In Part 1 of this Blog Series, Translation Quality Assessment: The Why, What and How, we discussed the importance of assessing translation quality. In Part 2, Translation Quality Assessment: Measuring Defects, we began our deep dive into the “how” of translation quality assessment with discussion of how to measure errors or “defects.”
In this Part 3, we’ll continue our detailed analysis of the “how” by looking at error categories and severity. Not all errors are created equally. Rather, there are two dimensions used to calculate an error’s weight in any translation. Error categories and severity differ from one standard and/or metric to another and from one company to another, however there are a lot of similarities between all of them.
The first dimension in evaluating quality is categorizing errors. Some error categories that are common for all types of content, while others are more important for specific types of content. Below is a list of common error categories and their definitions:
- Accuracy: Accuracy is about how accurate the translation is compared to source. It measures whether the words, meanings, messages and intent were clearly conveyed into the target language. Accuracy requires real understanding of the source, so linguists are always encouraged to ask questions about the source (especially if the source has more than one meaning) in order to avoid any ambiguity in the target.
- Terminology: Terminology is about adhering to product or client or domain glossary. Clients may have a glossary of terms for their products/services that is translated into other languages. The glossary may include brand names, product names, and/or proper names and their preferred translation, transliteration, or transcription into other languages. The glossary may also include the standard translations for user interface items in software localization, and/or operating system standard translations.
- Consistency: Consistency is a broad concept related to ensuring that translation is consistent within the current project, with previous translations, across products, and in context. It is about making sure that key source words, terms and phrases are consistently translated in the target language. Following the same style and tone is considered part of the consistency check. When abbreviations and/or acronyms are used in translations then they should be used consistently. It also includes using specific numbering and date formats consistently.
- Spelling: Spelling has to be correct in the target language and should follow the spelling standardized by the official language standards organization (if any) and/or official language dictionary. If there are different ways of spelling specific words/characters in the target language because it is spoken in different countries/regions, then the most common spelling in the target country should be used unless the client directs otherwise.
- Completeness: Completeness measures whether all words, terms, concepts and meaning in the source are fully conveyed in the target. For specific types of translation, it is crucial to translate every word, but for other types it is more important to translate the meanings and concepts, even if some source words were skipped in the translation. It is also meant to ensure that numbers, dates and any other variables that exist in source segments are translated properly in the target. If words have been added to the target that were not in the source, completeness looks as whether they were added to convey a message properly, or if they were added in error, changing the meaning of the final product.
- Structure: Structure measures sentence structure. Sentence structure differs from one language to the other, so translation into some languages means changing the whole sentence structure including positions of verbs, propositions, nouns, adjectives and articles (if any). Some languages require translating back to front. Part of correct translated sentence structure is to correctly position variables, placeholders and tags in accordance with language reading order.
- Flow/readability: Flow or readability evaluates ease of reading and understanding the translation for target audience profile. Ambiguity sometimes comes from the source language; but it is the linguists’ role to ask for clarifications (or even editing) before starting translation. If a source sentence has tags, is divided into different segments, or contains placeholders, then linguists have to ensure they adjust the translations accordingly, or even try to adjust the source segmentation before the translation starts.
- Formatting: Formatting means ensuring that translated content format is as close as possible to source format, but follows the target language formatting rules and reading order. Formatting includes layout, type styles, text direction, font, size, orientation, alignment, currency, numbering format and date format. Formatting is handled differently based on content file format. It can be represented by tags or physical formatting of the text.
- Punctuation: Punctuation differs between languages and must be translated from source to target language. This error category measures whether correct punctuation was used in translations. Full stops, commas, colons, semi colons, exclamation marks, and question marks have different rules of usage in many languages—they even have different shapes. Some marks have different usage conventions in various languages. Spacing before and after punctuation marks differ from one language to another.
- Style: Style may or may not be assessed, depending on the type of content. If the client has language style guides or multilingual branding guidelines, then style should be measured against those guidelines. Style also differs from one language to another and from one target audience profile to another, so it is highly recommended that linguists learn about the client’s style and tone of voice preferences before starting to translate. The tone and formality should be adjusted to the target audience profile. This is one of the categories that sparks debate since it’s very subjective and subject to personal taste. Often, stylistic changes are called “preferential changes” and are not counted as errors.
The second dimension is assigning a severity level to each error. Severity levels are mainly used to represent content usability and avoidance of liability. Errors in translations that may cause legal liability against the client are critical and will stop the translated content from being published or the product/service from being shipped/launched.
Another level of severity is assigned when errors mislead the user, block him/her from using the product/service, or provide incorrect and/or contradictory information. These types of errors are considered major. Translations with major errors are typically not usable.
Minor errors include those that might make the translation harder to understand, or cause some confusion, but ultimately do not stop the user from using the content, product, or service. Minor errors may affect the user’s comprehension of content, productivity, and/or the time required to read and understand.
Assessing Errors by Category and Severity
Translation quality is ultimately assessed by assigning a severity level to each error, assigning a value to each type of error and each severity level, and coming up with a “score.” Much like a teacher might still award a high grade to a student’s 30-page final paper, even though the student misspelled two words, the ultimate score that will be considered “passing” is not typically 100%. What constitutes a passing score will depend on the quality assessment model used, industry standards, and/or standards agreed upon with the client.
Understanding how to measure errors, and how to categorize them, gives us a solid foundation upon which to build a plan for avoiding them. In Part 4 of this Blog Series, we’ll turn our attention to planning for quality translation.
Want to learn more about Vocalink Global’s translation services? Connect with us today!