Build an MQM model for the task
Choose error categories and subcategories, severity levels, weights, the scoring method and the pass threshold for a given text type and service.
Measure translation quality with numbers
An online course for editors, QA specialists and translation managers. Participants build their own error typology based on MQM, annotate errors by category and severity, and calculate a quality score with a pass/fail threshold. RQS and CQS are calculated on training texts.
- Format
- online, with feedback
- Duration
- 4 weeks
- Language of instruction
- English
- Price
- on request
- Outcome
- your own MQM-based model
01.Who it is for
Course audience
For people responsible for translation quality who need to measure it and justify the findings.
Suitable for:
- editors who want to formalise assessment and support their corrections with a score;
- heads of translation or QA departments setting up quality control;
- project managers who review vendor test translations;
- translators and freelancers who want to offer LQA as a separate service;
- specialists who carry out external quality audits of translation departments.
Less suitable if the goal is:
- to learn translation rather than how to assess it;
- a single universal score that does not need to be configured for the task;
- theory only, without annotation or practice on texts.
The course involves spreadsheets, error categories and calculations.
02.Outcomes
After the course
Annotate errors consistently
Classify errors with the MQM 2.0 Core typology and assign severity: Neutral, Minor, Major, Critical.
Calculate RQS and CQS
Calculate the raw score (APT → PWPT → RQS) and the calibrated score, using a threshold and scaling so that results are not capped at 100.
Build LQA into processes
Apply assessment to pre-delivery checks, vendor test translations, periodic translator evaluation and external quality audits.
03.Programme
Four modules
-
01
What LQA is and where it is used
Linguistic quality assessment evaluates a translation by counting objective errors against a chosen model. Four scenarios: pre-delivery checks, test translations, periodic and ad hoc evaluation of translators, LQA as a separate service and external audit.
-
02
Components of an LQA model
Error categories and subcategories, severity levels, weights, the scoring method and the pass/fail threshold, and how they form a working assessment scheme.
-
03
MQM as a general-purpose metric
Multidimensional Quality Metrics (MQM), created in 2014 by the German Research Center for Artificial Intelligence (DFKI) on the basis of LISA QA. The module covers configuring MQM for different services, text types and subject areas; MQM 2.0 Core and Full, and why Core is used in practice; and severity levels and penalty points, calculated as category weight × severity multiplier.
-
04
Scoring: RQS, CQS and holistic assessment
Raw Quality Score without calibration (APT → PWPT → RQS) and the Calibrated Quality Score: RWC, maximum score, passing threshold, scaling factor. When to use non-linear formulas (samples above 5,000 words) and when to use holistic assessment on the Correspondence and Fluency scales.
Example: Terminology (weight 1) × Major (5) = 5 penalty points; then RQS = (1 − PWPT) × 100.
04.Format
How the course runs
Weekly online sessions
Theory and joint annotation. Recordings remain available to participants.
Practice on texts
Building an MQM model for a task, annotating errors, calculating RQS and CQS. Materials include an MQM master spreadsheet and sample LQA annotations.
Feedback
The teacher reviews annotations and calculations, points out disputable classifications and helps calibrate the model.
05.Teacher
Who teaches the course
Yuri Vodostoy, founder of INSIGHT, has taught Linguistic Quality Assessment and Translation Quality Metrics at Tomsk State University since 2024. The course is based on the university programme, taught to two cohorts of master's students, and adapted for translation teams and freelancers.
06.Options
Individual and team participation
Individual participation — price on request
- 4 weeks, online
- access to materials and recordings
- MQM master spreadsheet and sample LQA annotations
- weekly online sessions
- review of annotations and calculations
- an MQM model built for your task
By default, participants receive a certificate of completion.
For teams — on request
- training for a translation or QA department
- an LQA model for the company's processes
- implementation and quality audit
- flexible schedule and group composition
07.FAQ
Questions
.01Is a background in translation studies required?
The course is for people who already work in translation: editors, managers, translators. No mathematical background is required; we explain RQS and CQS calculations from the start, with examples.
.02How is LQA different from editing?
Editing corrects the text. LQA measures quality: it records objective errors against a chosen model, assigns severity and produces a numerical score with a pass threshold. LQA can be a separate stage or a standalone service.
.03Why MQM?
MQM is a general-purpose base metric that can be configured for different services, text types and subject areas. Most current models in LQA tools are built on it.
.04What materials are used?
Participants build an MQM model, annotate errors and calculate scores on training translations, using an MQM master spreadsheet and sample LQA annotations.
.05When does the next group start?
We form groups as applications come in. Contact us and we will send the start date and the detailed programme.
08.Enrol
Enrolment
Email edu@insgt.pro with the subject "LQA course", or write in Telegram to @insgtmgr. We confirm the dates of the next group and the participation format.