Dependency-length minimization and its limits: A possible role for a probabilistic version of the final-over-final condition: Supplementary Material

Dependency length minimization and its limits*: Statistical analyses∗ Yingqi Jing, Damián E. Blasi, and Balthasar Bickel July 12, 2022 Contents S1 Data Preprocessing 4 S1.1 Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 S1.2 Preprocess data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 S1.3 Check projectivity for each sentence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 S2 Random Baselines 7 S2.1 Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 S2.2 Perform randomizations and linearizations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 S2.2.1 Novel baseline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 S2.2.2 Free word order baseline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 S2.2.3 Fixed head position baseline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 S3 Overview of Languages in UD 2.5 18 S3.1 Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18 S3.2 Summary of treebank information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18 S4 BAM Modeling 24 S4.1 Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24 S4.2 Model comparison in parallel catenae . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26 S4.2.1 Distribution of DLM with head-finality and harmony . . . . . . . . . . . . . . . . . . . 26 S4.2.2 Model comparison via F test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32 S4.2.3 Inspection of summed effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33 S4.2.4 Additive model with interaction for each language . . . . . . . . . . . . . . . . . . . . 36 S4.2.5 Cross-validation in parallel catenae . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39 S4.2.6 Re-analyze interaction effects by discretizing explanatory variables . . . . . . . . . . . 39 S4.2.7 Relationship of head-finality and harmony on DLM in each langauge . . . . . . . . . . 41 S4.3 Model comparison in hierarchical catenae . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44 S4.3.1 Distribution of head-finality and harmony in hierarchical catenae . . . . . . . . . . . . 47 S4.3.2 Model comparison via F test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52 S4.3.3 Inspection of summed effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53 S4.3.4 Additive model with interaction for each language . . . . . . . . . . . . . . . . . . . . 56 S4.3.5 Cross-validation in hierarchical catenae . . . . . . . . . . . . . . . . . . . . . . . . . . 59 S4.3.6 Re-analyze interaction effects by discretizing explanatory variables . . . . . . . . . . . 59 S4.3.7 Relationship of head-finality and harmony on DLM in each langauge . . . . . . . . . . 61 S4.4 Variation of DLM at the language level . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65 S5 Linearization of Parallel and Hierarchical Catenae 66 ∗This document was rmarkdown::render’ed from https://osf.io/a2jvw/?view_only=74166131be1d4839b0bfc70875410389 1 Supplementary materials for ‘Dependency-length minimization and its limits: A possible role for a probabilistic version of the final-over-final condition’, by Yingqi Jing, Damián E. Blasi, and Balthasar Bickel. Language 98(3).397–418, 2022. S5.1 Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66 S5.2 Parallel catenae . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69 S5.2.1 Ordering of parallel dependents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69 S5.2.2 Distribution of number of parallel dependents at the same side . . . . . . . . . . . . . 71 S5.2.3 Relationship between head-finality and entropy of number of parallel dependents . . . 71 S5.2.4 Visualize parallel catenae . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74 S5.3 Hierarchical catenae . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78 S5.3.1 Modeling FOFC in hierarchical catenae . . . . . . . . . . . . . . . . . . . . . . . . . . 78 S5.3.2 Visualize hierarchical catenae . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79 References 84 List of Tables S1 Overview of 45 dependency treebanks selected in this study . . . . . . . . . . . . . . . . . . . 23 S2 Model performance via 10-fold cross-validation for parallel catenae . . . . . . . . . . . . . . . 39 S3 Summary of DLM in parallel catenae between different random baselines . . . . . . . . . . . . 45 S4 Model performance via 10-fold cross-validation for hierarchical catenae . . . . . . . . . . . . . 59 S5 Summary of DLM in hierarchical catenae between different random baselines . . . . . . . . . 64 List of Figures S1 Overview of head-finality and harmony in all languages . . . . . . . . . . . . . . . . . . . . . 19 S2 Relationship between head-finality, harmony and proportion of lexical dependencies . . . . . . 20 S3 Distribution of head-finality and harmony in parallel catenae . . . . . . . . . . . . . . . . . . 27 S4 Distribution of DLM with head-finality and harmony in parallel catenae for all languages . . 28 S5 Distribution of DLM with head-finality and harmony in parallel catenae for each language (novel baseline) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29 S6 Distribution of DLM with head-finality and harmony in parallel catenae for each language (free word order baseline) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30 S7 Distribution of DLM with head-finality and harmony in parallel catenae for each language (fixed head position baseline) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31 S8 Contour plots of head-finality and harmony on DLM with interaction (left) and without interaction (right) in parallel catenae (novel baseline) . . . . . . . . . . . . . . . . . . . . . . . . 34 S9 Contour plots of head-finality and harmony on DLM with s() term and fREML in parallel catenae (novel baseline) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34 S10 Contour plots of head-finality and harmony on DLM in comparison to free word order baseline (left) and fixed head position baseline (right) . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 S11 BAM model of DLM effects in parallel catenae for each language with the novel baseline . . . 36 S12 BAM model of DLM effects in parallel catenae for each language with the free word order baseline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37...