Transformer is a deep neural network that employs a self-attention mechanismto comprehend the contextual relationships within sequential data. Unlikeconventional neural networks or updated versions of Recurrent Neural Networks(RNNs) such as Long Short-Term Memory (LSTM), transformer models excel inhandling long dependencies between input sequence elements and enable parallelprocessing. As a result, transformer-based models have attracted substantialinterest among researchers in the field of artificial intelligence. This can beattributed to their immense potential and remarkable achievements, not only inNatural Language Processing (NLP) tasks but also in a wide range of domains,including computer vision, audio and speech processing, healthcare, and theInternet of Things (IoT). Although several survey papers have been publishedhighlighting the transformer's contributions in specific fields, architecturaldifferences, or performance evaluations, there is still a significant absenceof a comprehensive survey paper encompassing its major applications acrossvarious domains. Therefore, we undertook the task of filling this gap byconducting an extensive survey of proposed transformer models from 2017 to2022. Our survey encompasses the identification of the top five applicationdomains for transformer-based models, namely: NLP, Computer Vision,Multi-Modality, Audio and Speech Processing, and Signal Processing. We analyzethe impact of highly influential transformer-based models in these domains andsubsequently classify them based on their respective tasks using a proposedtaxonomy. Our aim is to shed light on the existing potential and futurepossibilities of transformers for enthusiastic researchers, thus contributingto the broader understanding of this groundbreaking technology.
URL
Affiliations
Abstract
Translation (by gpt-3.5-turbo)
Summary (by gpt-3.5-turbo)