Abstract
Traditionally, human–machine interaction to reach an improved machine translation (MT) output takes place ex-post and consists of correcting this output. In this work, we investigate other modes of intervention in the MT process. We propose a Pre-Edition protocol that involves: (a) the detection of MT translation difficulties; (b) the resolution of those difficulties by a human translator, who provides their translations (pre-translation); and (c) the integration of the obtained information prior to the automatic translation. This approach can meet individual interaction preferences of certain translators and can be particularly useful for production environments, where more control over output quality is needed. Early resolution of translation difficulties can prevent downstream errors, thus improving the final translation quality “for free”. We show that translation difficulty can be reliably predicted for English for various source units. We demonstrate that the pre-translation information can be successfully exploited by an MT system and that the indirect effects are genuine, accounting for around 16% of the total improvement. We also provide a study of the human effort involved in the resolution process.
Similar content being viewed by others
Notes
This problem no longer occurs when PE is complemented with adaptive learning. However, in such scenarios the human is not usually asked explicitly to provide generalized translations for all the corresponding contexts. Therefore, the corrections of recurrent translations may still be required.
Grid search with 5-fold cross-validation was used to tune the following parameters towards F-score: the optimizing criterion, the number of estimators, the maximum depth and the minimum number of leaf samples. All other parameters are those provided by default.
We use a Radial Basis Function (RBF) kernel. Grid search with 5-fold cross-validation was used to tune \(\gamma \) and C towards F-score. All other parameters are those provided by default.
We used the l-bfgs algorithm as the optimization algorithm. All other parameters are those provided by default. All the hyperparameters were tuned on the development set.
The first hidden layer contained the quantity of units equal to the total quantity of features (the embedding dimension was taken into account, e.g. 1822 hidden units for word-seg), the second layer used half that number of units. The following features served as inputs to embedding layers: word \(f_i\), its left and right context \(f_{i-1}\) and \(f_{i+1}\), its aligned word \(e^1_{j}\), its left and right context \(e^1_{j-1}\) and \(e^1_{j+1}\) for word-seg; sequence \(f_{[k:t]}\), its aligned translation \(e_{[r:g]}\) and its the left (\(f_{k-2}, f_{k-1}\)) and right (\(f_{t+1}, f_{t+2}\)) contexts for phrase-level segmentations. The length of a segment sequence was limited to 10 words, and masking was used for shorter segments.
We used the DummyClassifier with default parameters.
Translations of such nouns tend to vary greatly depending on the context, even when natural translation variability is taken into account. Some of them are also homonymous to verbs, which contributes further to their translation difficulty.
We artificially balanced the quantity of examples in both classes for the language pairs where we found an unbalanced proportion of ET and DT (EN-ES, EN-FR) by removing the least frequent examples of ET. This resulted in a reduction of around 34% of the initial training data and a prediction improvement of about 0.07 in \(F_{\texttt {DT}}\).
In all the experiments, we will translate complete texts rather than isolated random sentences. Our reference translations are thus likely to be somewhat “normalized”, i.e. to contain less translation variety than random sentences.
In a more elaborate version, the user could select these translations from the variants proposed by an MT system (Cheng et al. 2016), or/and from the cache of past translations of DT segments, thus potentially saving many keystrokes.
For this experiment, unaligned words in the reference \(\hat{e}\) are aligned (recursively) to the same word(s) as their syntactic heads. Dependencies were identified with the help of the Stanford Parser toolkit.
We round word averages to nearest integer, which may result in several TER values per each rounded value. We report average TER values for such cases.
Recall that DT segments represent approximately 50% of a sentence, cf. Sect. 3.1.1.
\(\varDelta \) TER is computed as the absolute difference between the initial MT quality score (a point with a pair of coordinates (0,n)) and a quality score resulting from a pre-translation experiment (any other point). A maximum quality gain is computed as the absolute difference of the y-coordinate values for points (0,n) and (max(x),m).
