Machine Translation: An Overview

Machine Translation relies on the use of algorithms and linguistic models to automatically translate text from one language to another. In recent years, machine translation technology has made significant progress thanks to the use of neural networks and deep learning, which have enabled translations that are more fluid and natural compared to previous methods based on rules and statistical translation. However, to ensure high-quality and relevant translation, MT needs to be trained and refined using specific data.

^{_{Related: Machine Translation, Hybrid, or Human Translation}}

Artificial Intelligence: A Broader Field

Artificial Intelligence, on the other hand, is a much broader field that includes the simulation of human intelligence through machine learning, deep learning, natural language processing (NLP), and more. AI can perform a wide range of tasks, from understanding natural language to generating text, from image recognition to aiding in medical diagnoses. While AI can be integrated with machine translation to create advanced translation systems, it is important to note that not all AI applications are specifically designed for translation.

The Benefits of a Trained Machine Translation System

When discussing machine translation tools and trained translation memories, the quality and consistency of the text can reach very high levels, often surpassing that of “generic” machine translation. Here are some key advantages of trained MT:

1. Personalization for Sector and Client: Translation memories specific to a sector and client allow for training the machine with a dataset that reflects the technical language and stylistic preferences of a given field. For instance, translating legal documents requires precise and formal language. A translation for the medical sector needs to be especially accurate and compliant with specialist terminology standards. A trained machine translation can ensure that these requirements are met, reducing the need for manual intervention and improving the final quality of the text.
2. Constant Updates: Specific translation memories can be regularly updated with new translations, feedback, and corrections, continuously improving the quality and adaptability of the system. This is a key advantage over generic translation models, which may not always be updated with relevant data for a particular client, sector, or geographic area. Ongoing updates ensure that translations keep pace with changes in terminology and style, maintaining a high level of consistency and precision.
3. Terminological Consistency: One of the main benefits of MT is its ability to maintain terminological consistency across large projects or repeated translations. Translation memories store previous translations and automatically reuse them when encountering similar terms or phrases, ensuring uniform terminology. This is particularly important for projects that require a high level of accuracy, such as technical manuals or patents.
4. Improved Quality: With training based on specific data and feedback, machines can learn from past errors and correct them in the future. This means that the quality of the translation not only remains high but also improves over time, adapting to the continuously evolving needs.

Conclusion

Machine translation trained with specific translation memories is one of the most advanced and high-performing options for achieving high-quality translations. Compared to artificial intelligence, machine translation offers a unique specialization and focus on linguistic translation. The ability to customize, update, and continuously improve translations with the use of specific memories makes machine translation the best choice for those seeking consistency, quality, and adaptability in content translation.

^{_{Image by cottonbro studio}}

L'articolo Machine Translation and Artificial Intelligence: A Comparison proviene da ASTW.

Machine translation is certainly not a novelty within the domain of quality language services. For years now, the combination of human translators and automatic translation systems has established itself in the market as the best reference model.

Related:

• Translation and Technical Writing: One Provider, Multiple Services
• Technical Translation and The Use of CAT Tools

Not only for the benefits provided to the professional, freed by these tools from the most repetitive and somewhat “routine” tasks of the translation process. But also thanks to the improved outcomes achieved, both in terms of quality and management.

A sector that benefits more than others from this technological progress is undoubtedly the one related to the translation of manuals and technical materials.

Linguistic Technologies and Technical Translations

Adaptive machine translation leverages advanced algorithms to continuously enhance the quality of translations based on linguists’ feedback, while simultaneously adapting outputs to the specific context of the text. Across different sectors, this approach enables faster and more accurate translation of various technical documents, thereby reducing the delivery times of the translated document.

If we add to this the use of the best computer-assisted translation tools (CAT Tools), then the work of human translators can rely on truly functional and comprehensive support.

These tools indeed offer highly useful functionalities to the work of professionals, allowing them to manage the translation process more efficiently and accurately. Among these features, we find the storage and possibility of reusing previous translations through terminology databases and translation memories, ensuring consistency and cohesion in translated manuals and technical materials. At ASTW, we take a further step forward by utilizing translation memories entirely dedicated to individual clients, thus further enhancing the effectiveness of these tools.

Last but not least, CAT Tools integrate quality control tools capable of automatically detecting common errors, such as lack of terminological consistency or linguistic inconsistencies, allowing translators to correct them quickly and further reducing the time needed for reviewing and quality assurance work.

In summary, better results and reduced timelines at a competitive cost.

For more information, click here!

^{_{Image by Tima Miroshnichenko from Pexels}}

L'articolo Manuals and Technical Materials: Translation and CAT Tools proviene da ASTW.

Several Instagram users have indeed noticed that if the English word “Palestinian” were included in a sentence containing the Palestinian flag emoji and the Arabic word “alhamdulillah” (translatable as “Praise be to God” or “thank God”), the algorithm would mistakenly translate the content by inserting the concept of “terrorist.”

A serious, extremely grave error. Capable of resulting in serious consequences for individual users and the movement they are affiliated with. It has the potential to offend an entire nation and its population.

Related:

• Machine, hybrid, or human translation
• The Translator of the Future According to The Economist
• The Fields Where the Human Translator is Still Essential

The news, initially reported by the technology magazine 404 Media, could only further inflame tensions, especially in this tragic period of history.

Meta’s response, the parent company of Instagram, was not long in coming. Through a spokesperson, it announced to The Guardian Australia the correct resolution of the issue, without, however, providing further details regarding its possible causes.

Meta and Machine Translation: Why the Term “Terrorists”?

The void created by the absence of explanations regarding the causes of the problem inevitably left ample room for personal reflections. Among these, the hypothesis of guilt towards the big tech Meta certainly could not be overlooked.

Certainly, it is not up to us to pass judgment. However, what we can hypothesize is the presence of biases within the linguistic model used to train the automatic translation system.

As reported by 404 Media, researchers Gabriel Nicholas and Aliya Bhatia from the Center for Democracy and Technology also believe that the bug is due to the training data of the model.

In summary, if the system is trained on data, often sourced online, containing biases, the model will reproduce these biases with each new occurrence unless direct intervention is made to eliminate them.

“Meta faces a dearth of available training data in languages other than English and specifically in Arabic dialects that might not be widely spoken or geopolitically strong” Batia stated. “The language model is making the connection [based on whatever is] in the available examples of either Arabic language speech or speech related to Palestinian people, and that means the output is reflective of the perspective that this text has.”

And it is precisely for this reason that, to date, the gold standard for translation services requires the indispensable presence of a human translator.

Sources: Multilingual – 404 Media – The Guardian

L'articolo The Risks of Machine Translation: Meta and the “Terrorists” proviene da ASTW.

These gender biases occur when automatic translation or writing systems embed gender-related prejudices within their textual outputs, reflecting biases typical of certain social contexts.

Namely, when, for instance, the feminine gender is associated with terms related to emotions, fragility, family, and beauty. Simultaneously, these biases occur when the masculine gender is attributed characteristics associated with professionalism, leadership, strength, and rationality.

Related: Schwa (Ə) and Inclusive Language

As reported by Slator, this new study conducted by researchers from the Kempelen Institute of Intelligent Technologies (accessible through the link provided in the conclusion) aims to identify fundamental patterns leading to the propagation of these biases, attributing over 3,500 examples to 16 core prejudice models.

How Gender Biases Describe Women and Men

To conduct the analysis in question, researchers collaborated with a team of gender bias experts, embedding instances of these biases in short sentences closely resembling those that might arise in everyday language.

The list, comprising a total of 3,565 examples, was then condensed into a few universal macro-stereotypes, considered as the foundational basis for prejudiced reasoning.

In detail, these gender biases portray women as ’emotional and irrational’ in 254 examples, ‘gentle, kind and submissive’ in 215, ’empathetic and caring’ in 256, ‘neat and diligent’ in 207, ‘social’ in 200, ‘weak’ in 197, and ‘beautiful’ in 243.

