{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/train-sort-explain-learning-to-diagnose","title":"Train, Sort, Explain: Learning to Diagnose Translation Models","arxiv_id":"1903.12017","date":"2019-03-28","proceeding":"NAACL 2019 6","authors":["Robert Schwarzenberg","David Harbecke","Vivien Macketanz","Eleftherios Avramidis","Sebastian Möller"],"abstract":"Evaluating translation models is a trade-off between effort and detail. On\nthe one end of the spectrum there are automatic count-based methods such as\nBLEU, on the other end linguistic evaluations by humans, which arguably are\nmore informative but also require a disproportionately high effort. To narrow\nthe spectrum, we propose a general approach on how to automatically expose\nsystematic differences between human and machine translations to human experts.\nInspired by adversarial settings, we train a neural text classifier to\ndistinguish human from machine translations. A classifier that performs and\ngeneralizes well after training should recognize systematic differences between\nthe two classes, which we uncover with neural explainability methods. Our\nproof-of-concept implementation, DiaMaT, is open source. Applied to a dataset\ntranslated by a state-of-the-art neural Transformer model, DiaMaT achieves a\nclassification accuracy of 75% and exposes meaningful differences between\nhumans and the Transformer, amidst the current discussion about human parity.","url_abs":"http://arxiv.org/abs/1903.12017v1","url_pdf":"http://arxiv.org/pdf/1903.12017v1.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"train-sort-explain-learning-to-diagnose","repo_url":"https://github.com/dfki-nlp/diamat","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"tf","reach":null}],"tasks":[{"task_slug":"translation","task_name":"Translation"}],"methods":[{"method_slug":"absolute-position-encodings","method_name":"Absolute Position Encodings"},{"method_slug":"adam","method_name":"Adam"},{"method_slug":"attention","method_name":"Attention"},{"method_slug":"bpe","method_name":"BPE"},{"method_slug":"dense-connections","method_name":"Dense Connections"},{"method_slug":"dropout","method_name":"Dropout"},{"method_slug":"label-smoothing","method_name":"Label Smoothing"},{"method_slug":"layer-normalization","method_name":"Layer Normalization"},{"method_slug":"linear-layer","method_name":"Linear Layer"},{"method_slug":"multi-head-attention","method_name":"Multi-Head Attention"},{"method_slug":"position-wise-feed-forward-layer","method_name":"Position-Wise Feed-Forward Layer"},{"method_slug":"relu","method_name":"ReLU"},{"method_slug":"residual-connection","method_name":"Residual Connection"},{"method_slug":"softmax","method_name":"Softmax"},{"method_slug":"transformer","method_name":"Transformer"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}