Resources

You can find most of my code on Github.

This page links to a couple of specific resources related to one of my Publications.

As part of our paper “Teaching Machines to Read and Comprehend” we published a tool to rebuild the large reading comprehension corpus described in the paper. You can find this tool on Github: https://github.com/deepmind/rc-data/

Together with our 2013 ACL paper on graph parsing with hyperedge replacement grammars, we released a set of tools for doing exactly that. Bolinas can process rooted, directed, cyclic, edge and node-labeled (hyper-)graphs. The software is hosted at ISI: http://www.isi.edu/licensed-sw/bolinas/

The code for my 2013 ACL paper on compositional semantics using CCG and vectors is hosted on Github: https://github.com/karlmoritz/oxcvsm. Note that this code is deprecated and I suggest you look at the BiCVM package (below) instead, which supersedes this package and should provide a cleaner, more easily extensible model.

The code for my 2014 ACL and ICLR papers on learning multilingual word embeddings and language models is hosted on Github: https://github.com/karlmoritz/bicvm. The code in BiCVM works for replicating the results in my papers and can easily be extended with more complex composition functions.

As part of these publications I have published a modified version of the TED corpus (https://wit3.fbk.eu/), which you can find here.