Automatic Synonym and Phrase Replacement Show Promise for Style Transformation

Khosmood, Foaad; Levinson, Robert

doi:10.1109/icmla.2010.153

Cited by 12 publications

(11 citation statements)

References 2 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Back-and-forth machine translation provides a simple but highly limited technique [3,7,10,14,44], as it does not allow targeting or avoiding any particular style. Another classical alternative is rule-based paraphrase replacement from knowledge bases [12, 36-40, 45, 61], 1 https://gitlab.com/ssg-research/mlsec/parchoice/ Techniques Transformations applied [40,46] synonym replacement from WordNet [45,61] word embedding neighbour replacement [38] word replacement from GNU Diction [26], hand-crafted rules [12] synonym replacement from FreeLing [54], hand-crafted rules [36] synonym/hypernym/definition replacement from WordNet or PPDB, hand-crafted rules Table 1. Prior rule-based style transfer techniques.…”

Section: Introductionmentioning

confidence: 99%

Effective writing style transfer via combinatorial paraphrasing

Gröndahl

Asokan

2020

Proceedings on Privacy Enhancing Technologies

View full text Add to dashboard Cite

Stylometry can be used to profile or deanonymize authors against their will based on writing style. Style transfer provides a defence. Current techniques typically use either encoder-decoder architectures or rule-based algorithms. Crucially, style transfer must reliably retain original semantic content to be actually deployable. We conduct a multifaceted evaluation of three state-of-the-art encoder-decoder style transfer techniques, and show that all fail at semantic retainment. In particular, they do not produce appropriate paraphrases, but only retain original content in the trivial case of exactly reproducing the text. To mitigate this problem we propose ParChoice: a technique based on the combinatorial application of multiple paraphrasing algorithms. ParChoice strongly outperforms the encoder-decoder baselines in semantic retainment. Additionally, compared to baselines that achieve nonnegligible semantic retainment, ParChoice has superior style transfer performance. We also apply ParChoice to multi-author style imitation (not considered by prior work), where we achieve up to 75% imitation success among five authors. Furthermore, when compared to two state-of-the-art rule-based style transfer techniques, ParChoice has markedly better semantic retainment. Combining ParChoice with the best performing rulebased baseline (Mutant-X [34]) also reaches the highest style transfer success on the Brennan-Greenstadt and Extended-Brennan-Greenstadt corpora, with much less impact on original meaning than when using the rulebased baseline techniques alone. Finally, we highlight a critical problem that afflicts all current style transfer techniques: the adversary can use the same technique for thwarting style transfer via adversarial training. We show that adding randomness to style transfer helps to mitigate the effectiveness of adversarial training.

show abstract

Section: Introductionmentioning

confidence: 99%

Effective writing style transfer via combinatorial paraphrasing

Gröndahl

Asokan

2020

Proceedings on Privacy Enhancing Technologies

View full text Add to dashboard Cite

show abstract

“…is an attempt to give simple mathematical representations to complex properties of language, and social norms and interactions in a group which will in turn facilitate automation in feature extraction. Features similar to these were also used in a recent work by Khosmood which attempted to disguise authorship through camouflage [16].…”

Section: Feature Extractionmentioning

confidence: 94%

Implicit Group Membership Detection in Online Text: Analysis and Applications

Ellen

Kaina

Parameswaran

2012

Social Computing, Behavioral - Cultural Modeling and Prediction

View full text Add to dashboard Cite

Abstract. Our thesis is that members of the same group have shared tendencies and nuances in communication style and substance, particularly online. In this paper, we dicuss some potential applications of accuarate authorship affiliation technology. We also discuss related work in similar author identification efforts and the research issues that currently exist when trying to perform automated authorship affiliation. We provide quantitative results from our recent Machine Learning experimenation using Support Vector Machines as some initial validation of our theory. In this paper, we applied our work towards the task of classifying website forum posts by the affiliation of their author. We discuss in detail the stylometric features we used to perform the automated classification and split the original features into individual groups to isolate their respective contributions and/or discriminating capability. Our results show promise towards automating group representation, an important first step in studying group formation.

show abstract

“…There are other works [8,9] on automatic replacement and style transformation. However, these works either provide a general approach or are not acceptable in practice.…”

Section: Related Workmentioning

confidence: 99%

“…In [8], it is suggested to transform the writing style of a document incrementally using a loop, where in each run of the loop, the style is slightly changed and this is repeated until some target condition is satisfied. An example of this latter category is [9] where all the words in a document are automatically replaced with their synonyms. This may decrease the readability, and could significantly affect the semantic of the document.…”

Section: Related Workmentioning

confidence: 99%

“…In the case of Anonymouth (an instantiation of [6]) which is not completely deterministic, using the data set and experiments that are reported in the paper, we can identify a set of size 2 that includes the author, with probability 14 32 = 0.438. In this experiment, in total 10 authors were considered and so random guessing of a set of size 2 that includes the author would have the success chance of 9 ( 10 2 ) = 0.2. Hence, our attack doubles the success chance of the attacker compared to this random guess.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Secure Obfuscation of Authoring Style

Galib

2015

Information Security Theory and Practice

View full text Add to dashboard Cite

Anonymous authoring includes writing reviews, comments and blogs, using pseudonyms with the general assumption that using these pseudonyms will protect the real identity of authors and allows them to freely express their views. It has been shown, however, that writing style may be used to trace authors across multiple Websites. This is a serious threat to privacy and may even result in revealing the authors's identities. In obfuscating authors' writing style, an authored document is modified to hide the writing characteristics of the author. In this paper we first show that existing obfuscation systems are insecure and propose a general approach for constructing obfuscation algorithms, and then instantiate the framework to give an algorithm that semi-automatically modifies an author's document. We provide a secure obfuscation scheme that is able to hide an author's document securely among other authors' documents in a corpus. As part of our obfuscation algorithm we present a new algorithm for identifying an author's unique words that would be of independent interest. We present a security model and use it to analyze our scheme and also the previous schemes. We implement our scheme and give its performances through experiments. We show that our algorithm can be used to obfuscate documents securely and effectively.

show abstract

Automatic Synonym and Phrase Replacement Show Promise for Style Transformation

Cited by 12 publications

References 2 publications

Effective writing style transfer via combinatorial paraphrasing

Effective writing style transfer via combinatorial paraphrasing

Implicit Group Membership Detection in Online Text: Analysis and Applications

Secure Obfuscation of Authoring Style

Contact Info

Product

Resources

About