Leace: Perfect linear concept erasure in closed form
Abstract Concept erasure aims to remove specified features from a representation. It can
improve fairness (eg preventing a classifier from using gender or race) and interpretability …
improve fairness (eg preventing a classifier from using gender or race) and interpretability …
Overview and challenges of machine translation for contextually appropriate translations
P Naveen, P Trojovský - Iscience, 2024 - cell.com
Machine translation facilitates cross-linguistic communication by converting text between
languages. However, producing contextually accurate translations remains a challenge …
languages. However, producing contextually accurate translations remains a challenge …
From rule-based models to deep learning transformers architectures for natural language processing and sign language translation systems: survey, taxonomy and …
With the growing Deaf and Hard of Hearing population worldwide and the persistent
shortage of certified sign language interpreters, there is a pressing need for an efficient …
shortage of certified sign language interpreters, there is a pressing need for an efficient …
Understanding transformer memorization recall through idioms
To produce accurate predictions, language models (LMs) must balance between
generalization and memorization. Yet, little is known about the mechanism by which …
generalization and memorization. Yet, little is known about the mechanism by which …
Translate meanings, not just words: Idiomkb's role in optimizing idiomatic translation with language models
To translate well, machine translation (MT) systems and general-purposed language models
(LMs) need a deep understanding of both source and target languages and cultures …
(LMs) need a deep understanding of both source and target languages and cultures …
Do gpts produce less literal translations?
Large Language Models (LLMs) such as GPT-3 have emerged as general-purpose
language models capable of addressing many natural language generation or …
language models capable of addressing many natural language generation or …
ACES: Translation accuracy challenge sets for evaluating machine translation metrics
As machine translation (MT) metrics improve their correlation with human judgement every
year, it is crucial to understand the limitations of such metrics at the segment level …
year, it is crucial to understand the limitations of such metrics at the segment level …
MIND your language: A multilingual dataset for cross-lingual news recommendation
Digital news platforms use news recommenders as the main instrument to cater to the
individual information needs of readers. Despite an increasingly language-diverse online …
individual information needs of readers. Despite an increasingly language-diverse online …
Semantics of Multiword Expressions in Transformer-Based Models: A Survey
Multiword expressions (MWEs) are composed of multiple words and exhibit variable
degrees of compositionality. As such, their meanings are notoriously difficult to model, and it …
degrees of compositionality. As such, their meanings are notoriously difficult to model, and it …
Neural Generative Models and the Parallel Architecture of Language: A Critical Review and Outlook
According to the parallel architecture, syntactic and semantic information processing are two
separate streams that interact selectively during language comprehension. While …
separate streams that interact selectively during language comprehension. While …