LG4AV: Combining Language Models and Graph Neural Networks for Author Verification

  title={LG4AV: Combining Language Models and Graph Neural Networks for Author Verification},
  author={Maximilian Stubbemann and Gerd Stumme},
The automatic verification of document authorships is important in various settings. Researchers are for example judged and compared by the amount and impact of their publications and public figures are confronted by their posts on social media platforms. Therefore, it is important that authorship information in frequently used web services and platforms is correct. The question whether a given document is written by a given author is commonly referred to as authorship verification (AV). While… 

Notebook for PAN at CLEF 2022

A framework using the author’s psychometric, contextual, and ironic features in a Gradient Boosting classifier based on the authors' theory is described, demonstrating the importance of this combination in identifying ironic spreader users.



Explainable Authorship Verification in Social Media via Attention-based Similarity Learning

This work proposes a substantial extension of a recently published hierarchical Siamese neural network approach, with which it is feasible to learn neural features and to visualize the decision-making process and shows that the proposed method is indeed able to latch on to some traditional linguistic categories.

Similarity Learning for Authorship Verification in Social Media

This work proposes a new neural network topology for similarity learning that significantly improves the performance on the author verification task with such challenging data sets.

Authorship verification for different languages, genres and topics

Author Verification Using Common N-Gram Profiles of Text Documents

This work proposes a proximity based method for one-class classification that applies the Common N-Gram (CNG) dissimilarity measure, and utilizes the pairs of most dissimilar documents among documents of known authorship.

Improved algorithms for extrinsic author verification

Two algorithms are proposed, one instance-based and one profile-based (all known documents are treated cumulatively) that are able to outperform state-of-the-art methods in several benchmark datasets and are robust when text length is reduced.

Cross-Domain Authorship Attribution Using Pre-trained Language Models

This paper modify a successful authorship verification approach based on a multi-headed neural network language model and combine it with pre-trained language models and demonstrates the crucial effect of the normalization corpus in cross-domain attribution.

Authorship Verification, Average Similarity Analysis

This work proposes an authorship analysis method that compares the average similarity of a text of unknown authorship with all the text of an author, and introduces a text filtering phase that delete all the sample text of a author that are more similar to the samples of other author.

Authorship Identification using Recurrent Neural Networks

This paper aims to use a deep learning approach for the task of authorship identification by defining a suitable characterization of texts to capture the distinctive style of an author by using an index based word embedding for the C50 and the BBC datasets.

SPECTER: Document-level Representation Learning using Citation-informed Transformers

This work proposes SPECTER, a new method to generate document-level embedding of scientific papers based on pretraining a Transformer language model on a powerful signal of document- level relatedness: the citation graph, and shows that Specter outperforms a variety of competitive baselines on the benchmark.

Author2Vec: Learning Author Representations by Combining Content and Link Information

A novel model, 'Author2Vec', is presented, which learns low-dimensional author representations such that authors who write similar content and share similar network structure are closer in vector space.