, defending and making the research data credible to the public and the scientific community.

In other words, the researcher should not want to exploit or twist the results of his work just to attain some kind of satisfaction (narcissistic, financial, professional, or to get a rush, etc.). Otherwise, his scientific mission— the search for the (scientific) truth—might be questioned. It turns out that this itch for attention or craving for success is harmful to science, because it creates fertile ground for crimes and unacceptable behavior to sprout up, the most blatant example of which is scientific misconduct.

Here I would like to delve deeper into the individual psychological aspects, social pressures and any epistemological particulars of biomedical research which, together or individually, could fuel the development of behaviors such as scientific misconduct that may be seen as real, genuine antisocial behaviors, firstly in the scope of the scientific society and secondly in that of all of society in general.


Definition issues


A preliminary definition of scientific misconduct could be: “the surreptitious and fraudulent addition of false notions into the body of scientific knowledge”. However, scientific misconduct is a transgression from scientific methods and not from scientific knowledge in and of itself. We could therefore consider scientific misconduct—when it is seen as the deliberate production of data—to be at the far end of a spectrum of more or less illegal or ethically reprehensible types of conduct.

These include plagiarism of your own data or that of others, copying publications, citing authors carelessly and improper behavior in reviewing the data or research of others. For example, while the production of data that doesn’t exist is certainly the most serious case, nevertheless vaguer forms of unacceptable scientific conduct exist, which I would link to the adjustment of data to fit one’s own expectations. This means that the results of the research are shrewdly selected (including some, leaving out others) to achieve the most satisfying, flat and pleasing narrative version of the scientific problem in question. Based on the above, science therefore is put on the same level—at least in part—as a form of narrative, as a literary genre just like the rest. Boldly and deliberately picking certain data rather than others (for example, to get a better statistical significance), publishing many more studies with a positive outcome than those with a negative outcome and the fact that new trials are not repeating the experiments of previous clinical trials (it is estimated that only 44% of the most cited clinical trials between 1990 and 2003 were done over yielding similar results), constitute a serious bias of scientific research.


Why people cheat: “publish or perish”, but that’s not all…


Numerous factors may set up, or are at the base of, the emergence of these scientific crimes. Researchers are certainly subjected to a great deal of financial pressure to obtain funding when resources are scarce and in a system where the funding agencies choose who to grant funds to according to the famous peer-reviewing mechanisms.


However, although this method is based on a valid principle, clearly distortion is going on when the referees have conflicts of interest in granting approval to colleagues who are competitors. Furthermore, researchers rarely are free to dedicate themselves to their own topic of study that they get to choose. Last but not least, important subjective factors may contribute to the emergence of these behaviors. The “risk factors”, psychological and personality-related in nature, may be: a narcissistic personality, a distorted perception of reality, an irrational belief that one knows the answer to a question in advance based with the research consequently following suit, an inveterate behavior of self-absolution of one’s “crimes” advancing absolutely captious justifications, as well as delusions of omnipotence. Either alone or in combination, all of these could fuel, or create fertile ground, for the development of scientific misconduct.

Another important consideration has to do with the nature of biomedical research itself, since we see that the large majority of this improper conduct occurs mainly in medical research and in closely linked fields and much less in psychological and social research for example.

The epistemological statutes of biomedical research could be prone to the emergence of these behaviors; living systems are so variable in terms of results and outcomes that copying published data is already hard as it is. This has brought some people to believe that no one will ever be able to discover the misconduct, because it’s not often that researchers check others’ results. In fact, scientific misconduct is often discovered due to major errors in the writing of the text. Figures and images copied from other already published works, calculation errors, inconsistent data, records of participants in clinical trials that are completely made up or false, are some of the most common examples.


In vino veritas (but not always)


Resveratrol is a non-flavonoid phenol in the skin of red grape berries that is said to have an antitumor and anti-inflammatory action. Some research holds that it may also have a beneficial effect on cardiovascular diseases.

Recently one of the biggest world experts on this substance and its potential role in diseases of the cardio circulatory system, Dr. Dipak Das, was removed from office at the University of Connecticut for having made up data on resveratrol in dozens of original articles that he authored. The researcher also had to return $890,000 he had received to the federal government. In these scientific works, resveratrol seemed to have a positive effect on cardiovascular health. Although there was a large amount of data (4,000 articles already published) according to which resveratrol is a molecule with potentially positive effects in a wide range of therapeutic applications, Dr. Dipak Das’s misconduct shook the scientific community. What emerged is that the person responsible had no conflict of interest (he had no apparent ties to the company that provided the substance to be tested) and what’s more, Das immediately denied all accusations against him, going on with a counterattack as rash as it was instrumental saying that the academic establishment was soiling itself with racism against Indian researchers. Despite all this, there is so much data on resveratrol that—although Das’s research was rejected—this molecule still has potential therapeutic applications, including for treating and preventing cardiovascular diseases.

