چکیده مقاله
The main problem in this survey paper that we want to speak about it is Author Masking, besides, we willtalk about Author Verification too In the author verification, given a pair of documents, decide whetherthey are written by the same author or not They propose a machine learning attitude based on a numberof different features that characterize documents from widely different points of view In the authormasking, given a document and a set of documents from the same author, changing the former so that itsauthor cannot be identified, anymore Masking the writing style of a writer has been useful and used bynovelists for the goal of passing unnoticed, as well as by people who aim to give information withoutbeing linked to it In fact, these two approaches are in front of together and by studying on authormasking methods we can detect the weaknesses of authorship verification system and simulate the systemthat will be able to find the text s author easily In this paper, we will state an approach that constructsnon overlapping groups of homogeneous features, use a random forest regressor for each features groupand combine the output of all regressors by their arithmetic mean to verify the text s author and someapproaches that are in against of that and try to mask the text s author by simple translation,transformation in sentences, word choice distribution, sentence length preference, and etc The main problem in this survey paper that we want to speak about it is Author Masking, besides, we willtalk about Author Verification too In the author verification, given a pair of documents, decide whetherthey are written by the same author or not They propose a machine learning attitude based on a numberof different features that characterize documents from widely different points of view In the authormasking, given a document and a set of documents from the same author, changing the former so that itsauthor cannot be identified, anymore Masking the writing style of a writer has been useful and used bynovelists for the goal of passing unnoticed, as well as by people who aim to give information withoutbeing linked to it In fact, these two approaches are in front of together and by studying on authormasking methods we can detect the weaknesses of authorship verification system and simulate the systemthat will be able to find the text s author easily In this paper, we will state an approach that constructsnon overlapping groups of homogeneous features, use a random forest regressor for each features groupand combine the output of all regressors by their arithmetic mean to verify the text s author and someapproaches that are in against of that and try to mask the text s author by simple translation,transformation in sentences, word choice distribution, sentence length preference, and etc
کلیدواژهها
نویسندگان
Farnaz Mahan
Assistant Professor of Tabriz University, University of Tabriz, Computer Science Department, Tabriz, Iran
Salar Khayat Mirza Rasluli
Master student of soft computing and artificial intelligence, University of Tabriz, Computer Science Department, Tabriz, Iran
Erfan Mohammadzadeh
Bachelor student of computer science, Computer Science Department, Ardabil, Iran
شیوه ارجاع
Mahan, Farnaz and Khayat Mirza Rasluli, Salar and Mohammadzadeh, Erfan,1397,Author Masking against Author Verification and a means to strengthen it using different features,Second National Conference on Computer, Information Technology and AI Applications,Ahvaz
ارائهشده در
مجموعه مقالات دومین کنفرانس ملی کامپیوتر، فناوری اطلاعات و کاربردهای هوش مصنوعی1 اسفند 1397 · اهواز