Revisiting Word Prediction: A GCN and ARIMA Framework for Wordle Mining
Qinyu He, Rui Liu, Jinchuan Li · 2023
Word prediction algorithms play an important role in natural language text retrieval and pattern mining.This article explores the hidden patterns of the word mechanism by starting from the word items in Wordle. The paper presents a preliminary ARIMA model based on player guessing data from January 7th, 2022 to December 31st, 2022, aimed at predicting the range of reported results with a 95% confidence level on March 1st, 2023. To extract hidden information about word frequency and guessing difficulty, a GCN model was constructed to establish the distribution of attempts for word structures, where GCN network nodes represent letter frequency and position information, and edges contain weights for letter combinations. The model demonstrated a significant improvement over traditional machine learning algorithms.