towards an enterprise aggregated solution for dqf - fred tuinstra (lionbridge)
TRANSCRIPT
TAUS QE Summit – Dublin Jun 8Productivity Metrics – the proverbial minefield…
What data do we want to capture, and do we even care?
Translation Productivity | What metrics do we need?
• Hardly anybody buys a car without looking at the labels
• Plenty of confusion –
– mpg, km/l, l/100km
– fuel types & carbon emissions
– Urban, City, Highway
– But mostly agreed units and measurement practices, although…
• Are Translation Productivity Metrics like Fuel Consumption Data?
Productivity Metrics | Areas where they provide value
• Cost / Effort / Value – continuous validation of pricing models– Source Complexity– Fuzzy grids– Edit Distance Data– Ratios between per hour / per word
pricing, both for Selling and Buying
• Scheduling & Planning – accurate prediction of throughputs
• Resourcing – “Improvement Potential” of new
resources– Optimized Task assignments
Productivity Metrics | More questions than answers
• Bare Productivity (throughput) numbers don’t mean much without reference to things like– Complexity of the input– Quality of the output– The actual activity (T/E/P)– Identity (experience level of a resource) – Location– MT engine in use– Etc., etc.
• Large data sets will take care of outliers but we need to be able to further dissect
• Case Study done on a distinct program –– Fuzzy grid in use for a particular program
was inadequate– however, net new words throughput was
much higher at nearly 400 words / hour
Productivity Metrics | Challenges for an LSP
• Variability in ‘work types’, customer quality standards, resource qualifications, instructions, toolsets, MT engines, and more
• “Linguistic” productivities are only part of the total TAT calculation
• Most translation tasks are paid ‘per unit’
– Productivity (speed) is rarely a goal in itself
– We don’t instruct translators that they will be ‘measured’ for productivity
Agreed definitions and metadata on how to represent productivity metrics