a web-based translation post-editing environment for use with cdec Realtime.


Michael Denkowski and Alon Lavie


TransCenter provides a web-based translation editing interface that remotely monitors and records user activity. The "live" version now uses cdec Realtime to provide on-demand MT that automatically learns from post-editor feedback. Translators use a web browser to access a familiar two-column editing environment from any computer with an Internet connection. The left column displays the source sentences, while the right column, initially empty, is incrementally populated with translations from the Realtime system as the user works. For each sentence, the translator edits the MT output to be grammatically correct and convey the same information as the source sentence. During editing, all user actions (key presses and mouse clicks) are logged so that the full editing process can be replayed and analyzed. After editing, the final translation is reported to the Realtime system for learning and the next translation is generated. The user is additionally asked to rate the amount of work required to post-edit each sentence immediately after completing it, yielding maximally accurate feedback. The rating scale ranges from 5 (no post-editing required) to 1 (requires total re-translation). TransCenter also records the number of seconds each sentence is focused, allowing for exact timing measurements. A pause button is available if the translator needs to take breaks. TransCenter can generate reports of translator effort as measured by (1) keystroke, (2) exact timing, and (3) actual translator post-assessment. Final translations are also available for calculating edit distance. Millisecond-level timing of all user actions further facilitates time sequence analysis of user actions and pauses. Figure 3 shows an example summary report generated by TransCenter showing a user’s activity on each sentence in a document. This information is also output in a simple comma-separated value format for maximum interoperability with other standards-compliant tools.


IP Agreement

Required Acknowledgment

Michael Denkowski; Alon Lavie; Isabel Lacruz; Chris Dyer, "Real Time Adaptive Machine Translation for Post-Editing with cdec and TransCenter" Proceedings of the EACL 2014 Workshop on Humans and Computer-assisted Translation, 2014
[PDF] [bib]

If this information is inaccurate or incomplete, please submit an update through this form.