Traditional editing versus Machine translation post editing: The good, the bad and the truth

2024-03-22 | Malina Prusinska-Kurpiewska

Traditional editing versus Machine translation post editing: The good, the bad and the truth

In translation and localization services, the process of editing is important. Both human-driven translation editing and Machine Translation Post-Editing (MTPE) play significant roles, each with its own strengths, limits, and applications. The critical factor lies in understanding which approach to use in a given situation. Continue reading to explore this further.

>> Learn more about the true value of Machine Translation Post Editing

Editing of a human-made translation

The whole process of translation and proofreading/editing in traditional services heavily relies on human linguists equipped with the industry-standard computer-assisted translation (CAT) tools to ensure consistency: the translated text should be consistent within itself and aligned with any reference materials or guidelines provided.

When editors review the translated text, they not only correct linguistic errors such as spelling, typos, grammar mistakes, and punctuation issues but also rephrase sentences to enhance the flow, improve the organization of ideas, and ensure that the target text conveys the intended message with the desired tone, aligning with the target audience. 

Traditional translation editing tasks demand the expertise of a proficient and highly skilled linguist who possesses a deep understanding of both the source and target languages. Exceptional writing and communication abilities are essential. Additionally, there is often a preference, if not a requirement, to engage an editor who possesses a certain level of familiarity with the subject matter.

The good 

There are several positive aspects to traditional editing services, including the predictability and controllability of quality, cost, and turnaround time.

  1. The combination of professional human translation and human editing yields the highest quality final product. This ensures that the text is flawless in terms of grammar, punctuation, style, and accuracy, as well as effectively delivs the intended message while being tailored to the target audience’s preferences.

Traditional editing also ensures optimal resource utilization by aligning skills with the task at hand. The active collaboration between translators and editors results in positive outcomes in terms of quality, often extending across multiple languages. This improves on project efficiency and leaves a lasting impact on future releases or additions. It also enhances the team’s language skills across all projects. 

  1. Cost and turnaround time of the service are much more predictable. Editing human translation typically incurs lower costs than MTPE, with editor rates typically ranging from 50% to 60% of the translation rate. Retranslation is seldom necessary, contributing to the overall predictability of editing costs. 

A professional editor is generally expected to handle up to 6,000 words of human-made translation per day, averaging 4,500-5,000 words per day for pre-planned tasks involving uniform text. However, these figures can vary significantly depending on factors such as source and target quality, language, subject matter, text type, and number of files.

Editing human-made translations is generally faster than MTPE, as editors usually deal with only a fraction of possible mistakes. There are certain types of issues that human translators rarely make, so editors can fully concentrate on the accuracy of the translation and its style. Their job is to “polish” the translation rather than “fix” it.

The bad

While traditional editing services have their merits, it’s important to acknowledge their limitations. 

  1. The main limitation of traditional editing services lies in their susceptibility to human error. Despite efforts to reduce them, human errors can still happen, and they frequently arise from contextual misunderstandings or divergent cultural interpretations of the text. Consequently, it emphasizes the critical importance of conducting thorough reviews and careful proofreading to ensure accuracy and consistency.
  2. The overall cost of the localization process may exceed that of Machine Translation followed by Post-Editing (MT+MTPE). This is particularly evident during the quoting stage, where the initial cost of human translation services very often appears higher. 

However, it’s important to recognize that due to the unpredictable nature of the MTPE process, the final cost of the human approach could ultimately be lower. This is attributed to the reduced need for extensive rework and revisions inherent in MTPE.

  1. Furthermore, the localization of content through human translation may necessitate a longer timeframe, especially when compared to the expedited nature of machine translation. However, this trade-off often results in a higher quality end-product, as human translators possess the ability to capture nuances and convey tone more accurately.
  2. It’s important to acknowledge that human translation services are sometimes viewed as “old-school”, particularly amidst the widespread excitement surrounding recent advancements in machine translation technology. In today’s rapidly evolving technological landscape, there is a prevalent desire to embrace and incorporate the latest innovations into workflows, leaving traditional methods seemingly outdated.

