arXiv subjects

Serge Sharoff

Publications and source records attributed to Serge Sharoff.

21 records · Page 2Linked to original sources

arXiv2020

Sentence Level Human Translation Quality Estimation with Attention-based Neural Networks

This paper explores the use of Deep Learning methods for automatic estimation of quality of human translations. Automatic estimation can provide useful feedback for translation teaching, examination and quality control. Conventional methods for solving this task rely on manually engineered features and external knowledge. This paper presents an end-to-end neural model without feature engineering, incorporating a cross attention mechanism to detect which parts in sentence pairs are most relevant for assessing quality. Another contribution concerns of prediction of fine-grained scores for measuring different aspects of translation quality. Empirical results on a large human annotated dataset show that the neural model outperforms feature-based methods significantly. The dataset and the tools are available.

cs.CL↗

arXiv2020

Know thy corpus! Robust methods for digital curation of Web corpora

This paper proposes a novel framework for digital curation of Web corpora in order to provide robust estimation of their parameters, such as their composition and the lexicon. In recent years language models pre-trained on large corpora emerged as clear winners in numerous NLP tasks, but no proper analysis of the corpora which led to their success has been conducted. The paper presents a procedure for robust frequency estimation, which helps in establishing the core lexicon for a given corpus, as well as a procedure for estimating the corpus composition via unsupervised topic models and via supervised genre classification of Web pages. The results of the digital curation study applied to several Web-derived corpora demonstrate their considerable differences. First, this concerns different frequency bursts which impact the core lexicon obtained from each corpus. Second, this concerns the kinds of texts they contain. For example, OpenWebText contains considerably more topical news and political argumentation in comparison to ukWac or Wikipedia. The tools and the results of analysis have been released.

cs.CL↗

arXiv2003

Core Lexicon and Contagious Words

We present the new empirical parameter $f_c$, the most probable usage frequency of a word in a language, computed via the distribution of documents over frequency $x$ of the word. This parameter allows for filtering the core lexicon of a language from the content words, which tend to be extremely frequent in some texts written in specific genres or by certain authors. Distributions of documents over frequencies for such words display long tails as $x>f_c$ representing a bunch of documents in which such words are used in abundance. Collections of such documents exhibit a percolation like phase transition as the coarse grain of frequency $Δf$ (flattening out the strongly irregular frequency data series) approaches the critical value $f_c$.

cond-mat.dis-nn↗