Pausanias Analysis

Book-number regression from stylometric features

Can Stylometric Features Predict Book Number?

This page uses Ridge regression, not classification, to predict the numeric book number of each grammar-parsed sentence. Chance is the per-fold training-set mean book number.

These models use only sentences with current LLM grammar-token rows, so the sample is smaller than the full annotation corpus.
Strict three-way and mythic-vs-historical tasks drop rare both/mixed rows; one-vs-rest tasks use the available mythic/historical flags where present.
Character 4-grams are included as a traditional baseline; morphosyntax is the primary non-lexical stylometric family.

Open feature trend diagnostics | Download regression metrics CSV | Download regression features CSV

Regression Metrics

Variant	Features	n	Book counts	Features	Alpha	R2	MAE	Chance MAE	MAE improvement	RMSE	Status
Including Books 4 and 8	Morphosyntax	8,029	1: 993, 2: 895, 3: 681, 4: 766, 5: 667, 6: 600, 7: 673, 8: 1,018, 9: 811, 10: 925	240	3162.278	0.050	2.595	2.696	0.101	2.958	ready
Including Books 4 and 8	Function Words	8,029	1: 993, 2: 895, 3: 681, 4: 766, 5: 667, 6: 600, 7: 673, 8: 1,018, 9: 811, 10: 925	160	1000.000	0.030	2.627	2.696	0.069	2.987	ready
Including Books 4 and 8	Character 4-Grams	8,029	1: 993, 2: 895, 3: 681, 4: 766, 5: 667, 6: 600, 7: 673, 8: 1,018, 9: 811, 10: 925	220	3162.278	0.056	2.581	2.696	0.114	2.947	ready
Excluding Books 4 and 8	Morphosyntax	6,245	1: 993, 2: 895, 3: 681, 5: 667, 6: 600, 7: 673, 9: 811, 10: 925	240	3162.278	0.065	2.730	2.867	0.137	3.132	ready
Excluding Books 4 and 8	Function Words	6,245	1: 993, 2: 895, 3: 681, 5: 667, 6: 600, 7: 673, 9: 811, 10: 925	160	3162.278	0.039	2.784	2.867	0.082	3.175	ready
Excluding Books 4 and 8	Character 4-Grams	6,245	1: 993, 2: 895, 3: 681, 5: 667, 6: 600, 7: 673, 9: 811, 10: 925	220	3162.278	0.072	2.722	2.867	0.144	3.121	ready

Predictive Stylometric Features

Variant	Feature family	Largest standardized coefficients
Including Books 4 and 8	Morphosyntax	later_books: `Head>child PROPN>DET` (0.126) earlier_books: `Feature Voice=Act` (-0.091) later_books: `DepRel/UPOS cc:PART` (0.081) later_books: `Feature PronType=Prs` (0.075) earlier_books: `DepRel/UPOS obj:PROPN` (-0.074) later_books: `DepRel/UPOS obj:NOUN` (0.072) later_books: `Feature Gender=Neuter` (0.072) later_books: `UPOS NOUN` (0.072)
Including Books 4 and 8	Function Words	later_books: `Function form/UPOS οἱ:DET` (0.148) earlier_books: `Function UPOS PRON` (-0.133) earlier_books: `Function DepRel advmod` (-0.110) earlier_books: `Function word δὲ` (-0.106) later_books: `Function form/UPOS ἡ:DET` (0.098) earlier_books: `Function DepRel obl` (-0.098) later_books: `Function word ὑπὸ` (0.092) later_books: `Function form/UPOS ὑπὸ:ADP` (0.092)
Including Books 4 and 8	Character 4-Grams	earlier_books: `char 4-gram αὶ τ` (-0.123) earlier_books: `char 4-gram ἀθην` (-0.116) later_books: `char 4-gram αὶ ἐ` (0.108) earlier_books: `char 4-gram λέγ` (-0.105) later_books: `char 4-gram ἐκ` (0.086) later_books: `char 4-gram τοῦ` (0.085) earlier_books: `char 4-gram τον` (-0.083) later_books: `char 4-gram ὑπὸ` (0.082)
Excluding Books 4 and 8	Morphosyntax	later_books: `Head>child PROPN>DET` (0.127) earlier_books: `Feature Voice=Act` (-0.098) later_books: `Feature Number=Plur` (0.098) later_books: `DepRel/UPOS cc:PART` (0.094) earlier_books: `DepRel/UPOS obj:PROPN` (-0.090) later_books: `Feature Gender=Neuter` (0.086) earlier_books: `Feature Number=Sing` (-0.086) later_books: `Feature Degree=Pos` (0.084)
Excluding Books 4 and 8	Function Words	later_books: `Function word τε` (0.093) later_books: `Function form/UPOS ὑπὸ:ADP` (0.091) later_books: `Function word ὑπὸ` (0.091) later_books: `Function form/UPOS τε:PART` (0.087) later_books: `Function form/UPOS οἱ:DET` (0.086) later_books: `Function form/UPOS ἡ:DET` (0.084) earlier_books: `Function word δὲ` (-0.082) later_books: `Function word ἡ` (0.078)
Excluding Books 4 and 8	Character 4-Grams	later_books: `char 4-gram ντο` (0.139) earlier_books: `char 4-gram θηνα` (-0.126) earlier_books: `char 4-gram αὶ τ` (-0.116) later_books: `char 4-gram αὶ ἐ` (0.115) later_books: `char 4-gram τε` (0.113) earlier_books: `char 4-gram καλ` (-0.110) later_books: `char 4-gram ἐκ` (0.104) earlier_books: `char 4-gram λέγ` (-0.102)