Personalized Machine Translation: Preserving Original Author Traits

By Shuly Wintner, Shachar Mirkin, Lucia Specia, Ella Rabinovich, Raj Nath Patel. 2016

125 citations recorded by the source corpus.

The language that we produce reflects our personality, and various personal and demographic characteristics can be detected in natural language texts. We focus on one particular personal trait of the author, gender, and study how it is manifested in original texts and in translations. We show that author’s gender has a powerful, clear signal in originals texts, but this signal is obfuscated in human and machine translation. We then propose simple domain-adaptation techniques that help retain the original gender traits in the translation, without harming the quality of the translation, thereby creating more personalized machine translation systems.

Subfields: Artificial Intelligence, Molecular Biology

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Molecular Biology

Citations in the corpus

Getting Gender Right in Neural Machine Translation (2019, methodology): “When translating from one language into another, original author traits are partially lost, both in human and machine translations (Mirkin et al., 2015; Rabinovich et al., 2017).”
Style Transfer Through Back-Translation (2018, background): “sentence from a source language to a target language retains the meaning of the sentence but does not preserve the stylistic features related to the author’s traits (Rabinovich et al., 2016).”
Automatic Gender Identification and Reinflection in Arabic (2019, background): “The approach is system-independent and can be used with MT, dialogue systems, etc., as well as, to balance corpora through augmentation by adding reinﬂected copies of gender-speciﬁc constructions.”
Extreme Adaptation for Personalized Neural Machine Translation (2018, methodology): “Recent work has incorporated side information about the author such as personality (Mirkin et al., 2015), gender (Rabinovich et al., 2017) or polite-ness (Sennrich et al., 2016a), but these methods can only handle phenomena where there are ex-Source plicit labels for the traits.”
Towards Cross-Lingual Generalization of Translation Gender Bias (2021, background): “its identification [17, 22, 26] and resolution [10, 23] have been dis-”
Gender-Aware Reinflection using Linguistically Enhanced Neural Models (2020, background): “…etc.) biases across a variety of core tasks such as coreference resolution (Rudinger et al., 2018; Zhao et al., 2018a), machine translation (Rabinovich et al., 2017; Vanmassenhove et al., 2018; Font and Costa-jussà, 2019; Moryossef et al., 2019; Stanovsky et al., 2019; Stafanoviˇcs et al.,…”
Gender Bias in Machine Translation (2021, unknown): “, 2019) and personalized MT models (Mirkin et al., 2015; Bawden et al., 2016; Rabinovich et al., 2017).”
Filling Gender & Number Gaps in Neural Machine Translation with Black-box Context Injection (2019, background): “Rabinovich et al. (2017) showed that given input with author traits like gender, it is possible to retain those traits in Statistical Machine Translation (SMT) models.”
Towards Robust and Privacy-preserving Text Representations (2018, background): “Moreover, models ﬁt to language corpora often ﬁxate on author attributes which correlate with the target variable, e.g., gender correlating with class skews (Zhao et al., 2017), or translation choices (Rabinovich et al., 2017).”
Europarl Datasets with Demographic Speaker Information (2018, methodology): “We used meta-information of the members of the European Parliament (MEPs) released by Rabinovich et al. (2017) to retrieve the demographic annotations.”
Fair Transfer of Multiple Style Attributes in Text (2019, background): “Further [10] argued on using author’s characteristics as pointed out by [11] who argued that in a machine translation system, author’s characteristics are very well obfuscated.”
Does It Capture STEL? A Modular, Similarity-based Linguistic Style Evaluation Framework (2021, background): “Additionally, style can be important to address in natural language generation (Ficler and Goldberg, 2017), including identity modeling in dialogue systems (Li et al., 2016) and style preservation in machine translation (Niu et al., 2017; Rabinovich et al., 2017).”
Revision in Continuous Space: Unsupervised Text Style Transfer without Adversarial Learning (2019, methodology): “sentangle the content and the attribute with adversarial learning, another line of research (Prabhumoye et al. 2018; Logeswaran, Lee, and Bengio 2018; Guillaume Lample 2019) applies back-translation (Wintner et al. 2016) to rephrase a sentence while reducing the stylistic properties and encourage co”
Case Study: Deontological Ethics in NLP (2020, background): “systems (Rabinovich et al., 2017; Michel and Neubig, 2018; Mirkin et al., 2015; Mirkin and Meunier, 2015) while preserving author attributes as well as controlling structural information like voice (Yamagishi et al.”
Predicting Authorship and Author Traits from Keystroke Dynamics (2018, background): “, 2018) or can be used for a wide range of applications like customer support, healthcare and personalized machine translation (Mirkin et al., 2015; Rabinovich et al., 2017).”
A Survey on Data Augmentation for Text Classification (2021, unknown): “By the translation of text, the content is preserved and only stylistic features based on the traits of the author are excluded or altered [95].”