LingoHub’s Marketplace: Introducing a new feature in LingoHub Next
In this article, we will focus on the main intricacies connected to translation service providers and their services. There are several industry and translation standards, as well as implemented modes of operation in place.
We’ll start off with the obvious one first, ISO standards. Later on, we’ll take a closer look at price calculations – in theory and practice. It will also illustrate LingoHub’s new take on price transparency and calculation with our new release coming soon. Here's the index:
- ISO standards
- Translation price calculation: prerequisites
- Translation price calculation: calculation methods
- Actual price calculation
- Price calculation in LingoHub’s new Marketplace
Translation managers and translators, amongst others, will all know about the ISO standards in place to regulate translations. Nonetheless, I’d like to introduce the standards to those who are unfamiliar with them.
ISO 17100:2015: Translation services — Requirements for translation services, was published in 2015. In the standard, requirements for quality translations are listed, such as core aspects of translations, and necessary resources. Translation service providers (TSPs) may use this standard to indicate their conformity, by either stating that they are adhering to the standard, or by being certified and audited regularly. The latter, of course, illustrates that an independent auditor regularly checks the conformity. If you are in search of expert translators or TSPs, the ISO 17100:2015 certification is a very good starting point. In addition to ISO 17100:2015, another standard might be relevant, ISO 18587:2017.
ISO 18587:2017: Translation services — Post-editing of machine translation output — Requirements was published in 2017. In the standard, requirements for the process of human post-editing of machine translated content are listed. In addition, the competences of post-editors are recorded.
ISO 9001:2015: Quality management systems — Requirements was published in 2015. It is the only standard in the ISO 9000 series that companies can obtain certification to. While it is not directly associated with translation services, it still ensures that companies are documenting their processes well. It also ensures that constant improvement and evaluation of processes take place. This, of course, also benefits customer satisfaction.
Now that the three relevant standards have been introduced, let’s take a look at how prices are calculated in the translation industry.
Translation price calculation: prerequisites
Several things have to be considered when calculating translation costs. Language pairs, text difficulty, proofreading, sometimes even urgency. Various options affect the price, way before a quote is generated.
Language pair: source and target language
The combination of source and target language is often referred to as a language pair. First and foremost, the language pair determines the price. Translating English source texts to Romance languages, e.g., Spanish or French, is usually less complicated and therefore cheaper than translating English to, e.g., Norwegian or Finnish. Another thing to consider is that some language pairs are less common than others. While English as a source language can be translated into almost any other language, it might be very challenging to find a translator for Spanish to Tamil.
Industry: text difficulty and text type
Many translators have a field of expertise. Some are very specialized in legal texts, others are more versed in search engine optimization or marketing. Depending on the complexity of the source texts, TSPs typically provide three different tiers for their services, e.g., standard, advanced, and expert.
A second set of eyes: proofreading
The majority of (if not all) translation service providers offer a proofreading service for translation orders. Dual control helps ensure that even lengthy, or complex translations are on point. In addition, ambiguity may also be reduced through the four-eyes principle. While the proofreading option often comes with higher costs, it also ensures great accuracy for complicated translations. There is one catch, of course: Not every TSP offers proofreading only services, meaning that proofreading is only available when paired with prior translations. In addition, TSPs handle proofreading differently. The majority of TSPs compare source and target texts when proofreading, but some perform proofreading on target text segments only, without comparing it to the source texts. Some might also only offer sole proofreading for specific languages instead of supporting all language options.
Post-editing of machine translations
A fairly new service that TSPs offer is post-editing machine translated text segments. Machine translations are getting better by the minute. The ISO 18587 standard has been introduced for 5 years now, and TSPs adapt to that situation. There is a major difference between post-editing already machine translated text segments and starting from scratch. Still, no machine translation can match translations performed by a professional translator. There are always some details or cultural intricacies that a machine is unable to predict or detect. Depending on the TSP, machine translation post-editing is sometimes treated like proofreading, but some may list this as an extra service as well.
Translation price calculation: calculation methods
Depending on the TSP and the industry, four main ways of price calculation are common. The list below illustrates these ways, from most common to least common:
- Per word calculation
- Per hour calculation
- Standard page
- Standard line
Per word calculation
Most commonly, translation prices are calculated based on the count of source words. While the target language word count is different to the source, it is still a standard used by the majority of TSPs. Customers get an estimate of costs, even before the translation has started. As we pointed out in a previous post, average text expansion can range between 130 % and up to 300 %. So this model favors target languages with shorter translations or wordy source languages.
Per hour calculation
It is especially common for German translators to use an hourly rate for price calculations. While this is definitely less common nowadays in general, it is easily understandable when looking at the German language. German is filled with long and complex compound words, e.g., Donaudampfschiffahrtskapitän (Danube steamship captain), Rechtsschutzversicherungs- gesellschaften (insurance companies providing legal protection), or Siebentausendzweihundert- vierundfünfzig (7,254). As you can see, translating one compound word might result in multiple words in other languages. While translations calculated per hour are rather uncommon, they still do exist.
A standard page contains 1500 characters without spaces. Prices vary, depending on industry, translator qualifications, text difficulty and much more.
Standard lines are 55 characters long. Again, prices vary, depending on many factors. Standard lines are less common around the world, but quite common in Austria and Germany.
Actual price calculation
With the prerequisites established, and price calculation methods elaborated, let’s focus on real life examples. Since the majority of TSPs calculate price by word count, we will focus on that. There is still a fundamental part unexplained in price calculation, which is connected to text matches. Matching is an integral part of translation cost calculation. Of course, the first translation will always be priced at 100 % of per-word costs. But in time, and with more translations, a translation memory (TM) can be built. This introduces the match variable. In general, two types of matches can be made, these will be explained below.
