wordspend.symmachus.org

Word Spend

Translation residuals as evidence for relative lexical compression.

Current Results

The live analysis fits a source-to-target length baseline and then looks for source and target terms associated with unexpectedly long or short translations. Positive residuals mean the translation spends more words than expected after controlling for source length.

Segments3170
Works1
Metrictokens
Baselinelog
R squared0.862
Residual SD0.100
Source features2000
Target features2000

Source Length vs Target Length

Source Terms Predicting Residuals

Feature Coefficient Segments Mean residual with Mean residual without
ἑλένης0.0463180.091-0.001
τὸ ἱερόν-0.041538-0.0260.000
πελοποννησίων0.0393310.057-0.001
καθέστηκεν0.0379400.030-0.000
ἄλλοι-0.037477-0.0470.001
τοῦ μὲν0.0368180.033-0.000
οὐκ ἂν0.0356220.055-0.000
unknown-0.034630-0.0300.000
θεὸς0.0335570.030-0.001
παλαιοῦ0.0332300.053-0.001
μεσσηνίοις0.0331590.039-0.001
ἡρῷον0.0331260.084-0.001
καὶ ταῖς-0.032818-0.0570.000
ἱερόν0.03221690.027-0.002
τῇ πόλει-0.031876-0.0100.000
ἐκ παλαιοῦ0.0318290.053-0.000
νομίζοντες0.0317220.028-0.000
χρημάτων0.0317450.027-0.000
δρόμον0.0315230.040-0.000
δὲ ἢ-0.031528-0.0310.000
πολλά-0.031521-0.0350.000
αὐτοῖς καὶ0.0314180.020-0.000
τὸ ἀπὸ-0.030923-0.0690.001
ἀνδρῶν0.0308670.022-0.000
ἐντεῦθεν0.0306500.036-0.001

Target Terms Predicting Residuals

Feature Coefficient Segments Mean residual with Mean residual without
make0.0350530.045-0.001
dedicated to0.03421660.042-0.002
even to0.0334510.053-0.001
because he0.0325370.063-0.001
rivers-0.032128-0.0730.001
a great0.0312300.039-0.000
the sons0.03071040.006-0.000
who won0.0305300.066-0.001
year of0.0303250.024-0.000
orchomenus-0.029632-0.0730.001
who composed-0.029526-0.0320.000
that a0.0293380.053-0.001
proclaimed-0.029231-0.0090.000
armor0.0289270.027-0.000
a temple0.02892080.039-0.003
poetry-0.028742-0.0310.000
use0.0287250.055-0.000
here the-0.028436-0.0240.000
road from-0.027830-0.0290.000
cavalry-0.027735-0.0490.001
are said-0.027635-0.0140.000
me to0.0276250.043-0.000
established the0.0272350.052-0.001
shall0.0272760.041-0.001
fought against0.0270300.023-0.000

Downloads

metrics.json segments.csv source_predictors.csv target_predictors.csv