Distilling Task Knowledge from How-To Communities



The HowToKB dataset consists of input WikiHow (crawled and structured), as well the resulting KB here

HowToKB data

==========================================================================================================================================
- wikihow-data-all.json: wikihow data in .json format
==========================================================================================================================================
- task-frame-before-clustering.json: task frames extracted from wikihow data before clustering in .json format
==========================================================================================================================================
- task-frame-after-clustering.json: task frames after clustering: merging all task frames in a same cluster into an task frame

Gold Corpus
- task-frame-gold-corpus.json: prominent task frames after clustering (including post-processing): merging all task frames in a same cluster into an task frame
==========================================================================================================================================
- task-frame-id-before-to-task-frame-id-after: mapping task frame id before clustering to after clustering

- wikihow-id-category.txt: mapping id to category of wikihow

- wikihow-id-url: mapping id to url of wikihow


- articles-word2vec-word-pos.model.txt: word2vec model, trained on all wikihow data

==========================================================================================================================================
==========================================================================================================================================

Data for usecase

- task-frame-have-video.json: wikihow task frames which have youtube video (extract from wikihow data)

- uniq-youtube-video.json: youtube video information crawled from youtube in .json format

- id-frame-wikihow-url-youtube-link: mapping: taskframe id -- wikihow url -- youtube link