The power of language and the language of power in M. T. Anderson’s Feed and Gary Stheyngart’s Super Sad True Love Story. A comparative analysis

The aim of my dissertation is to analyze how selected elements of language are addressed in two contemporary dystopias, Feed by M. T. Anderson (2002) and Super Sad True Love Story by Gary Shteyngart (2010). I chose these two novels because language plays a key role in both of them: both are primarily focused on the pervasiveness of technology, and on how the use/abuse of technology affects language in all its forms. In particular, I examine four key aspects of language: books, literacy, diary writing, as well as oral language. In order to analyze how the aforementioned elements of language are dealt with in Feed and Super Sad True Love Story, I consider how the same aspects of language are presented in a sample of classical dystopias selected as benchmarks: We by Yevgeny Zamyatin (1921), Brave New World by Aldous Huxley (1932), Animal Farm (1945) and Nineteen Eighty-Four (1949) by George Orwell, Fahrenheit 451 by Ray Bradbury (1952), and The Handmaid's Tale by Margaret Atwood (1986). In this way, I look at how language, books, literacy, and diaries are dealt with in Anderson’s Feed and in Shteyngart’s Super Sad True Love Story, both in comparison with the classical dystopias as well as with one another. This allows for an analysis of the similarities, as well as the differences, between the two novels. The comparative analysis carried out also takes into account the fact that the two contemporary dystopias have different target audiences: one is for young adults (Feed), whereas the other is for adults (Super Sad True Love Story). Consequently, I also consider whether further differences related to target readers affect differences in how language is dealt with. Preliminary findings indicate that, despite their different target audiences, the linguistic elements considered are addressed in the two novels in similar ways.

Vocabulary Transfer and Knowledge Distillation for Language Model Compression

Natural Language Processing (NLP) has seen tremendous improvements over the last few years. Transformer architectures achieved impressive results in almost any NLP task, such as Text Classification, Machine Translation, and Language Generation. As time went by, transformers continued to improve thanks to larger corpora and bigger networks, reaching hundreds of billions of parameters. Training and deploying such large models has become prohibitively expensive, such that only big high tech companies can afford to train those models. Therefore, a lot of research has been dedicated to reducing a model’s size. In this thesis, we investigate the effects of Vocabulary Transfer and Knowledge Distillation for compressing large Language Models. The goal is to combine these two methodologies to further compress models without significant loss of performance. In particular, we designed different combination strategies and conducted a series of experiments on different vertical domains (medical, legal, news) and downstream tasks (Text Classification and Named Entity Recognition). Four different methods involving Vocabulary Transfer (VIPI) with and without a Masked Language Modelling (MLM) step and with and without Knowledge Distillation are compared against a baseline that assigns random vectors to new elements of the vocabulary. Results indicate that VIPI effectively transfers information of the original vocabulary and that MLM is beneficial. It is also noted that both vocabulary transfer and knowledge distillation are orthogonal to one another and may be applied jointly. The application of knowledge distillation first before subsequently applying vocabulary transfer is recommended. Finally, model performance due to vocabulary transfer does not always show a consistent trend as the vocabulary size is reduced. Hence, the choice of vocabulary size should be empirically selected by evaluation on the downstream task similar to hyperparameter tuning.

Audiovisual materials and second-language acquisition: a comparison between native language subtitles and target language captions.

Over the last few years, the massive popularity of video streaming platforms has managed to impact our daily habits by making the watching of movies and TV shows one of the main activities of our free time. By providing a wide range of foreign language audiovisual content, these entertainment services may represent a powerful resource for language learners, as they provide them with the possibility to be exposed to authentic input. Moreover, research has shown the beneficial role of audiovisual textual aids such as native language subtitles and target language captions in enhancing language skills such as vocabulary and listening comprehension. The aim of this thesis is to analyze the existing literature on the subject of subtitled and captioned audiovisual materials used as a pedagogical tool for informal language learning.