fako / discourse_scope Goto Github PK

Deduplication of entries now takes a lot of time in Pandas on the request. The deduplication of entries is much faster when precomputed during growth phase

The creation of a corpus takes multiple steps. It should be clear which steps are involved, how long it takes and how much it costs. At any time in the process the user should have an end-in-sight.

Details about the steps are in the flow diagram below.

• https://gist.github.com/sj26/7079674
• https://medium.com/ninjaconcept/interactive-dynamic-force-directed-graphs-with-d3-da720c6d7811
• https://github.com/ninjaconcept/d3-force-directed-graph/tree/master/example
• https://itnext.io/d3-force-directed-graph-forces-experiments-for-dummies-20a5682935
• http://bl.ocks.org/sebbacon/6138241 (best)

Make sure that it's possible to deselect an item without scrolling. More bugs might be hiding there :(

We should create an interview questionnaire and find 3 to 5 users to have the interview with. The users should be using the product in order to see how they are using it.

This research is limited to the "exploration of the corpus" and does not involve any "creation of a corpus" functionality.

get_ranking_results works with individual iterators, but because of this it has to solve its problems in pure Python instead of in Cython. This makes it unnecessarily slow.

Currently you can view and add discourses. But in fact they are corpora. The interface should reflect this change and explain what a corpus is to people that don't know and are interested.

Related to this is that I think I prefer an empty "corpus" with a big + inside for a user to add their own corpus. Instead of a small button in the header.

Currently only the newer discourses have the new data layout. The data layout should remain consistent.

Computing these on request is very slow. So we need a place to precompute and store these.

Here's the list of missing filters

• Sources
• Vocabulairy (n-grams)
• Interesting people
• Interesting organisations

Currently nothing is done with multiple authors. For now it should take the primary author if there are multiple.

Sometimes a growth is started upon request, but it doesn't match the configuration and that screws up the request. Either redirect to correct request, or indicate a bad request, disallow starting a request or allow growths to take place on request.

This bug is related to the inability to refresh on details pages in app.

Required input is:

• A topic
• A description
• List of subjects
• Pro framing words
• Con framing words

The input on this page is assumed to be flawed and an assistant (automated or not) will have to give suggestions to improve on this input. When the user first fills the information, no suggestions are now. However when they come back to it a second time the original input and suggestions should show side-by-side. There should be a way for users to accept and reject such suggestions and/or to add and delete their own input based on those suggestions.

Below is an image of the exact flow where this page will fit in. It describes in more detail the information that we need.

It takes up to 26s to complete in some cases

Frontend uses the wrong DSN

https://github.com/lejafar/factrank