Pausanias Analysis

Sentence-level words and phrases that predict skepticism vs. non-skepticism

Model Performance Metrics

The following metrics are from the logistic regression classifier's performance on the test set:

Class Precision Recall F1-Score Support
Non-skeptical 0.900 0.749 0.817 867
Skeptical 0.268 0.526 0.356 152
Overall Accuracy 0.715 1019

Sentence Predictors of Skeptical Content

Sort by:
Word/Phrase English Coefficient Skeptical Count Non-skeptical Count p-value q-value
ποιῆσαι to make; to do 3.7172 23 9 6.26e-13 3.13e-10
εἴτε whether 3.0190 14 2 3.15e-10 1.05e-07
γενέσθαι to become 2.9623 59 85 4.63e-14 4.63e-11
εἰ if 2.6451 29 47 1.04e-06 0.000104
ὑπὸ under; beneath; by (of agent) 2.6434 104 283 1.07e-09 2.68e-07
ταύτην this one 2.4353 35 58 1.15e-07 1.44e-05
παῖδα child 2.3697 36 79 1.36e-05 0.000973
ἅιδου of Hades 2.1515 9 6 7.66e-05 0.00454
ὡσ as; like; how; that; when 2.0636 116 352 2.01e-08 4.01e-06
οὐδὲν nothing 2.0615 19 45 0.00349 0.0908
οὐκ not 2.0451 74 205 6.96e-07 7.73e-05
οἱ δὲ but they 1.8535 29 81 0.00259 0.0786
μεγαρεῖσ Megarians 1.8341 7 8 0.00412 0.0981
ὄνομα name 1.8272 47 94 6.8e-08 1.13e-05
τοῦτον this one 1.7733 38 77 1.71e-06 0.000155
ἄνδρασ men 1.7003 10 21 0.0184 0.236
ὅτωι to whomever 1.6809 7 13 0.0299 0.284
αὐτήν her 1.6246 8 12 0.00761 0.132
ἔργα deeds 1.6161 9 20 0.0334 0.298
ταῦτα these (things) 1.5929 21 80 0.139 0.541
ἐλθεῖν to come 1.5478 11 13 0.000392 0.0196
ἀποθανεῖν to die 1.5242 15 26 0.000822 0.0353
τελευτήσαντοσ of the deceased 1.5230 8 15 0.0211 0.244
παῖδασ children 1.4901 14 42 0.0592 0.389
τὸν the 1.4899 186 679 9.44e-08 1.35e-05
καὶ ἀπὸ and from 1.4882 11 14 0.000622 0.0283
ποταμὸν river 1.4762 7 12 0.0217 0.244
λόγοσ word; speech; account; reason; argument 1.4429 16 37 0.00613 0.116
δʼ but; and 1.4344 29 85 0.00472 0.103
οἰκῆσαι to dwell 1.4197 7 9 0.00668 0.121

Sentence Predictors of Non-skeptical Content

Sort by:
Word/Phrase English Coefficient Skeptical Count Non-skeptical Count p-value q-value
θάλασσαν sea -2.2047 4 56 0.0697 0.422
λακεδαιμονίοισ to the Lacedaemonians -1.7903 1 57 0.00138 0.0553
δὲ τῆσ but of the -1.7882 5 65 0.0358 0.309
ἦν was -1.7640 30 197 0.359 0.788
σφισιν to them -1.7290 11 139 0.00259 0.0786
τε and -1.6765 132 882 0.0207 0.244
τότε then; at that time -1.3560 14 114 0.145 0.541
κατʼ according to; against; down -1.3160 2 29 0.215 0.652
πόλισ city -1.3045 1 38 0.0232 0.252
ὁδὸν road -1.2902 2 41 0.054 0.389
ἐστι is -1.2779 23 190 0.0505 0.377
ἔχων having -1.2383 2 38 0.0768 0.435
λίθου of a stone -1.2288 4 66 0.0193 0.244
παῖδεσ children -1.2155 2 44 0.0378 0.309
ἔσχε he/she/it had -1.2051 0 27 0.0266 0.266
ἔχουσα having (feminine) -1.1997 1 20 0.235 0.656
ἔσχεν he/she/it had -1.1727 0 27 0.0266 0.266
ἐστίν is (he/she/it is) -1.1654 1 38 0.0232 0.252
ὁδὸσ road; way -1.1485 0 23 0.0386 0.309
καὶ and -1.1457 463 2736 0.0297 0.284
ἐσ τὸ into the -1.1418 15 123 0.124 0.521
πρὶν before -1.1166 9 46 0.831 0.993
παρέχεται is provided -1.1062 2 22 0.567 0.899
μάλιστα especially; very much; certainly -1.1025 21 168 0.0881 0.472
ἀνάκειται is set up -1.0633 0 18 0.0954 0.472
θέασ of a view -1.0606 1 52 0.00317 0.0882
ἀρχήν at first -1.0467 3 30 0.466 0.881
τὴν ἀρχὴν the beginning -1.0355 2 25 0.418 0.844
εἰσι they are -1.0335 2 27 0.3 0.748
αἱ the -1.0334 7 93 0.0103 0.162

Simplified Checklist Version

This reduced model keeps only the words or phrases with q-value below 0.1, then turns them into a simple checklist.

