Search Results for author: Blaž Škrlj

Found 33 papers, 19 papers with code

E8-IJS@LT-EDI-ACL2022 - BERT, AutoML and Knowledge-graph backed Detection of Depression

no code implementations LTEDI (ACL) 2022 Ilija Tavchioski, Boshko Koloski, Blaž Škrlj, Senja Pollak

Depression is a mental illness that negatively affects a person’s well-being and can, if left untreated, lead to serious consequences such as suicide.

AutoML

BERT meets Shapley: Extending SHAP Explanations to Transformer-based Classifiers

no code implementations EACL (Hackashop) 2021 Enja Kokalj, Blaž Škrlj, Nada Lavrač, Senja Pollak, Marko Robnik-Šikonja

Transformer-based neural networks offer very good classification performance across a wide range of domains, but do not provide explanations of their predictions.

Exploring Neural Language Models via Analysis of Local and Global Self-Attention Spaces

1 code implementation EACL (Hackashop) 2021 Blaž Škrlj, Shane Sheehan, Nika Eržen, Marko Robnik-Šikonja, Saturnino Luz, Senja Pollak

Large pretrained language models using the transformer neural network architecture are becoming a dominant methodology for many natural language processing tasks, such as question answering, text classification, word sense disambiguation, text completion and machine translation.

Machine Translation Natural Language Processing +5

Interesting cross-border news discovery using cross-lingual article linking and document similarity

no code implementations EACL (Hackashop) 2021 Boshko Koloski, Elaine Zosa, Timen Stepišnik-Perdih, Blaž Škrlj, Tarmo Paju, Senja Pollak

Team Name: team-8 Embeddia Tool: Cross-Lingual Document Retrieval Zosa et al. Dataset: Estonian and Latvian news datasets abstract: Contemporary news media face increasing amounts of available data that can be of use when prioritizing, selecting and discovering new news.

Retrieval-efficiency trade-off of Unsupervised Keyword Extraction

1 code implementation15 Aug 2022 Blaž Škrlj, Boshko Koloski, Senja Pollak

Efficiently identifying keyphrases that represent a given document is a challenging task.

Keyword Extraction

Out of Thin Air: Is Zero-Shot Cross-Lingual Keyword Detection Better Than Unsupervised?

no code implementations14 Feb 2022 Boshko Koloski, Senja Pollak, Blaž Škrlj, Matej Martinc

We find that the pretrained models fine-tuned on a multilingual corpus covering languages that do not appear in the test set (i. e. in a zero-shot setting), consistently outscore unsupervised models in all six languages.

Keyword Extraction Pretrained Multilingual Language Models

Unsupervised Feature Ranking via Attribute Networks

1 code implementation25 Nov 2021 Urh Primožič, Blaž Škrlj, Sašo Džeroski, Matej Petković

The need for learning from unlabeled data is increasing in contemporary machine learning.

Recommendation Systems

Link Analysis meets Ontologies: Are Embeddings the Answer?

1 code implementation23 Nov 2021 Sebastian Mežnar, Matej Bevec, Nada Lavrač, Blaž Škrlj

The increasing amounts of semantic resources offer valuable storage of human knowledge; however, the probability of wrong entries increases with the increased size.

Anomaly Detection

Knowledge Graph informed Fake News Classification via Heterogeneous Representation Ensembles

2 code implementations20 Oct 2021 Boshko Koloski, Timen Stepišnik-Perdih, Marko Robnik-Šikonja, Senja Pollak, Blaž Škrlj

Increasing amounts of freely available data both in textual and relational form offers exploration of richer document representations, potentially improving the model performance and robustness.

Classification Fake News Detection +4

Compressibility of Distributed Document Representations

no code implementations14 Oct 2021 Blaž Škrlj, Matej Petkovič

Contemporary natural language processing (NLP) revolves around learning from latent document representations, generated either implicitly by neural language models or explicitly by methods such as doc2vec or similar.

Natural Language Processing Text Classification

Semantic Reasoning from Model-Agnostic Explanations

no code implementations29 Jun 2021 Timen Stepišnik Perdih, Nada Lavrač, Blaž Škrlj

The derived semantic explanations are potentially more informative, as they describe the key attributes in the context of more general background knowledge, e. g., at the biological process level.

Transfer Learning for Node Regression Applied to Spreading Prediction

1 code implementation31 Mar 2021 Sebastian Mežnar, Nada Lavrač, Blaž Škrlj

This work is one of the first to explore transferability of the learned representations for the task of node regression; we show there exist pairs of networks with similar structure between which the trained models can be transferred (zero-shot), and demonstrate their competitive performance.