Machine translation post editing 

The translation process is fully automated using industry-standard CAT tools and Machine Translation engines with varying levels of professional usability. Translation memories, glossaries, or prompts are employed to produce higher-quality output, but the approach is not standardized and may vary from one context to another, resulting in differing degrees of generated outcomes

The good 

Machine Translation Post Editing can bring enormous benefits when integrated with robust support from Translation Memory, glossaries, and tailored prompts during output preparation. These tools not only contribute to the stability of translation outcomes but also render the MTPE process more predictable and efficient in terms of time and cost.  

  1. The synergy between MTPE and these support mechanisms not only ensures accuracy but also potentially reduces the overall cost of localization services. By leveraging automated processes and refining them with human expertise through both prompting and post-editing, companies can achieve high-quality translations at a lower cost compared to that typically associated with traditional human translation and editing workflows.
  2. Moreover, this optimized approach to localization has the potential to expedite the entire process. With the aid of machine translation enhanced by post-editing and supportive tools, the turnaround time for localization projects can be significantly shortened. This accelerated timeline not only benefits companies seeking to reach global markets quickly but also allows for more agile responses to updates and changes in content. 

The bad 

Due to the diverse practices employed during the output stage, MTPE faces significant limitations. These variations in approach can hinder the effectiveness and consistency of MTPE processes, leading to challenges in achieving desired translation quality and efficiency

  1. The speed of work delivered by MT post-editors heavily depends on the quality of the translation output and is virtually impossible to predict with certainty. 

When dealing with high-quality MT output, the pace of work may resemble that of editing human translation.

Conversely, lower-quality MT output may significantly reduce efficiency, potentially slowing down the process to a level resembling, or even below, that of retranslation (up to 2500 words a day).

  1. Without proper MT preparation, MTPE is typically slower than editing human translation, as editors often contend with a wide array of errors, some of which can render the text illogical or even nonsensical. The primary focus of the post-editor is thus on ‘fixing’ the text and reducing the number of obvious grammatical, accuracy, consistency, or punctuation mistakes, leaving little to no room for improving tone or making necessary adjustments for local audiences. 
  2. It is important to remember that MT output may turn out to be very unstable and unpredictable, as well as suffer from sudden drops in quality after consistent stability. This unpredictability also makes any efficiency assessments particularly challenging and risky. 
  3. Customers commonly presume that MTPE will be significantly cheaper than traditional translation services. However, this assumption is often inaccurate. Drawing from LocalizeDirect’s experience, it’s revealed that utilizing a combined Translation, Editing, and Proofreading (TEP) approach may yield a service cost similar to or only slightly higher than that of MTPE, especially for texts that require creativity, are highly specialized, or involve certain languages.
  4. Since the linguistic constructs delivered by AI can be highly non-standard, MTPE is much more energy-consuming, making the post-editor less efficient compared to editing human translation. MTPE requires the involvement of a professional, highly skilled linguist with additional experience in and willingness to provide MTPE services. This narrows down the pool of available linguists significantly. The limited availability of freelancers offering MTPE makes it challenging to select the best-suited individuals for the job. 
  5. MTPE rates typically range from 70% to even 100% of a translation rate, with some instances where linguists charge even more for MTPE than for actual translation. Due to the unpredictable nature of MT output, editors often find themselves needing to request rate modifications mid-project, especially when the text requires retranslation rather than simple editing, or when post-editing proves to be particularly challenging.

The truth 

When seeking translation services, it’s common for individuals and companies to expect a combination of cost-effectiveness, rapid turnaround times, and impeccable quality, all at once. Is it true that achieving this trifecta is impossible in reality? 

MTPE-is-fast-good-and-cheap

Achieving this ideal trifecta always requires extensive experience in working with both machine translation and human linguistics. With over 10 years of experience in game translation, proofreading, editing, and language quality assurance (LQA) projects, LocalizeDirect happily advises the clients on most optimal prioritization of goals that in turn significantly impacts preferences for services. 

Besides, LocalizeDirect looks for more understanding of the motivation and background of the services choice, for example, if the editing service request involved machine translation before, which machine translation solutions are employed (free or paid), the prompt process, etc.  LocalizeDirect’s clients, after discussion and reflection, hence can weigh their priorities and make informed decisions based on specific needs and situations.

Feel free to share your inquiries, and let’s start a conversation about how we can tailor a localization plan specifically to meet your service preferences.