Meeting Time: March 1, 2020, 3-5 pm (Pacific)Meeting Location: Hollister Brewing CompanyMeeting Zoom: We'll use Alan's "instant" Zoom ID (our default meeting Zoom): https://ucsb.zoom.us/j/760-021-1662 |
Red font = notes from the meeting
1. Dataset Releases
- Primary public release: on Web site or Github?
- Large dataset releases as organized by keyword/conceptual unit in MongoDB (e.g., humanities, science, comparison data, Reddit)
- Flat file datasets
- Accompanied by Manifests
- Queryable access via Manager?
- Notes from our Discussion:
- Datasets will be large batches of zipped JSON files (of "bags of tokens"), with a manifest, in a Frictionless Data package
- Those batches will need to be sectioned and otherwise made useable for the laptop-based user (who will otherwise find it impossible to download, open, and work with millions of files)
- We will also include with the datasets utilities that make the JSONs useable in other common formats:
- Scott's notebook for converting JSONs to CSV; and also a tool for converting to plain text
- Datasets will be deposited in GitHub and Zenodo.
- We intend to do a demo by the time we meet with Greg in early April.
- In regard to queryable access:
- Manager will not be available for that within the scope of this grant.
- But see below on hosting our topic models on our site
- "Collection" dataset releases--for each collection:
- JSONs of "bags of tokens" with citations
- Mallet files
- Vizualization files
- Notes from our discussion:
- These collections, models, and visualizations will be packaged like the large-batch datasets above for downloading
- However, we will also host the models live from our web server.
- And we will also make available containers for each collection (able to serve up the live models).
2. Social media datasets
3.
4.
5.
Planning for Future Meetings