Unsupervised Pairwise Entity Classification Model to Detect Free Associations
Alex Romanova · 2022 International Conference on Electrical, Computer and Energy Technologies (ICECET) · 2022
Outstanding success of Convolutional Neural Network (CNN) image classification influenced application of these techniques to many data mining domains. Highly accurate CNN image classifications that are based on supervised machine learning require labeling of huge volumes of input data. One of the ways to resolve this problem is unsupervised pairwise entity classification model that classifies entity pairs to symmetric and asymmetric classes. In this study we will examine a possibility to train this model on one data domain and apply it to classifying pairwise entities from another data domain. We will demonstrate how to use pairwise entity classification model trained on time series climate data to detect ’free associations’, i.e. how to find in text documents semantically dissimilar co-located word pairs.