Acquiring Semantic Mode Signal from Tweets of Cryptoassets
Hiroshi Uehara, Wataru Souma, Yuichi Ikeda · 2023
This study proposes a method for detecting collective motions, the time series situation where dispersed information becomes inclined to a unique direction.Our proposal, Semantic mode signal, is distinctive among related methods in providing the situation with semantic contexts extracted from time series texts.Furthermore, the proposal is characterized by its applicability to high-dimensional word space with sparsity, such as numerous tweets.We applied the method to tweets concerning 19 cryptoassets.The empirical results indicated that the method appropriately detected the collective motions representing the contexts semantically coincident with the news events and the price trends, supporting the efficiency of our proposal.