Paper: Watermarking the Outputs of Structured Prediction with an application in Statistical Machine Translation.

ACL ID D11-1126
Title Watermarking the Outputs of Structured Prediction with an application in Statistical Machine Translation.
Venue Conference on Empirical Methods in Natural Language Processing
Session Main Conference
Year 2011
Authors

We propose a general method to water- mark and probabilistically identify the structured outputs of machine learning al- gorithms. Our method is robust to lo- cal editing operations and provides well defined trade-offs between the ability to identify algorithm outputs and the qual- ity of the watermarked output. Unlike previous work in the field, our approach does not rely on controlling the inputs to the algorithm and provides probabilistic guarantees on the ability to identify col- lections of results from one’s own algo- rithm. We present an application in statis- tical machine translation, where machine translated output is watermarked at mini- mal loss in translation quality and detected with high recall. 1 Motivation Machine learning algorithms provide structured results to input q...