Search

As large Pre-trained Language Models (PLMs) trained on large amounts of data in an unsupervised manner become more ubiquitous, identifying various types of bias in the text has come into sharp focus. Existing "Stereotype Detection" datasets mainly adopt a diagnostic approach toward large PLMs. Blodgett et. al (2021a) show that there are significant reliability issues with the existing benchmark datasets. Annotating a reliable dataset requires a precise understanding of the subtle nuances of how stereotypes manifest in text. In this paper, we annotate a focused evaluation set for "Stereotype Detection" that addresses those pitfalls by de-constructing various ways in which stereotypes manifest in text. Further, we present a multi-task model that leverages the abundance of data-rich neighboring tasks such as hate speech detection, offensive language detection, misogyny detection, etc., to improve the empirical performance on "Stereotype Detection". We then propose a reinforcement-learning agent that guides the multi-task learning model by learning to identify the training examples from the neighboring tasks that help the target task the most. We show that the proposed models achieve significant empirical gains over existing baselines on all the tasks., Long paper at ACL 2022 main conference

Additional file 1.

Mestrado Vinifera EuroMaster - Instituto Superior de Agronomia Ageing over lees has long been considered to benefit the overall quality of wine, enhancing the body and mouthful as well as sensorial complexity, color stability. Despite of all the positive attributes conferred by this technique, it is a complex process which could last up to years and is affected by several variables of different nature and complexities. This process has several economic impacts representing large investments for all producers to store the wines in cellars as well as the wine maker has to bear the potential risk associated with the organoleptic and microbiological alterations in the wines. Thus, it is important for the winemakers to optimize the time of ageing on lees. Further more in today’s fiercely competitive market it is reasonable to develop new strategies and techniques to accelerate the ageing over lees process, shorten the storage time and achieve better quality. In this study two novel techniques: use of non Saccharomyces strains coupled with ultrasound treatment were tested to see their efficiency for accelerating aging over lees. The combined effects of the techniques were tested to see their impacts on the polysaccharide release and on the organoleptic properties of red wine. Release of polysaccharides was analyzed by HPLC-RI. Anthocyanins and aroma compounds were analyzed by using HPLC-PDAD/ESI-MS, GC-FID respectively. Also Color and Total Phenolic Index were recorded periodically along the experiment. Results showed that ultrasound treatment is a reliable technique for shortening the ageing on lees process by strongly increasing the concentration of polysaccharides released into the wine after only two weeks treatment and without adversely affecting the sensorial quality of the wine. Additions of sand as an abrasives agent increased the polysaccharide release. Furthermore the non Saccharomyces strains of Schizosaccharomyces pombe, Saccharomyces ludwigii and Brettanomyces showed better results with regards to the amount of polysaccharide release compared to the control Saccharomyces strains. Interesting results with Brettanomyces were observed, as the use of this particular strain did not impart any off flavors to the wine. This can be explained by the use of lyophilized strains used in this study which were dosed in the wine to carry out the aging over lees. Ultrasonic treatment coupled with aging over lees resulted in reduction of anthocyanin content of the wine and also effected the aroma compounds.In conclusion this study illustrated the use of Ultrasounds and Non Saccharomyces strains as novel techniques for aging over lees, however more research in this field is required to study the clear effects on ultrasounds on the chemical composition of wine before replicating the application on a large scale production