Towards a CAT tool agnostic standard for User Activity Data

The dominance of the source word pricing model combined with the fact that most translators work as freelancers has led to a scenario in which until recently most buyers (direct and intermediary) who work with freelancers neither knew nor cared how many words per hour the translators they hire translate. However, this situation is beginning to change. Machine translation has shown that it is possible for translation requesters to impact positively on words per hour productivity. In addition to classical full-sentence MT, advances in various typeahead technologies have resulted in a situation in which a number of options are available to impact positively on a translator’s working speed and terminological consistency with previous translations. Finally, evidence is beginning to emerge that productivity gains can be achieved where translators use Automatic Speech Recognition to dictate rather that type the target text. In this paper we will provide a brief overview of these technologies and use cases the impact on translator productivity and describe an architecture to gather translation process data to measure their impact from working translators in a maximally unobtrusive way. We propose an open-standard for User Activity Data in CAT tools (CAT-UAD) so that they can work in any CAT tool that implements this standard and outline a technical architecture to gather such data conveniently and a privacy model that respects translator, intermediary and end-client data sharing concerns and discuss various A/B testing scenarios that can be tested using Segment Level A/B testing.