Skip to content
No description, website, or topics provided.
Python HTML R
Branch: master
Clone or download
Type Name Latest commit message Commit time
Failed to load latest commit information.
generation_projects debug Aug 2, 2019
output_processing add new folders for benchmark Jul 8, 2019
outputs redo some nli presuppositions datasets Aug 2, 2019
results/structure_dependent_experiments clean output file May 6, 2019
utils work on presuppositions Aug 1, 2019
.gitignore update gitignore Apr 2, 2019
collectivepredicates.tsv init for better conjugation Mar 6, 2019
dependencies add matrix question npi, irregular pl agreement Jul 31, 2019
generation_script remove "category" metadata Jul 31, 2019
intr_verbs.csv Added institutions Mar 21, 2019
vocabulary.csv fix simple anaphor gender agreement metadata Aug 1, 2019
wilcox_npi_paradigm.csv sentences from Wilcox et al. (2019) Apr 8, 2019


This project includes utilities for generating sentences with certain grammatical properties.


A shared vocabulary is vocabulary.csv to be read by the code.

The utils package contains shared functionality for reading the vocab and accessing fields.

The generation_projects package contains scripts for specific generated datasets.


The vocabulary lives in vocabulary.csv.

If you add a new column, you must update utils/

If you add a new row, definitely fill out all the relevant selectional restrictions.

Selectional restrictions are written in disjunctive normal form: A single condition is written as LABEL=VALUE. The symbol ";" is used for disjunction. The symbol "^" is used for conjunction. The entire selectional restriction should be written in the from a1^...^an;...;z1^...^zn. This matches any vocab item which matches conditions all of a1, ...., and an, OR ..., OR all of z1, ..., and zn

UTILS utils.conjugate includes functions which conjugate verbs and add selecting auxiliaries/modals utils.constituent_building includes functions which "do syntax": - build a subject relative clause from a head (subject_relative_clause) - gather all arguments of a verb (verb_args_from_verb) utils.data_type contains the all-important data_type necessary for the numpy structured array data structure used in the vocabulary utils.string_utils contains functions for modifying strings utils.vocab_table contains functions for creating and accessing the vocabulary table - get_all gathers all vocab items with a given restriction - get_all_conjunctive gathers all vocab items with the given restrictions

DOCUMENTATION Within each project's output directory, there is (should be) a docs document which explains: - the metadata in the output file - the data paradigm

GENERATION PROJECTS long distance npi plurality structure dependence

You can’t perform that action at this time.