Simon Berger, Larissa Kleppel, Ralf Schlüter · 2026 · arXiv (Cornell University)
DOI: 10.48550/arxiv.2610.31259 · cited by 0 works (OpenAlex) · open-access PDF
Die Suche ist ein zentraler Bestandteil der Systeme der automatischen Spracherkennung (ASR) und wurde für klassische ASR-Modellierungen auf der Grundlage von Hidden Markov Models (HMMs) tief untersucht. Bei der Arbeit an aktuellen Sequenz-zu-Sequenz-ASR-Architekturen wird es jedoch oft auf eine kleine Festgröße-Blitzsuche ohne externes Sprachmodell (LM) vereinfacht. In dieser Arbeit präsentieren wir eine vergleichende Analyse der Suchraumkomplexität und Effizienz für die Connectionist Temporal Classification (CTC) und neuronale Transducer-Modelle unter verschiedenen Beschneidungsstrategien und LM-Konfigurationen. Wir zeigen, dass diese End-zu-End-Faktoren (E2
Machine translation (Meta M2M100); the English original follows.
Search is a central component of Automatic Speech Recognition (ASR) systems and has been deeply investigated for classical ASR modeling based on Hidden Markov Models (HMMs). However, in work on current sequence-to-sequence ASR architectures it is often simplified to a small fixed-size beam search without external Language Model (LM). In this work, we present a comparative analysis of search space complexity and efficiency for Connectionist Temporal Classification (CTC) and neural transducer models under different pruning strategies and LM configurations. We show that these end-to-end (E2E) models work with a small search space most of the time but exhibit large temporary peaks in hypothesis count during periods of uncertainty. Score-based pruning dynamically adapts the search to model uncertainty, leading to a significantly lower Real-Time Factor (RTF). Our results highlight the limitations of search with low fixed-size beam pruning and provide practical guidelines for efficient search in E2E ASR.
Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese
Topics: Beam search, Pruning, Computer science, Hidden Markov model, Connectionism, Artificial intelligence
GET https://store.agentexchange.work/gw/papers/verify?title=...&year=... (0.003 USD), fetch this record as JSON with /gw/papers/paper?id=2610.31259, or list what it cites with /gw/papers/references?id=.... Pay per call in USDC (x402) or with a 19 USD card key for 1,000 calls. Catalogue · llms.txtSources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange