cltl/dutchdescriptions

Dutch descriptions for the Flickr30K validation and test data, plus a cross-lingual comparison tool.

DutchDescriptions

Dutch descriptions for the Flickr30K validation data, plus a cross-lingual comparison tool. This repository is part of this paper, which will be presented at INLG 2017.

Here is a temporary BibTeX entry, which will be updated when the paper has been included in the ACL anthology.

@unpublished{miltenburg2017cross,
	Author = {Emiel van Miltenburg and Desmond Elliott and Piek Vossen},
	Note = {Paper accepted at INLG 2017, \url{https://arxiv.org/abs/1707.01736}},
	Title = {Cross-linguistic differences and similarities in image descriptions},
	Year = {2017}}

Folder structure

• Crowdsourcing Contains the data used as input for the crowdsourcing task.
• Data Contains all the image description data, including the newly collected Dutch data.
• InspectionTool Contains code and data for the inspection tool.
• Stats Contains scripts to generate statistics about the data, and the output of these scripts.
• Utils Contains a script to anonymize the data, and a script to generate Dutch validation files.

Requirements

We used Python 3.6.1 with the following external libraries:

• Flask 0.1.2
• NLTK 3.2.2
• Numpy 1.12.0
• Matplotlib 2.0.0
• Matplotlib-venn 0.11.5

Notes

This repository does NOT contain the Flickr30K images. Please visit the Flickr30K website and fill in the form to obtain the images.