Physiological time‑series such as ECG and EEG show intricate temporal patterns, high acquisition variability, and a strong demand for transparent decisions. Deep models achieve strong detection scores but often cannot explain why a segment is anomalous, how local anomalies evolve over time, or whether they belong to a larger recurring pattern. Signal2Symbol offers a neuro‑symbolic pipeline that first converts raw signals into symbolic sequences using either a learned VQ‑VAE codebook or a SAX baseline. It then builds token‑window transactions enriched with bigram information and scores each window by mining minimal rare itemsets as evidence of rarity. Detected anomalous windows are merged into intervals and related through Allen interval algebra, enabling composite temporal explanations such as escalation chains, artifact overlap, and cross‑channel synchrony. Finally, a rare temporal concept lattice based on Formal Concept Analysis groups intervals by shared rare symbolic evidence, Allen relations, channel context, and robustness attributes, producing a Galois lattice that compresses many local detections into interpretable families of temporal‑symbolic anomalies. Evaluation on three public benchmarks—MIT‑BIH Arrhythmia (beat‑level ECG), PTB‑XL (record‑level ECG), and Bonn EEG (segment‑level EEG)—includes robustness tests under additive noise and baseline wander. The findings highlight that neuro‑symbolic tokenization benefits temporal anomaly analysis and that Allen/FCA reasoning yields compact, explainable summaries of detections.
Review