Discrete Quantum Random Walks for Semantic Text Similarity

Context-based determination of text similarity is a fundamental computation task that enables identification and attribution of emerging topics and narratives within and across information platforms. Most existing techniques for text similarity scoring are not context-based, or are context-based but struggle to process large amounts of data. In our work, we have developed an end-to-end quantum random walk protocol that determines semantic text similarity. As part of our work on the quantum protocol, we also developed a novel regularization protocol for irregular graphs, and a clean implementation of a weighted and directed graph decomposition protocol, all of which may provide speedup in context-based analysis of texts.