One gathers from this example how easy it is to steal or invent data and how whatever the control bodies did—including the referees for scientific magazines—didn’t manage to stop dozens and dozens of false or entirely made up scientific works from being published. This case is not the only one of its kind. Just think of the recent scientific misconduct by Dr. Diederik Stapel discovered in Holland at the University of Tilburg, where he made up the data in dozens and dozens of research papers published in highly prestigious magazines. All this reveals that whatever measures are supposed to deter such dishonest conduct are entirely ineffective and scientific integrity cannot be forced by third parties. Rather, it is a quality that has to do with the subjectivity of each individual researcher.


Scientists or storytellers?


Louis-Ferdinand Céline (real name Louis-Ferdinand Auguste Destouches), a French doctor and one of the most important novel writers of the last century, dedicated his college thesis to Ignác Fülöp Semmelweis, a forerunner of antisepsis long before microbiology and Pasteur. Céline wrote, «The experimental method is nothing more than a technique, infinitely valuable, but depressing. It requires of the researcher an increase in fervor if he is not to be discouraged before attaining the goal he has set for himself, along that naked path that he must follow in its company. Man is a creature of feeling,» adds Céline. Being able to handle the frustrations is an absolutely essential condition and—as Céline said—requires an increase in fervor which must not give way to a search for fraudulent shortcuts.

So that science does not become a literary genre.

Written by Letizia Merello

Translation of patents: between functionality, rigor and creativity

Patents are a unique type of text, with specific characteristics, not just linked to bureaucratic and legal language and technical industry-specific language (which varies based on the topic in question), but above all inherent to the central purpose of the patent itself: that of illustrating aninvention, that is, an absolutely innovative solution for a certain technical problem, whether it be a product or a procedure.

The text of a patent is therefore a sort of compromise between different needs:

  • functionality: the text of patent has to be functional to the objective to legally protect an invention and guarantee that it be used exclusively by the owners;
  • rigor: it is necessary to follow the structure of the patent’s text, set in a very precise way and comprised of a title, a detailed description (in some cases accompanied by a summary), and a series of claims and some illustrative drawings;
  • creativity: the innovative nature of the patent requires you to create neologisms suitable to explain the innovative concepts that it is based on.

In light of these aspects, translators of patent texts must stay within many limits, inherent not just to the technical and legal value of the text, but also to its practical function. The first rule that they have to follow is matching the original text entirely, so as to avoid any misunderstanding of the legal implications of the patent granted. Secondly, regarding the stylistic elegance of the translation which in general entails a broad use of synonyms to improve the variety of the text and make it more readable and flow better, a dry style centered around frequent repetition of terms and expressions prevails in order to express the concept of the invention unequivocably and deliberately redundantly. Despite this, translating patents doesn’t mean you don’t need skills like the ability to adapt the text and be creative. Since one must deal precisely with innovative concepts, it isn’t uncommon for the translator (just like the author of the original text) to have to develop new terms ad hoc and coin novel expressions. Simultaneously, the focus must be kept on the functional aspects, yet still achieving a certain level of formal elegance.

This is no small challenge for the translator, who in the majority of cases has an education in the humanities, is specialized in a limited number of technical areas and therefore needs reference materials and terminological resources in order to make up for inevitable gaps in knowledge. From this standpoint, integrating assisted translation software programs or CAT tools and creating translation memories can be useful—useful to maintaining terminological consistency in the texts that present numerous repetitions, but also availing of automatic translation tools which provide a translation from a natural language to another. However, considering the numerous problems of automatic translation—especially if performed with generic tools—it is essential to have bilingual text corpora in the source and target languages with which to develop a translation engine that can guarantee greater terminological accuracy, which can be enriched and updated over time as you complete other projects.

Essential to the research and development progress and protecting intellectual property, patent translation is a delicate and complex sector, but at the same time—or perhaps precisely for this reason— is extremely fascinating.

On our blog we offer ideas, suggestions and food for thought on various related topics, ranging from the more technical aspects to the bread-and-butter of a translator’s profession. Follow us!