Answers are grounded in this paper and cite their sources.
The article argues that the architectural originality of artificial neural networks (Large Language Models, Recurrent Neural Networks, Long Short-Term Memory, Bidirectional Encoder Representations from Transformers) opens up new possibilities for recognizing linguistic markers: first of all, it is about innovations in the structure of the latter.It is emphasized that they will lead to more accurate and efficient analysis of textual data, more valid output results, etc. Accordingly, the development of artificial neural networks correlates with progress in modern humanities research (in particular, digital humanities), producing the emergence of new, innovative tools and modern methodology for discourse analysis of textual data.The author notes that the main architectural innovations that have significantly increased the efficiency of artificial neural networks with linguistic markers are: a) attention mechanisms that produce a focus on the most relevant parts of the input data, improving the accuracy, efficiency, and other processing of the data; b) pre-training, which is effective for models such as Bidirectional Encoder Representations from Transformers, which use it to learn general language patterns before additional training on specific tasks; c) in deep learning architectures, the inclusion of many hidden layers leads to a greater degree of model ability to detect complex linguistic markers.Thus, innovations in the architecture of artificial neural networks produce new possibilities for processing linguistic markers, causing transformational changes in modern linguistic research (in particular, those written within the digital humanities).The ChatGPT artificial neural network, whose architecture is specific (a combination of Recurrent Neural Networks and Bidirectional Encoder Representations from Transformers), trained on huge amounts of text data, is illustrative in this regard.For linguists, such a model may be of interest in two ways: the first is the emergence of the phenomenon of "meaningfulness" in the texts it generates, and the second is related to its architecture and functionality.In particular, we consider the following structural features to be indicative: a) an exponential increase in the size of the model, b) actualization of the role of context (discourse), and c) a significant improvement in architectural capabilities, which manifest themselves in the potential emergence of hybrid models.The researcher concludes that the possibilities for the development of artificial neural networks are enormous, as humanity is only beginning to understand the range of their applications, and as for linguistics, the development of such models will lead to more accurate generation of textual data, etc. against the background of correlation with the search and development of new approaches to applied integrated tasks (Natural Language Processing, Sentiment Analysis, etc.).In turn, this will lead to a specific rethinking of the role of science (in particular, modern linguistic research), as its essence will become more applied and correlated with the development of artificial intelligence, becoming a specific basis for the latter.
No takes yet. Share an insight, caveat, or question.
Oleksii Dovhan (2024) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: