CrystalASR: Hierarchical Phoneme-Grounded Speech Decoding

Archived by the author. A methodological flaw in this version undermines one of its central conclusions, so it should not be cited. It has been superseded by a later paper.

While preparing follow-up work I found a methodological problem in this version. It is not a typo or a presentation issue: it sits in how one of the central analyses was set up, and the conclusion drawn from that analysis does not survive a correct treatment. Nothing in this version should be read as a result.

I have removed the DOI, the BibTeX entry, and the machine-readable citation metadata from this page, and the PDF is no longer linked from it. The page itself stays up. The work happened, and quietly deleting it would make for a worse record than leaving it marked.

This was one of my earliest pieces of research, and it reads that way. The question was worth asking; the execution was greener than I would like. The work has since been redone from the ground up and published separately as How Much of Speech Recognition Must Be Learned? A Parameter-Free Analysis of Lexical Decoding. That paper replaces this one; please cite it instead.