Conversely, men are depicted as ‘tough and rough’ in 251 examples, ‘self-confident’ in 229, ‘professional’ in 215, ‘rational’ in 231, ‘leaders’ in 222, ‘childish’ in 194, and ‘strong’ in 221.

These unacceptable stereotypes are capable of portraying a distorted and clearly misleading reality. In an attempt to definitively address this issue, researchers have introduced the GEST dataset, a tool for measuring gender bias within automatic translation systems.

In conclusion, to quote the researchers: “If we are to understand the stereotypes in these models, we need to have them properly defined. […] Our results show a pretty bleak picture of the state of the field today. Different types of models have seemingly very similar patterns of behavior, indicating that they all might have learned from very similar poisoned sources. At the same time, as we now have a more fine-grained view of their behavior, we can try and focus on specific issues, e.g, how to stop models from sexualizing women. This might be more manageable compared to when gender bias is taken as one vast and nebulous problem.”

For those interested in delving deeper, we recommend reading the entire paper:

Women Are Beautiful, Men Are Leaders: Gender Stereotypes in Machine Translation and Language Modeling

L'articolo Gender Biases: ‘Women are beautiful, men are leaders’ proviene da ASTW.

In these texts, we have dealt with the issue of machine translation from time to time, spreading our vision of language services and our daily work.

The ASTW workflow has been based for years on human-machine interaction, or rather, on the linguistic-mt binomial.

All those who have experience in our sector are well aware of the limits that still “afflict” machine translation engines. However, the benefits that these technologies can offer are also well known.

The task of machine translation is not to replace the linguist, but to help him in his profession. And it is precisely within this perspective that all the quality control systems present in the main translation tools come into play.

Terminological consistency, correct use of sectoral language and, why not, grammatical and syntactical accuracy. These are just some of the main advantages provided to the linguist in terms of textual quality.

Moreover, the mere presence of these automated processes will allow the human professional to focus almost exclusively on the most complex translation tasks.

But let’s get back to our topic.

The translator of the future

As already mentioned, within the article the magnifying glass is placed on our work, describing what will be the translator of the future, or rather, the translator “from now on“.

The author cites as an example the opinion of a legal translator from Madrid. In one of the last projects addressed by the linguist, machine translation had converted the term “negligent” with “malice”. This could represent a serious problem within the courtroom, since “the law in question considerably increases penalties when misconduct is intentional”.

The text continues by exposing what was foreshadowed at the beginning of our article, that is, talking about the pros and cons and the different positions of linguists towards machine translation.

Johnson concludes his contribution with an idea that I hope, ironically, will enter soon into our sectoral language. In essence, if the ideal translator will be a perfect combination of human skills and technological components, they might as well be called CYBORGS.

To learn more about our specialised translation services, click here.

^{_{Image by fauxel from Pexels}}

L'articolo The Translator of the Future According to The Economist proviene da ASTW.

eTranslation is a neural machine translation tool designed to make it easier for citizens to access information about the European Union in their preferred language. […] It allows users to automatically translate both texts or parts of texts and entire documents, in all 24 official languages of the EU.

The decision follows the current evolution of the information services global landscape, profoundly revolutionised by the internet and the use of social networks. In an increasingly connected and digitalised world, news spreads much faster than in the past, allowing continuous “minute-by-minute” updating.

This increasing speed meant, for example, that newspapers around the world opened their digital portals and social channels to reach users at every news story.

Immediacy in communication has become of primary importance, and it is precisely from this need that the EU program for an instant automatic translation of all the news broadcast by the Commission is born.

Related:

• Artificial Intelligence and Perfect Translation

 EU and machine translation

The use of eTranslation will therefore be intended for the immediate translation of the press review into languages for which the work of a professional translator is not yet available.

As Mamer himself stated, “We will continue to provide professional human translation of our print material. Which will replace machine translations as soon as it becomes available.”

