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

Blaž Škrlj, Shane Sheehan, Nika Eržen, Marko Robnik-Šikonja, Saturnino Luz, Senja Pollak

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract / Description of output

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. Commonly comprising hundreds of millions of parameters, these models offer state-of-the-art performance, but at the expense of interpretability. The attention mechanism is the main component of transformer networks. We present AttViz, a method for exploration of self-attention in transformer networks, which can help in explanation and debugging of the trained models by showing associations between text tokens in an input sequence. We show that existing deep learning pipelines can be explored with AttViz, which offers novel visualizations of the attention heads and their aggregations. We implemented the proposed methods in an online toolkit and an offline library. Using examples from news analysis, we demonstrate how AttViz can be used to inspect and potentially better understand what a model has learned.

Original languageEnglish
Title of host publicationEACL Hackashop on News Media Content Analysis and Automated Report Generation, Hackashop 2021 at 16th conference of the European Chapter of the Association for Computational Linguistics, EACL 2021 - Proceedings
EditorsHannu Toivonen, Michele Boggia
PublisherAssociation for Computational Linguistics (ACL)
Pages76-83
Number of pages8
ISBN (Electronic)9781954085138
Publication statusPublished - 16 Apr 2021
Event2021 EACL Hackashop on News Media Content Analysis and Automated Report Generation, Hackashop 2021 - Virtual, Online
Duration: 19 Apr 2021 → …

Publication series

NameEACL Hackashop on News Media Content Analysis and Automated Report Generation, Hackashop 2021 at 16th conference of the European Chapter of the Association for Computational Linguistics, EACL 2021 - Proceedings

Conference

Conference2021 EACL Hackashop on News Media Content Analysis and Automated Report Generation, Hackashop 2021
CityVirtual, Online
Period19/04/21 → …

Fingerprint

Dive into the research topics of 'Exploring Neural Language Models via Analysis of Local and Global Self-Attention Spaces'. Together they form a unique fingerprint.

Cite this