Skip to content
A psycholinguistic modeling toolkit
C++ Python Prolog Makefile Perl 6 R Other
Branch: master
Clone or download
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
docs resource-rt %toks doc correction Jan 27, 2018
resource-alice updated default directory paths to /data/compling and /data/corpora Dec 4, 2019
resource-bmmm
resource-bnc updated default directory paths to /data/compling and /data/corpora Dec 4, 2019
resource-ccl updated default directory paths to /data/compling and /data/corpora Dec 4, 2019
resource-chgcg updated default directory paths to /data/compling and /data/corpora Dec 4, 2019
resource-childes updated default directory paths to /data/compling and /data/corpora Dec 4, 2019
resource-conlen updated default directory paths to /data/compling and /data/corpora Dec 4, 2019
resource-dmv updated default directory paths to /data/compling and /data/corpora Dec 4, 2019
resource-dtsr
resource-dundee updated default directory paths to /data/compling and /data/corpora Dec 4, 2019
resource-epic updated default directory paths to /data/compling and /data/corpora Dec 4, 2019
resource-fedorenkoetal16ecog updated default directory paths to /data/compling and /data/corpora Dec 4, 2019
resource-fmri Brought fMRI data munging up to spec Oct 2, 2019
resource-forrest updated default directory paths to /data/compling and /data/corpora Dec 4, 2019
resource-gcg patch so sem annots are recognized in the presence of # comments Nov 15, 2019
resource-general Fixed new bug in natstor, edits to conlen and fedorenko165 Sep 19, 2019
resource-gigaword updated default directory paths to /data/compling and /data/corpora Dec 4, 2019
resource-glove updated default directory paths to /data/compling and /data/corpora Dec 4, 2019
resource-huth updated default directory paths to /data/compling and /data/corpora Dec 4, 2019
resource-incrsem updated default directory paths to /data/compling and /data/corpora Dec 4, 2019
resource-kenlm updated default directory paths to /data/compling and /data/corpora Dec 4, 2019
resource-lcparse removed unused dependencies Aug 16, 2018
resource-linetrees get rid of blankfail, fix grep single quotes Sep 26, 2019
resource-lmefit updated default directory paths to /data/compling and /data/corpora Dec 4, 2019
resource-logreg-torch Major push of new code from internal repo to release. Previously-exis… Feb 6, 2017
resource-logreg fixes and bounds checks stopping warnings from compiler and valgrind May 20, 2019
resource-lvpcfg updated default directory paths to /data/compling and /data/corpora Dec 4, 2019
resource-naturalstories updated default directory paths to /data/compling and /data/corpora Dec 4, 2019
resource-naturalstoriesfmri updated default directory paths to /data/compling and /data/corpora Dec 4, 2019
resource-ontonotes updated default directory paths to /data/compling and /data/corpora Dec 4, 2019
resource-passages updated default directory paths to /data/compling and /data/corpora Dec 4, 2019
resource-regression
resource-rgcn rough notes on using the RGCN output Nov 8, 2019
resource-rhacks updated default directory paths to /data/compling and /data/corpora Dec 4, 2019
resource-rt r formula matching for warnings Dec 4, 2019
resource-rvtl fixes and bounds checks stopping warnings from compiler and valgrind May 20, 2019
resource-segment-tokenize fix to sentence segmenter, used in wikisem Feb 7, 2019
resource-skipdep updated default directory paths to /data/compling and /data/corpora Dec 4, 2019
resource-srilm updated default directory paths to /data/compling and /data/corpora Dec 4, 2019
resource-tiger updated default directory paths to /data/compling and /data/corpora Dec 4, 2019
resource-tokenizer updated default directory paths to /data/compling and /data/corpora Dec 4, 2019
resource-tom updated default directory paths to /data/compling and /data/corpora Dec 4, 2019
resource-treebank change 'home/corpora' to 'data/corpora' for treebank Nov 22, 2019
resource-ucl updated default directory paths to /data/compling and /data/corpora Dec 4, 2019
resource-upparse updated default directory paths to /data/compling and /data/corpora Dec 4, 2019
resource-williams updated default directory paths to /data/compling and /data/corpora Dec 4, 2019
resource-word2vec updated default directory paths to /data/compling and /data/corpora Dec 4, 2019
resource-wordnet updated default directory paths to /data/compling and /data/corpora Dec 4, 2019
resource-xlsx2csv updated default directory paths to /data/compling and /data/corpora Dec 4, 2019
shainetal_2016_cl4lc minor bugfixes Oct 25, 2017
shainetal_2016_coling added missing filtering script May 9, 2017
.gitignore maintenance: added .gitignore Feb 23, 2016
LICENSE coling 2012 nguyen&al published Dec 19, 2012
Makefile Fix to prediction from LMER models Feb 15, 2019
QUICKSTART.md Update QUICKSTART.md Aug 7, 2018
README.md tweak to readme Jun 13, 2017
RESOURCES.md added Roark parser hooks; Aug 7, 2017
getResourceDescr.py tweak to RESOURCES.md formatting Jun 13, 2017
workspace_makefile_src.txt added to passages pipeline May 31, 2017