In fact, eTranslation will fill the time gap that separates the output of the news from its translation by a qualified linguist.

Generally, these contents are available on the European Commission portal – in the Press corner section of the official EC website – in four European languages.

Namely English, German, French and Spanish.

These machine translations will not only cover the gap created by the physiological timing of the professional’s work, but will also expand the language coverage of the EC.

Finally, given the unreliability of unsupervised translation produced by MT systems, the pages translated by eTranslation will be marked with a disclaimer aimed at warning users of the possible inaccuracy of what they have just read.

Fortunately, and once again, the authorities do not question the role of the human translator. Professionals will have the final say, to guarantee the highest quality and correctness of the language conversion.

^{_{Image by Angel Bena from Pexels}}

L'articolo EU, machine translation and press review proviene da ASTW.

Pareschi is one of the Italian literary translators most appreciated by the public and publishing houses. She has translated works such as Jonathan Franzen’s Le correzioni, Shirley Jackson’s Paranoia, and Ernest Hemingway’s Il vecchio e il mare – New Edition. She is also the writer of I jeans di Bruce Springsteen e altri sogni americani.

As reported by several news agencies and by Pareschi herself through her Twitter profile, the input came during a conference held in the Italian Senate entitled “Artificial intelligence and the future of cultural diversity”.

“I have five years left, I’m told. Five years to continue doing my job, literary translation, before being replaced by a machine that will be able to translate exactly like me”.

These are the words reported by various information channels, capable of stimulating the fears of translators who have seen the concept of substitution as the greatest danger.

In general, when we talk about machine translation, we refer to the fields of translation where human creativity plays a marginal role. We are talking about technical and sectoral translations, where the focus lies on the precision of the text, in compliance with industry requirements, terminological accuracy and, of itself, semantic accuracy.

But what are the dangers in literary translation?

Will good enough quality be the future?

Again, reporting what Pareschi said: “The most immediate danger, however, is that the principle of good enough quality passes. Translating a book with AI will be faster and cost less, and if the quality suffers, it doesn’t matter; the frog quickly gets used to boiling water”.

Technological advances, even in the linguistic field, are often aimed at savings.

The implementation of machine translation guarantees a reduction in costs and time related to the translation process. A reduction that, in the aforementioned areas and with the presence of a human linguist, does not negatively affect the quality of the text.

However, the same cannot be said for literary translation.

When translating works of this type, it is not only the textual aspect that captures the concentration of the linguist. It is necessary to translate the emotions, the rhythm, the pauses, the stylistic choices of the author, and any other nuance.

A task that requires completely human cognitive effort and sensitivity, which makes the presence of the translator an essential condition. It is therefore surprising that an artificial intelligence can replace its talents. Although, considering the sudden developments we have witnessed in recent times, a time span of 5 years could lead AI to achieve goals that are unthinkable today.

In any case, according to Pareschi, quality will be affected, assuming a future scenario in which the market will be divided into two bands: “a mass one in which translations will be generated by machines [with a quality considered acceptable] and an elite one for those who want translations made by humans”.

What are your thoughts? Share your opinion in the comments section of our social channels!

^{_{Image by Markus Spiske from Pexels}}

L'articolo The danger of good enough quality proviene da ASTW.

The study aims to create a system for automatic detection of the translation mode of a given text. The classifier, trained on multilingual linguistic models and on textual inputs including the source text accompanied by its translation, showed very promising results.

For more information: Automatic Discrimination of Human and Neural Machine Translation in Multilingual Scenarios

But how is it possible to distinguish the translation process behind a text? And most importantly, how are they different from each other?

Machine, hybrid, or human translation: the differences

Many industry experts are aware of the information in the following paragraphs. Most professional translators, for example, will be able to detect the process behind a translation simply by reading the first few lines of the text.

Nevertheless, it can be an arduous task for non-experts, especially considering the achievements of new technologies.

Machine Translation

This expression indicates a type of linguistic conversion that is the exclusive prerogative of machine translation systems. The source text is entered into the machine translation tool, which returns a text in the language of interest.

