• Publications
  • Influence
Step-by-Step: Separating Planning from Realization in Neural Data-to-Text Generation
TLDR
The results demonstrate that decoupling text planning from neural realization indeed improves the system’s reliability and adequacy while maintaining fluent output, and improvements both in BLEU scores and in manual evaluations are observed.
Filling Gender & Number Gaps in Neural Machine Translation with Black-box Context Injection
TLDR
This work proposes a black-box approach for injecting the missing information to a pre-trained neural machine translation system, allowing to control the morphological variations in the generated translations without changing the underlying model or training data.
Improving Quality and Efficiency in Plan-based Neural Data-to-text Generation
TLDR
A trainable neural planning component is introduced that can generate effective plans several orders of magnitude faster than the original planner and a verification-by-reranking stage that substantially improves the faithfulness of the resulting texts is introduced.
Including Signed Languages in Natural Language Processing
TLDR
This position paper calls on the NLP community to include signed languages as a research area with high social and scientific impact and urges the adoption of an efficient tokenization method, development of linguistically-informed models, and the inclusion of local signed language communities as an active and leading voice in the direction of research.
Real-Time Sign Language Detection using Human Pose Estimation
TLDR
A lightweight real-time sign language detection model based on human pose estimation is proposed, which shows improvements of up to 91% accuracy, while still working under 4ms.
Data Augmentation for Sign Language Gloss Translation
TLDR
This work proposes two rule-based heuristics that generate pseudo-parallel glosstext pairs from monolingual spoken language text that improve translation from American Sign Language to English and German Sign Language (DGS) to German by up to 3.14 and 2.20 BLEU, respectively.
ABI Neural Ensemble Model for Gender Prediction Adapt Bar-Ilan Submission for the CLIN29 Shared Task on Gender Prediction
TLDR
The final results suggested that using tokenized, non-lowercased data works best for most of the neural models, while a combination of word clusters, character trigrams and word lists showed to be most beneficial for the majority of the more "traditional" models.
Evaluating the Immediate Applicability of Pose Estimation for Sign Language Recognition
TLDR
This paper analyzes representations based on skeleton poses, as these are explainable, person-independent, privacy-preserving, low-dimensional representations, and characterize the current limitations of skeletal pose estimation approaches in sign language recognition.
At Your Fingertips: Automatic Piano Fingering Detection
TLDR
This work shows that when running a previously proposed model for automatic PIANO-FINGERING on the authors' dataset and then fine-tuning it on manually labeled piano fingering data, it achieves state-of-the-art results.
ABI Neural Ensemble Model for Gender Prediction
TLDR
The system for the CLIN29 shared task on cross-genre gender detection for Dutch suggested that using tokenized, non-lowercased data works best for most of the neural models, while a combination of word clusters, character trigrams and word lists showed to be most beneficial for the majority of the more “traditional” models.