Machine translation and human translation are both widely used in international business services today – but, which is inherently better or more valuable? In this 3-part series, we’re showing you an inside look at our latest white paper, Man vs. Machine: A Guide to Translation Technology in Modern Business Use Cases.
This downloadable white paper explores the current and future states of the translation industry in terms of services for organizations, and provides guidance on how companies can better understand and navigate the complicated world of language services. Stay tuned on the blog for the next part of the series, and see a copy of the white paper here.
Industry Overview: Where is Translation Today?
An Introduction to Human Translation
Translation is commonly misunderstood as a simple process involving the word-for-word localization of written text from one language into another. as one of the single most important investments to multinational organizations of all sizes, translation is, in fact, a complex process requiring the adaptation and interpretation (vs. the literal conversion) of written text from one language into another.
Human translation is performed by “live experts” that have dedicated education and training in translation from one language to another as well as a unique passion for the intricacies of language, dialect, and culture. those interested in human translation have three main options [1]:
1. Traditional Agencies: Usually comprised of small groups of freelancers, full-time translators or individuals, translation agencies manage individual projects and files for businesses. Boutique agencies, on the other hand, offer this as well, along with more exclusive service and dedicated customer care.
2. Crowd Platforms: Crowd platforms host a mass of translators (sometimes thousands of translators at once) to work simultaneously on a given file or project. Through this option, there is never any one-on-one contact with the translators involved, and inconsistencies due to the stylistic preferences of each translator tend to run rampant.
3. Individual Freelancers: Freelance translators have received their training and certifications, and work exclusively through online forums and directories to connect with businesses looking for work. translators who are operating independently typically offer lower rates, but with added risk involved, as these service providers don’t generally provide quality assurance resources.
Despite innovation in the translation market and the introduction of machine learning and algorithmic powered solutions, human translations remain the most accurate option [5]. providing a higher standard of precision than most translation options [6]; human translation, chiefly human translation by a traditional agency, freelancer or an independent resource, is the most widely-used method to this day [1].
An Introduction to Machine Translation
Since the discovery and implementation of machine translation in 1956 [2], several studies published by a variety of highly-regarded institutions have evaluated the practical applications of machine translation, as well as the actual vs. perceived value that these translations can provide. Machine translation is powered by computational linguistic rules and automated technical processes that can translate source content into its target languages. Machine translation is preferred by some organizations primarily for expediency, as opposed to accuracy. Most machine learning translation solutions are able to provide nearly instantaneous results for the vast majority of translation projects [3].
There are a few main classes of machine translation solutions [4]:
1. Rules-Based Machine Translation: These systems use large collections of rules based on the source language and target language, which is developed by real, human experts over time. Due to the manual element, this type of machine translation can be costly and time-consuming to maintain and improve upon.
2. Statistical Machine Translation: These systems produce translations through statistical models that consist of words and phrases learned automatically from bilingual parallel sentences. through this algorithmic process, the computer creates a “database” of translations, making it the most automated and expensive way to translate text.
3. Hybrid Machine Translation: With a combination of rules-Based and Statistical translation systems, this hybrid system uses the core technology of each to provide higher quality translations overall. However, combining systems can be very costly and infinitely more complex to manage simultaneously.
4. Augmented Neural Machine Translation: The newest form of translation systems have evolved beyond the limitations of Rules-Based and Statistical translation systems. these augmented neural systems introduce sophisticated data pre-processing, along with terminology management solutions, to produce high-quality and multilingual solutions.
Machine-produced translations attract organizations and individuals with projects that have very tight budgets and require super-fast results. However, despite having greatly evolved over the last decade, these systems can produce results of varying degrees of accuracy that can cause costly consequences when improperly used.
Humans vs. Machines: The Current State of the Translation Agency
Due to the vast and changing landscape of the translation industry, the greater reliance on machines to provide translation output is raising questions about the role that people play in the field. Executed in some capacity by both humans and machines, there are two major methodologies to producing translations from one language to another:
1. Rule-Based Translation is a system that relies on inherent knowledge of the grammatical structures of both languages. a bilingual dictionary system allows
the system to translate one word at a time after the words are rearranged into “correct sentences” [5].
2. Experience-Based Translation leverages a database of past materials across previous translation projects. Often referred to as a “Knowledge Base” by human translation solutions, this system operates under the assumption that phrases and sentences produce a general understanding of the meaning, but are only useful to the casual reader [5].
For machine translation, results are vulnerable to errors that a live professional with experience and capabilities of understanding context and tone couldn’t make. When selecting machine translation over human translation, the sentence structure of the source document is often reviewed and re-architected. However, despite taking time-consuming additional steps, the translation output is consistently lower than that which an experienced human translator can produce.
There are a few major machine limitations that lead to this discrepancy in accuracy:
Barriers to Communication
Adapted to communicate across international borders. Supported by Cat (Computer-assisted translation) tools and digital dictionaries, human translators have made long-distance working and communication easier than ever [5].
Post-Editing
Despite boasting advanced capabilities provided by machine translation, businesses often subject language projects to budget cuts. In the commonly misheld view that machine translations provide comparable outputs at a fraction of the price, many businesses overlook the incurred cost of revisions and post-editing activities [5].
The Real World Implications of “Cut-Rate” Translations
The appeal of affordability can often impede on the desire to attain seamless, culturally-accurate translations. Implementing translation outputs with questionable accuracy into an organization’s processes and workflows can have an extremely negative impact on business, including:
- Damaging the brand due to public cultural mishaps that could be construed as inappropriate or insensitive.
- Promoting and using phrasing that contains cultural inaccuracies or failure to understand nuances.
- Receiving embarrassing publicity surrounding the improper use of language and cultural insensitivity, among other errors.
- Increasing the overall risk of customers and/or employees, due to public company mistakes and cultural inconsistencies.
Regarding more high-risk industries, such as legal and medical/pharmaceutical, poorly-translated texts can also directly impact the health and well-being of their customers and patients. It only takes a single translation mishap to severely damage a brand in the public’s eye. Choosing language services providers that are reliable and accurate should always supersede the desire to save money.
Stay tuned for the rest of the series on the blog, and see the full white paper here.
Resources
1. https://gengo.com/human-translation
2. https://nilservices.com/machine-translation-vs-human-translation
3. https://www.gala-global.org/what-machine-translation
4. http://www.machinetranslation.net/quick-guide-to-machine-translation/machine-translation-technologies
5. http://blog.ititranslates.com/is-there-a-difference-between-machine-and-human-translation
6. https://www.theguardian.com/education/2014/sep/19/tech-removing-language-barriers-jobs-lost-translation
7. https://www.gala-global.org/publications/intro-machine-translation-understand-when-use-mt-and-when-avoid-it
8. https://blog.memrise.com/2017/04/19/machine-translation-vs-human-translation
9. https://www.inc.com/geoffrey-james/the-20-worst-brand-translations-of-all-time.html
10. http://translatorthoughts.com/2014/01/5-common-translation-mistakes
11. https://blog.hubspot.com/marketing/global-marketing-and-international-business
12. https://languagedepartment.com/blog/global-events-translation-case-study-rpmc