Misinformation Transfer Learning

Extending Neural Keyword Extraction with TF-IDF tagset matching

1 code implementation EACL (Hackashop) 2021 Boshko Koloski, Senja Pollak, Blaž Škrlj, Matej Martinc

Keyword extraction is the task of identifying words (or multi-word expressions) that best describe a given document and serve in news portals to link articles of similar topics.

Keyword Extraction

ReliefE: Feature Ranking in High-dimensional Spaces via Manifold Embeddings

1 code implementation23 Jan 2021 Blaž Škrlj, Sašo Džeroski, Nada Lavrač, Matej Petković

The utility of ReliefE for high-dimensional data sets is ensured by its implementation that utilizes sparse matrix algebraic operations.

Multi-Label Classification

Identification of COVID-19 related Fake News via Neural Stacking

no code implementations11 Jan 2021 Boshko Koloski, Timen Stepišnik Perdih, Senja Pollak, Blaž Škrlj

Identification of Fake News plays a prominent role in the ongoing pandemic, impacting multiple aspects of day-to-day life.

Fake News Detection General Classification

Predicting Generalization in Deep Learning via Metric Learning -- PGDL Shared task

no code implementations16 Dec 2020 Sebastian Mežnar, Blaž Škrlj

The competition "Predicting Generalization in Deep Learning (PGDL)" aims to provide a platform for rigorous study of generalization of deep learning models and offer insight into the progress of understanding and explaining these models.

Metric Learning

Ensemble- and Distance-Based Feature Ranking for Unsupervised Learning

1 code implementation23 Nov 2020 Matej Petković, Dragi Kocev, Blaž Škrlj, Sašo Džeroski

In this work, we propose two novel (groups of) methods for unsupervised feature ranking and selection.

SNoRe: Scalable Unsupervised Learning of Symbolic Node Representations

1 code implementation8 Sep 2020 Sebastian Mežnar, Nada Lavrač, Blaž Škrlj

Learning from complex real-life networks is a lively research area, with recent advances in learning information-rich, low-dimensional network node representations.

Node Classification Structural Node Embedding

Fuzzy Jaccard Index: A robust comparison of ordered lists

2 code implementations5 Aug 2020 Matej Petković, Blaž Škrlj, Dragi Kocev, Nikola Simidjievski

In real-life, and in particular high-dimensional domains, where only a small percentage of the whole feature space might be relevant, a robust and confident feature ranking leads to interpretable findings as well as efficient computation and good predictive performance.

BIG-bench Machine Learning

Propositionalization and Embeddings: Two Sides of the Same Coin

2 code implementations8 Jun 2020 Nada Lavrač, Blaž Škrlj, Marko Robnik-Šikonja

This paper outlines some of the modern data processing techniques used in relational learning that enable data fusion from different input data types and formats into a single table data representation, focusing on the propositionalization and embedding data transformation approaches.

Relational Reasoning

AttViz: Online exploration of self-attention for transparent neural language modeling

1 code implementation12 May 2020 Blaž Škrlj, Nika Eržen, Shane Sheehan, Saturnino Luz, Marko Robnik-Šikonja, Senja Pollak

Neural language models are becoming the prevailing methodology for the tasks of query answering, text classification, disambiguation, completion and translation.

Language Modelling Text Classification +1

TNT-KID: Transformer-based Neural Tagger for Keyword Identification

1 code implementation20 Mar 2020 Matej Martinc, Blaž Škrlj, Senja Pollak

With growing amounts of available textual data, development of algorithms capable of automatic analysis, categorization and summarization of these data has become a necessity.

Keyword Extraction Language Modelling

Feature Importance Estimation with Self-Attention Networks

no code implementations11 Feb 2020 Blaž Škrlj, Sašo Džeroski, Nada Lavrač, Matej Petkovič

Black-box neural network models are widely used in industry and science, yet are hard to understand and interpret.

Feature Importance

Embedding-based Silhouette Community Detection

1 code implementation17 Jul 2019 Blaž Škrlj, Jan Kralj, Nada Lavrač

Mining complex data in the form of networks is of increasing interest in many scientific disciplines.

Community Detection

Language comparison via network topology

1 code implementation16 Jul 2019 Blaž Škrlj, Senja Pollak

In our experiments, we employ eight different network topology metrics, and empirically showcase on a parallel corpus, how the methods can be used for modeling the relations between nine selected languages.

Natural Language Processing

RaKUn: Rank-based Keyword extraction via Unsupervised learning and Meta vertex aggregation

1 code implementation15 Jul 2019 Blaž Škrlj, Andraž Repar, Senja Pollak

Keyword extraction is used for summarizing the content of a document and supports efficient document retrieval, and is as such an indispensable part of modern text-based systems.

Keyword Extraction

Cannot find the paper you are looking for? You can Submit a new open access paper.