Generally, this type of translation returns an output that is semantically consistent with the source text, often at the expense of the textual style, terminological consistency, and respect for the fields to which it belongs.

Often, but not always. There are, in fact, areas and textual typologies that lend themselves more to this process, although experts advise using it solely for comprehending the message and for private translation use by individuals with some, albeit limited, knowledge of the foreign language and/or context.

Related: The Singularity of Machine Translation

Hybrid translation

Otherwise called post-editing, it defines the synergy between human beings and machine translation systems.

Today, it sets the gold standard for translation services, capable of blending the superior quality ensured by human linguists with the control of timing and service costs.

In this working mode, the text is first processed by a machine translation system that has been carefully trained and instructed by the linguist. The output text is then checked and edited by the human translator, who can identify and correct any incorrect or inaccurate textual aspects.

Related: Light and Full Post-Editing: Let’s Clarify

Human translation

As is easily understood, this modality provides for the almost exclusive presence of a human linguist within the process.

It goes particularly well with creative texts, in which the use of machine translation could be disadvantageous, weighing down the work of the professional.

To date, the market tends to penalise this type of language conversion, as competition in terms of costs and timing means that adaptation to new workflows becomes almost essential.

Don’t know which type is best suited to your needs? Do not hesitate to contact us, our experts will be happy to answer any questions you may have.

^{_{Image by Deva Darshan from Pexels}}

L'articolo Machine, hybrid, or human translation proviene da ASTW.

Post-editing

Let’s start from the beginning. The first concept on which it is necessary to dwell for an in-depth analysis is precisely post-editing.

This technique has increasingly established itself as a modus operandi within our industry, becoming an essential service within the offer of language service providers all around the world.

Wishing to give a definition, post-editing is the process of elaboration carried out by the linguist (post-editor) on a text previously translated by machine translation tools.

A kind of textual revision similar to that carried out by the reviewer on a text translated by the human translator.

Moreover, the similarities are not limited to the working method adopted by the linguist. Quoting a European Commission document: “the difference between the two groups [translation with or without the aid of machine translation] is not statistically significant. Quality is comparable if we evaluate quality only in terms of translation errors. Post-editing is a bit faster than human translation, with productivity increasing by 6-15%. ”

However, this translation technique is more suitable for the most technical and sectoral types of text. We are talking about patent, medical, technical, and legal translations.

Instead, the experts tend to discourage the use of post-editing for more creative texts, such as literary and advertising ones.

Light and Full Post-editing

Let’s get to the heart of our topic: the differences between light and full post-editing.

The main distinction between the two modes of PE lies in the commitment required from the post-editor. A slight elaboration, characteristic of light post-editing, will engage the linguist for a shorter time, considering that he will focus exclusively on the detection of significant errors.

Specifically, the expert must:

• Correct mistakes that negatively affect the comprehension of the text;
• Rewrite phrases resulting without meaning;
• Delete any extra text;
• Ensure that the document is translated in full.

Light post-editing is particularly suitable for texts used for internal corporate communication, as well as for all texts whose purpose is purely informative.

Full post-editing, on the other hand, will engage the professional more. In this case, the post editor will have to evaluate all the textual aspects, including the specific terminology, syntax, and style.

The goal of complete post-editing is to achieve an ideal translation, within which the intervention of machine translation is not recognizable in any way.

The quality of the document must therefore be maximum.

In this case, the post-editor must:

• Correct all errors in the text;
• Verify the correct use of the specific terminology;
• Check the intra- and extra-textual consistency;
• Keep the textual characteristics mentioned above unchanged;
• Check the formatting of the document.

In light of what has been said so far, it is clear how the real needs of the customer determine which post-editing mode is best to use, in order to optimize the process and obtain the desired results.

^{_{Image by cottonbro studio from Pexels}}

L'articolo LIGHT AND FULL POST-EDITING: LET’S CLARIFY proviene da ASTW.