In fuzzy matching, source text segments are compared against each another. The match rate is calculated based on the similarity of two source segments that need to be translated.
Let’s look at an example: You have previously ordered the translation of the text segment “This is a blue car”. A TM was built storing this text segment.
A second order two months later contains the text segment “This is a blue car”. The system will calculate a 100 % TM match. In the second order, there is a second text segment, “The radios are broken”. For this segment, the system will calculate a 0 % match, since there is no word match with the translation memory. A third text segment, “This is a red car”, will yield an 80 % match, since 4 out of 5 words are matched with the TM.
|Order #||Text segment||Match in TM||Calculated TM match|
|1||This is a blue car.||No||0 %|
|2||This is a blue car.||Yes||100 %|
|2||The radios are broken.||No||0 %|
|2||This is a red car.||Partially||80 %|
Depending on the TSP, this simplified example might not apply. There are several other methods to calculate the fuzzy match percentage. Algorithms like the Levenshtein Distance and the Jaro-Winkler Similarity can also be used. You can learn more about these calculations in this recent blog post. You can find short and simplified descriptions below, however.
The Levenshtein Distance is the amount of single character edits to change a source word to a target word. Three edit types are performed by the algorithm: insert, delete or substitute.
The Jaro-Winkler Similarity algorithm uses a prefix scale. The similarity value is between 1 and 0. A rating of 1 shows that the two compared strings are identical, 0 means the strings have no similarity.
Longest common subsequence
The longest common subsequence checks each character of two or more text segments and finds the longest common subsequence in them. Similar to the Levenshtein Distance, it uses insert and delete operations, but no substitutions.
Translation service providers create match percentage brackets and calculate different prices for different brackets. As an example, these brackets might be used to calculate prices per word:
- Repetition or 100 % match
- 95 % - 99 % match
- 85 % - 94 % match
- 75 % - 84 % match
- 50 % - 74 % match
- < 50 % | No match
Even with repetitions and 100 % matches, it is prudent to make sanity checks. That’s why TSPs usually charge for 100 % matches, too. This is often referred to as a base rate.
Similar to fuzzy matching is fuzzy distance. However, instead of comparing source text segments, target text segments are compared. This method is used to calculate the effort needed for post-editing or proofreading target text segments. Fuzzy distance comes into play when calculating how high the necessary changes of target text segments will be. A fuzzy distance of 50 % or higher, typically requires a complete rewrite of the target text, whereas a 0 % differences indicates that the text segments are identical. As with fuzzy matching, a 0 % change score still requires a basic check by translators/reviewers to be sure that the translation is correct. The base rate applies here, too. Of course the same algorithms may be used for calculating the fuzzy distance.
Price calculation in LingoHub’s new Marketplace
All the information so far was a necessary prerequisite to better understand how LingoHub’s price calculation in our brand-new Marketplace works. With the word Hub being in our very name, we always dreamed about a place where translators and customers can freely collaborate. With our next big release, we will do just that and open up our brand-new Marketplace. And all that will be possible with transparent pricing models, too!
For our marketplace, we wanted to utilize the best of the best in the translation industry. Therefore, we established a modern, transparent pricing model that is fair to both customers and translators alike. Our marketplace billing is based on online-research about how translators are paid correctly. You can find all the details below. In short, we will utilize two rates, the base rate and the edit effort rate.
Every translation order has a base rate. It will always be charged, regardless of the necessary change efforts. The base rate will be multiplied with the source word count.
Let’s look at an example: Your source language is English, 100 words are present in your current project. You want to order the Spanish translation for your project, the price for this language pair is €0.20. This means that your base rate will be 100 words x €0.20 = €20.
Of course, this is only an example, the base rate will be defined by the language pair or a default rate.
Edit effort calculation and rate
The edit effort is calculated based on the target content changes. Every change in the source text segment will result in a new price calculation. When the text fully changes the edit effort will be 100 % and if the text doesn’t change at all the edit effort will be 0 %. Three algorithms are used for calculation (follow the links for full equations):
Based on the combination of these equations, we defined three rates: low, middle, and high. Of course, there is also the possibility that no edit effort is necessary. So in total, four options are possible:
- Edit effort is 0 % (zero changes to the target text): Only the base rate will be charged.
- Edit effort is between 1 % and 9 %: The base rate and the low rate will be charged.
- Edit effort is between 10 % and 49 %: The base rate and the middle rate will be charged.
- Edit effort is between 50 % and 100 %: The base rate and the high rate will be charged.
Let’s look at some basic examples
Prices: base rate: €0.20, low rate: €0.02, middle rate: €0.03, high rate: €0.05
7 source words were translated to 9 target words
The calculated edit effort is 100 %: base and high rate will be used
Calculation: 7 words x €0.20 + 7 words x €0.05 = €1.40 + €0.35 = €1.75
Prices: base rate: €0.20, low rate: €0.02, middle rate: €0.03, high rate: €0.05
7 source words were translated to 9 target words
The calculated edit effort is 30 %: base and middle rate will be used
Calculation: 7 words x €0.20 + 7 words x €0.03 = €1.40 + €0.21 = €1.61
You can find more information on price calculation in our Help Center soon.
Wrapping it up
With this aggregated information on translation orders, prices, calculations, and TSPs, We hope you have a proper understanding of all the things involved when ordering translations. Our new Marketplace will open up a lot of possibilities and more transparency for everyone involved. And the cool thing? You can try it out very soon! Our LingoHub Next beta is already on the way. Any questions? Want to join the beta? Let us know! Simply write to our support.