Our new LQA service Perfectionist is based on the most recent advancements in language quality assessment science as described in a standards proposal (WK54884) created in 2014, with new significantly updated version of it recently submitted to ASTM.
You are welcome to use LQA service as is with one of the pre-defined LQA metrics, or you can ask us to tailor the service to your own LQA metrics.
Professional, trained terminologists and linguists are available for all languages.
Additional services available, such as validating your own LQA metrics, or creating your metrics from scratch.
Over and over again in our industry, as we struggle to meet tighter and tighter deadlines while managing costs, we ask ourselves, “Is it possible that something as fascinating as human language can be squeezed into a Procrustean box rife with MT, mechanical processing and unqualified providers? Is there something wrong with how our language is treated, and is the mystery of language the reason why translation evaluation is subjective? Over and over again we are perplexed with the following questions:
- Why different companies are using so many different LQA metrics?
- Is there a way to validate them?
- Why linguists and experts sometimes disagree with the results of a particular LQA metrics?
- Evaluations are subjective – Why? And how big is the difference of opinion?
- Could there be one common approach?
- How LISA QA Model, MQM, DQF and other evaluation methods relate to each other?
- We are all fascinated by translation, so why is the gold standard so elusive?
- How to build a good detailed LQA metrics?
- Is there a cheaper and faster way than counting all errors and inserting them into a formula?
- Is there a way to do LQA quickly?
All these and other questions are answered in the proposed ASTM WK54884 standard which establishes the following (summarized):
- Text (and even more, translation) is a complex system which bears many properties of other complex systems, most notably:
- (a) the whole is more than its parts,
- (b) elements of text interact with each other, and enter into a complex relationship.
- Mechanisms from systems theory are applicable to text as a medium consisting of active, iridescent parts (words, phrases, sentences, paragraphs and chapters), which dynamically interact to form varying perceptions of human readers. These parts and their interaction are what makes text such a magnificent and amazing means of human communication, so fascinating for readers and of course (even more) for translators.
- The key to translation evaluation is holistic assessment of two key independent dimensions: Accuracy and Readability.
- The nature of text comprehension and impression is sequential, the reader progressively builds a “running total” holistic assessment in his mind. In parallel, he conducts (consciously or not) an atomistic error-based assessment, automatically recalibrating atomistic error “weights” to align with his holistic impression.
- Based on trials, it was demonstrated that holistic evaluations should be measured by a standard 10-point universal grade scale with specified categories.
- For every text, an atomistic metrics exists which is calibrated by correct error categories and weights so that it gives the same result as a holistic assessment of accuracy and readability of the same text. Such an atomistic metrics is “well-formed”.
- A method to build a well-formed detailed metrics, validated by holistic assessments, is proposed.
Please, contact us for more information about ASTM standard draft WK54884 and Cloud-based LQA assessment service.
You are welcome to use LQA service as is with one of the pre-defined LQA metrics, or you can ask us to tailor the service to your own LQA metrics.
Professional, trained terminologists and linguists are available for all languages.
Additional services available, such as validating your own LQA metrics, or creating your metrics from scratch.