References
Alabau V, González-Rubio J, Ortiz-Martínez D, Sanchis Trilles G, Casacuberta F, García-Martínez M, Mesa-Lao B, Petersen DC, Dragsted B, Carl M (2014) Integrating online and active learning in a computer-assisted translation workbench. In: Proceedings of the AMTA 2014 workshop on interactive and adaptive machine translation, Vancouver, BC, Canada, pp 1–9
Aronson AR, Lang FM (2010) An overview of MetaMap: historical perspective and recent advances. J Am Med Inform Assoc 17(3):229–236
Bahdanau D, Cho K, Bengio Y (2014) Neural machine translation by jointly learning to align and translate. CoRR abs/1409.0473
Blanchon H (1992) A solution for the problem of interactive disambiguation. In: COLING 1992 volume 4: the 15th international conference on computational linguistics. Nantes, France, pp 1234–1238
Bojar O, Chatterjee R, Federmann C, Graham Y, Haddow B, Huck M, Jimeno Yepes A, Koehn P, Logacheva V, Monz C, Negri M, Neveol A, Neves M, Popel M, Post M, Rubino R, Scarton C, Specia L, Turchi M, Verspoor K, Zampieri M (2016) Findings of the 2016 conference on machine translation. In: Proceedings of the first conference on machine translation: volume 2, shared task papers, Berlin, Germany, pp 131–198. https://doi.org/10.18653/v1/W16-2301
Bojar O, Chatterjee R, Federmann C, Graham Y, Haddow B, Huang S, Huck M, Koehn P, Liu Q, Logacheva V, Monz C, Negri M, Post M, Rubino R, Specia L, Turchi M (2017) Findings of the 2017 conference on machine translation (WMT17). In: Proceedings of the second conference on machine translation, Copenhagen, Denmark, pp 169–214
Breiman L (2001) Random forests. Mach Learn 45(1):5–32
Brown RD, Nirenburg S (1990) Human-computer interaction for semantic disambiguation. COLING 1990 volume 3: papers presented to the 13th international conference on computational linguistics. Helsinki, Finland, pp 42–47
Bubeck S, Cesa-Bianchi N (2012) Regret analysis of stochastic and nonstochastic multi-armed bandit problems. Found Trends Mach Learn 5(1):1–122
Chatterjee R, Negri M, Turchi M, Federico M, Specia L, Blain F (2017) Guiding neural machine translation decoding with external knowledge. In: Proceedings of the eighth workshop on statistical machine translation, Copenhagen, Denmark, pp 157–168
Cheng S, Huang S, Chen H, Dai XY, Chen J (2016) PRIMT: A pick-revise framework for interactive machine translation. In: Proceedings of the 2016 conference of the North American chapter of the association for computational linguistics: human language technologies, San Diego, California, pp 1240–1249, https://doi.org/10.18653/v1/N16-1148
Cherry C, Foster G (2012) Batch tuning strategies for statistical machine translation. In: Proceedings of the 2012 conference of the North American chapter of the association for computational linguistics: human language technologies, Montrèal, Canada, pp 427–436
Cortes C, Vapnik V (1995) Support-vector networks. Mach Learn 20(3):273–297
Denkowski M, Dyer C, Lavie A (2014) Learning from post-editing: Online model adaptation for statistical machine translation. In: Proceedings of the 14th conference of the European chapter of the association for computational linguistics, Gothenburg, Sweden, pp 395–404
Du J, Srivastava A, Way A, Maldonado-Guerra A, Lewis D (2015) An empirical study of segment prioritization for incrementally retrained post-editing-based SMT. In: Proceedings of the XV machine translation summit, Miami, Florida, pp 172–185
Dyer C, Chahuneau V, Smith NA (2013) A simple, fast, and effective reparameterization of IBM model 2. In: Proceedings of the 2013 conference of the North American chapter of the association for computational linguistics: human language technologies, Atlanta, Georgia, pp 644–648
Eisele A, Chen Y (2010) MultiUN: A multilingual corpus from United Nation documents. In: Proceedings of the seventh conference on international language resources and evaluation (LREC’10), Valletta, Malta, pp 2868–2872
Federico M, Koehn P, Schwenk H, Trombetti M (2013) Matecat: machine translation enhanced computer assisted translation. In: Proceedings of the XIV machine translation summit, Nice, France, p 425