README.md

README for ModelBlocks

This is the Modelblocks software package. It includes several resources for constructing and evaluating broad-coverage probabilistic models of cognitive processes, organized into projects centered around different tasks and data sets.

Quickstart

To quickly get started using modelblocks, read through the quickstart guide in the neighboring QUICKSTART.md file. Below is more info about the structure of Modelblocks.

Use of Makefiles

Each project exists in a subdirectory of the main modelblocks directory. In order to ensure the reproducibility of experiments conducted using this resource, these project directories each contain a Makefile, which specifies how data sets, output files, and evaluation results are constructed. The repository contains several resource-XXX repositories that contain reusable libraries of code, project repositories that contain recipes for reproducing published experimental results, and a workspace for experimentation and development. Recipes will fail unless all dependencies to external resources (text corpora, experimental data, external code libraries, etc.) are satisfied. If you attempt to make a recipe that has a missing dependency, Make will exit with an error message about which dependency is missing and how you can access it.

For sandboxing and development, nearly all ModelBlocks recipes can be created from a single workspace. To initialize your workspace, simply type make at the repository root, then navigate to the workspace directory. To reproduce a published experiment, navigate to the relevant experiment directory (named by author/year) and type make. NOTE: We do not guarantee indefinite future support of all published results recipes. In some cases it may be necessary to revert the repository to some previous state in order to reproduce a result. If you are encountering errors as you try to reproduce a result, please contact the ModelBlocks development team.

Documentation for the targets provided by each Makefile is provided in the docs directory of this repository. For help building targets for common recipes, you can use the browser-based ModelBlocks Assistant tool. This can be created locally by navigating to the docs directory of ModelBlocks and typing make mbassist. A remote version can also be accessed online at http://go.osu.edu/mbassist.

Included Resources

Modelblocks makes use of several third-party data and software resources. Where licenses permit, these have been included directly in the modelblocks package, so as to avoid version compatibility issues and thereby ensure reproducibility. In some cases open-source software has been modified so as to produce a common data file format required by other software. All resources included in this package are distributed under the Gnu General Public License (see LICENSE file in this directory).

External Resources and USER-*.TXT Files

When licenses of resources used in Modelblocks projects do not permit redistribution, or when (usually data) resources are too large to be included, the Makefile will generate an appropriately-named user-*.txt configuration file in modelblocks-release/config/, in which a user may specify a path to an external copy of the resource. When Make is first invoked, ModelBlocks will create an incorrect pointer for each configuration file in the dependency chain, along with console output indicating which configuration files are needed to create the recipe. Before re-running Make, the needed third-party resources will need to be downloaded and the pointers updated in modelblocks-release/config. This use of user-*.txt files is intended to allow users to specify external resources or other user- specific data without having to modify the Makefile, which may be overwritten in subsequent updates to ModelBlocks.

External resources are described in RESOURCES.md. RESOURCES.md also specifies which user-*.txt files are associated with each external resource.

You can’t perform that action at this time.