FSLM: An Intelligent Few-Shot Learning Model Based on Siamese Networks for IoT Technology

Abstract

As an important application of the Internet of Things (IoT) devices, sentiment analysis has been paid more attention with the rapid development of artificial intelligence. As a widely used method in artificial intelligence applications, traditional deep learning methods need massive data for training. However, due to the limitations of hardware, IoT devices have deficiencies in processing big data. In the case of insufficient sample size, how to carry out a machine learning method for IoT devices has become a common concern of the industry. In order to perform sentiment analysis on text with few data samples from the IoT devices, we propose FSLM, which is an intelligent few-shot learning model based on Siamese networks. The FSLM model consists of two self-attention models with the same parameters, which are divided into two parts. First, for two input texts, a self-attention model is used to extract sentiment features, and then the Mahalanobis distance is adopted to measure the similarity between two feature vectors to determine whether they belong to the same category. The FSLM is tested on the Amazon Review Sentiment Classification (ARSC) data set. The extensive experimental results on this data set demonstrate that the FSLM model has better accuracy and robustness for text sentiment analysis than other main existing models with a small number of samples.