Gao Q, Vogel S (2008) Parallel implementations of word alignment tool. In: Proceedings of the conference on software engineering, testing, and quality assurance for natural language processing, Columbus, Ohio, pp 49–57
Germann U (2015) Sampling phrase tables for the Moses statistical machine translation system. Prague Bull Math Linguist 104(1):39–50
Gong L, Max A, Yvon F (2014) Incremental development of statistical machine translation systems. In: Proceedings of the 11th international workshop on spoken language translation (IWSLT), South Lake Tahoe, California, Nevada, pp 214–222
González-Rubio J, Ortiz-Martínez D, Casacuberta F (2012) Active learning for interactive machine translation. In: Proceedings of the 13th conference of the European chapter of the association for computational linguistics, Avignon, France, pp 245–254
Green S, Manning CD (2010) Better Arabic parsing: baselines, evaluations, and analysis. In: Proceedings of the 23rd international conference on computational linguistics, Beijing, China, pp 394–402
Hokamp C, Liu Q (2017) Lexically constrained decoding for sequence generation using grid beam search. In: Proceedings of the 55th annual meeting of the association for computational linguistics (volume 1: long papers), Vancouver, Canada, pp 1535–1546. https://doi.org/10.18653/v1/P17-1141
Ive J, Max A, Yvon F, Ravaud P (2016) Diagnosing high-quality statistical machine translation using traces of post-edition operations. In: Proceedings of the LREC 2016 workshop on translation evaluation—from fragmented tools and data sets to an integrated ecosystem, Portorož, Slovenia, pp 55–62
Kay M (1973) The MIND system. In: Rustin R (ed) Natural language processing: courant computer science symposium 8: December 20–21, 1971, Algorithmics Press, pp 155–188
Kay M (1997) The proper place of men and machines in language translation. Mach Transl 12(1):3–23
Kingma DP, Ba J (2014) Adam: A method for stochastic optimization. CoRR abs/1412.6980
Kneser R, Ney H (1995) Improved backing-off for m-gram language modeling. In: Proceedings of the international conference on acoustics, speech, and signal processing (ICASSP’95), Detroit, MI, USA, pp 181–184
Koehn P, Hoang H, Birch A, Callison-Burch C, Federico M, Bertoldi N, Cowan B, Shen W, Moran C, Zens R, Dyer C, Bojar O, Constantin A, Herbst E (2007) Moses: open source toolkit for statistical machine translation. In: Proceedings of the 45th annual meeting of the association for computational linguistics companion volume proceedings of the demo and poster sessions, Prague, Czech Republic, pp 177–180
Koehn P, Carl M, Casacuberta F, Marcos E (2013) CASMACAT: cognitive analysis and statistical methods for advanced computer aided translation. In: Proceedings of the XIV machine translation summit, Nice, France, p 411
Lafferty JD, McCallum A, Pereira FCN (2001) Conditional random fields: probabilistic models for segmenting and labeling sequence data. In: Proceedings of the 18th international conference on machine learning, Morgan Kaufmann Publishers Inc., Williamstown, MA, USA, pp 282–289
Lavergne T, Cappé O, Yvon F (2010) Practical very large scale CRFs. In: Proceedings of the 48th annual meeting of the Association for Computational Linguistics, Uppsala, Sweden, pp 504–513
Levenshtein VI (1966) Binary codes capable of correcting deletions, insertions and reversals. Sov Phys Doklady 10(8):707–710
Logacheva V, Specia L (2015) Phrase-level quality estimation for machine translation. In: Proceedings of the 12th international workshop on spoken language translation (IWSLT), Da Nang, Vietnam, pp 143–150
Logacheva V, Hokamp C, Specia L (2016) MARMOT: a toolkit for translation quality estimation at the word level. In: Proceedings of the tenth international conference on language resources and evaluation (LREC-2016), Portorož, Slovenia, pp 3671–3674
Mathur P, Mauro C, Federico M (2013) Online learning approaches in computer assisted translation. In: Proceedings of the eighth workshop on statistical machine translation, Sofia, Bulgaria, pp 301–308
Mathur P, Cettolo M, Federico M, de Souza JGC (2014) Online multi-user adaptive statistical machine translation. In: Proceedings of the 11th conference of the association for machine translation in the Americas (Volume 1: Research Papers), Vancouver, BC, Canada, pp 152–166
Mikolov T, Chen K, Corrado G, Dean J (2013) Efficient estimation of word representations in vector space. CoRR abs/1301.3781
Mohit B, Hwa R (2007) Localization of difficult-to-translate phrases. In: Proceedings of the second workshop on statistical machine translation, Prague, Czech Republic, pp 248–255
Monroe W, Green S, Manning CD (2014) Word segmentation of informal Arabic with domain adaptation. In: Proceedings of the 52nd annual meeting of the association for computational linguistics (volume 2: short papers), Baltimore, Maryland, pp 206–211. https://doi.org/10.3115/v1/P14-2034
Nelder JA, Mead R (1965) A simplex method for function minimization. Comput J 7(4):308–313
Papineni K, Roukos S, Ward T, Zhu WJ (2002) Bleu: a method for automatic evaluation of machine translation. In: Proceedings of the 40th annual meeting of the association for computational linguistics, Philadelphia, Pennsylvania, USA, pp 311–318
Pedregosa F, Varoquaux G, Gramfort A, Michel V, Thirion B, Grisel O, Blondel M, Prettenhofer P, Weiss R, Dubourg V, Vanderplas J, Passos A, Cournapeau D, Brucher M, Perrot M, Duchesnay E (2011) Scikit-learn: machine learning in Python. J Mach Learn Res 12:2825–2830
Schmid H (1995) Improvements in part-of-speech tagging with an application to German. In: Proceedings of the ACL SIGDAT-workshop, Dublin, Ireland, pp 47–50
Sharoff S, Nivre J (2011) The proper place of men and machines in language technology: processing Russian without any linguistic knowledge. In: Proceedings of the dialogue 2011, Russian conference on computational linguistics, Bekasovo, Russia, pp 657–670
Snover M, Dorr B, Schwartz R, Micciulla L, Makhoul J (2006) A study of translation edit rate with targeted human annotation. In: Proceedings of the 7th conference of the association for machine translation in the Americas, Cambridge, Massachusetts, USA, pp 223–231
Specia L, Shah K, de Souza JG, Cohn T (2013) QuEst – A translation quality estimation framework. In: Proceedings of the 51st annual meeting of the association for computational linguistics: system demonstrations, Sofia, Bulgaria, pp 79–84
Stolcke A (2002) SRILM—an extensible language modeling toolkit. In: Proceedings of the 7th international conference on spoken language processing (ICSLP), Denver, Colorado, pp 901–904
Tiedemann J (2012) Parallel data, tools and interfaces in OPUS. In: Proceedings of the eighth international conference on language resources and evaluation (LREC-2012), Istanbul, Turkey, pp 2214–2218
Tomita M (1985) Feasibility study of personal/interactive machine translation systems. In: Proceedings of the conference on theoretical and methodological issues in machine translation of natural languages, Colgate University, Hamilton, New York, pp 289–297
Toutanova K, Klein D, Manning CD, Singer Y (2003) Feature-rich part-of-speech tagging with a cyclic dependency network. In: Proceedings of the 2003 human language technology conference of the North American Chapter of the Association for Computational Linguistics, Edmonton, Canada, pp 173–180
Way A (2013) Traditional and emerging use-cases for machine translation. In: Proceedings of translating and the computer 35, London, UK
Wuebker J, Green S, DeNero J (2015) Hierarchical incremental adaptation for statistical machine translation. In: Proceedings of the 2015 conference on empirical methods in natural language processing, Lisbon, Portugal, pp 1059–1065
Funding
Funding was provided by ANRT (partially).
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
About this article
Cite this article
Ive, J., Max, A. & Yvon, F. Reassessing the proper place of man and machine in translation: a pre-translation scenario. Machine Translation 32, 279–308 (2018). https://doi.org/10.1007/s10590-018-9223-9
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10590-018-9223-9