The Book of Isaiah: Morphological Clustering and Disputed Authorship
What it is about
Kasher asks whether non-stylistic hypotheses about the authorship and integrity of the Book of Isaiah can be tested with stylistic criteria. Each chapter is treated as a point in an n-dimensional Euclidean space, whose coordinates are statistical measures of the morphological features of the chapter's words. Clustering is then applied to these points, and the clusters are used to confirm or refute the hypotheses. The text file we have is the one-page abstract only, so the dimensions, the features and the outcome of the clustering are not stated here.
Ideas
- I1. Chapter vectors from the MACULA Hebrew trees: for each chapter, count morphological features (verb stems, suffix types, pronominal forms, construct states) and cluster the chapters. Build it as a report next to
macula compare. - I2. A stylometric check with a fixed rule: cluster with a deterministic method (sorted maps, no random seed), then test whether the clusters recover the traditional split of the book into chapters 1–39 and 40–66. Count the chapters that cross the split; a clean split is the claim to test, and it can fail.
- I3. The weak point Kasher's abstract leaves open is the number of dimensions. Fix the dimension in advance and report how stable the clusters are when one feature is dropped.
For Mova
I1 and I2 are a report and a test on macula compare output, with the feature list kept in languages/hbo/04-noun.md and languages/hbo/06-verb.md. I3 is a note for languages/hbo/13-mova-tools.md.
Cite
@inproceedings{kasher1969book,
title = {{The Book of Isaiah: Morphological Clustering and Disputed Authorship}},
author = {Kasher, Asa},
booktitle = {International Conference on Computational Linguistics COLING 1969: Preprint No. 69: Collection of Abstracts of Papers},
year = {1969},
doi = {10.3115/990403.990486},
url = {https://doi.org/10.3115/990403.990486}
}