English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Motivated by the seemingly high accuracy levels of machine learning models inMoldavian versus Romanian dialect identification and the increasing researchinterest on this topic, we provide a follow-up on the Moldavian versus RomanianCross-Dialect Topic Identification (MRC) shared task of the VarDial 2019Evaluation Campaign. The shared task included two sub-task types: one thatconsisted in discriminating between the Moldavian and Romanian dialects and onethat consisted in classifying documents by topic across the two dialects ofRomanian. Participants achieved impressive scores, e.g. the top model forMoldavian versus Romanian dialect identification obtained a macro F1 score of0.895. We conduct a subjective evaluation by human annotators, showing thathumans attain much lower accuracy rates compared to machine learning (ML)models. Hence, it remains unclear why the methods proposed by participantsattain such high accuracy rates. Our goal is to understand (i) why the proposedmethods work so well (by visualizing the discriminative features) and (ii) towhat extent these methods can keep their high accuracy levels, e.g. when weshorten the text samples to single sentences or when we use tweets at inferencetime. A secondary goal of our work is to propose an improved ML model usingensemble learning. Our experiments show that ML models can accurately identifythe dialects, even at the sentence level and across different domains (newsarticles versus tweets). We also analyze the most discriminative features ofthe best performing models, providing some explanations behind the decisionstaken by these models. Interestingly, we learn new dialectal patternspreviously unknown to us or to our human annotators. Furthermore, we conductexperiments showing that the machine learning performance on the MRC sharedtask can be improved through an ensemble based on stacking.

The Unreasonable Effectiveness of Machine Learning in Moldavian versus Romanian Dialect Identification