Start at -0.64 points. If one of the items below appears at least once, add its points once. If the final score ends above 0, classify the text as Skeptical. That is the same as saying the listed word-points must add up past +0.64.

Full logistic model: 71.5% | Simplified checklist: 74.0% | Always guess Non-skeptical: 85.1%

This shorter model uses 44 statistically strong vocabulary items.

These confusion matrices show which texts each approach gets right and where it makes false alarms or misses.

Full Logistic Model

Predicted
Non-skeptical Skeptical
Actual Non-skeptical 649 218
Skeptical 72 80

Simplified Checklist

Predicted
Non-skeptical Skeptical
Actual Non-skeptical 668 199
Skeptical 66 86

Always Guess Non-skeptical

Predicted
Non-skeptical Skeptical
Actual Non-skeptical 867 0
Skeptical 152 0
Word/Phrase English Points If Seen Pushes Toward q-value
εἴτε whether +5.54 Skeptical 1.05e-07
οὐ μὴν however +4.33 Skeptical 0.00899
ποιῆσαι to make; to do +2.97 Skeptical 3.13e-10
μὴν indeed; surely -2.46 Non-skeptical 0.0272
μεγαρεῖσ Megarians +2.35 Skeptical 0.0981
λακεδαιμονίοισ to the Lacedaemonians -2.32 Non-skeptical 0.0553
ἅιδου of Hades +2.05 Skeptical 0.00454
θέασ of a view -2.03 Non-skeptical 0.0882
καὶ ἀπὸ and from +1.48 Skeptical 0.0283
ἐλθεῖν to come +1.42 Skeptical 0.0196
ταύτην this one +1.41 Skeptical 1.44e-05
δʼ ἂν but would +1.37 Skeptical 0.00454
εἰ if +1.21 Skeptical 0.000104
γενέσθαι to become +1.17 Skeptical 4.63e-11
σφισιν to them -1.14 Non-skeptical 0.0786
τὸ ὕδωρ the water +1.12 Skeptical 0.0699
παῖδα child +1.02 Skeptical 0.000973
αὐτόν him +1.02 Skeptical 0.0579
καὶ τοῦτο and this +1.01 Skeptical 0.0682
οὐδὲν nothing +0.99 Skeptical 0.0908
τοῦτον this one +0.94 Skeptical 0.000155
τὴν γῆν the land +0.91 Skeptical 0.0814
ἐπεὶ since +0.87 Skeptical 0.0908
λόγον word +0.84 Skeptical 0.00404
εἴη may it be; let it be +0.80 Skeptical 0.0848
ὑπὸ under; beneath; by (of agent) +0.80 Skeptical 2.68e-07
ὄνομα name +0.76 Skeptical 1.13e-05
ἀποθανεῖν to die +0.75 Skeptical 0.0353
ἄνδρα man +0.69 Skeptical 0.000418
οἱ δὲ but they +0.68 Skeptical 0.0786
οὐκ not +0.59 Skeptical 7.73e-05
οὐδὲ nor +0.50 Skeptical 0.0353
καὶ ὡσ and as +0.49 Skeptical 0.0564
ὡσ as; like; how; that; when +0.33 Skeptical 4.01e-06
ὁμήρου of Homer +0.32 Skeptical 0.0981
ἄλλωσ otherwise +0.31 Skeptical 0.0908
ἂν a modal particle indicating potentiality, often rendered "would" or "might"; with relatives/temporals it adds "-ever" (e.g., "whoever," "whenever") +0.31 Skeptical 0.00962
τὸν the +0.25 Skeptical 1.35e-05
τοῦτο this +0.22 Skeptical 0.0981
τῆι to the +0.19 Skeptical 0.0786
ἐστιν is -0.19 Non-skeptical 0.0699
τὸ ὄνομα the name +0.18 Skeptical 0.000155
τῆι πόλει for the city -0.04 Non-skeptical 0.0908
τὴν the +0.03 Skeptical 0.0981