Read on Twitter
#roamcult, @RoamResearch, & @Conaw, have you seen @codexeditor 's SPEEDy editor? This thing is amazing: . There's an interesting conversation to be had here regarding resolution (block entity vs text annotation).
1/12
1/12
Advantages of text annotations: 1) No additional in-text markup ('[[', '#', '@', etc) to create/link to entities (the entities are in an entirely separate data structure). 2) As a result of 1., a wider variety of entity types are possible.
2/12
2/12
3) Entities are not limited to a hierarchical model (entities can overlap in the text). 4) Again, as a result of 1., text can be annotated without changing the text itself. 5) As a result of 3., there is greater information density possible.
3/12
3/12
6) As @codexeditor demonstrates, text annotation seems to pair extremely well with NLP and syntax/sentiment analysis. 7) Again, as a result of 1. (and 6?), you could extend the system to annotate texts that you don't have a copy of ...
4/12
4/12
...(using NLP to annotate the text on any website you visit to connect it to your database of entities in real time). NLP gets good enough and you can maybe just eat the web with this?
5/12
5/12
Advantages of block entities: 1) In-text markup limits the types of entities. 2) As a result of 1, users means you need few menus, making entry and writing incredibly fast.
6/12
6/12
3) Course-grained entities are excellent for things that are best understood as an atomic unit (code blocks, query blocks, widgets, etc) 4) As a result of 1., @RoamResearch eliminates some paradox of choice that users might get distracted by...
7/12
7/12
...(dealing with entity schemas/categorization in text annotation could be as big a time suck as file/folder structure). 5) As a result of 1, users can get productive with roam within minutes.
8/12
8/12
My takeaway is that, as a tool for writing/productivity, block entities are faster while providing enough power to be useful without providing so many options as to become distracting. Block entities also have a lot of potential for formal languages like queries and code.
9/12
9/12
That said, interop is not going to be great. In contrast text annotations seem dramatically more powerful for analysis. Entry and writing is quite a bit slower. But NLP may make integrating any text information into your knowledge graph fairly straight forward...
10/12
10/12
...as a separate step from creating the text itself). And future NLP could create a layer of text annotation that sits on top of the web. One can imagine the power of a search engine that has annotated the web.
11/12
11/12
So if that summary is accurate, two questions: 1) What can be done to make entry/writing in a text annotation system faster. 2) What can be done for a block entity system to take advantage of NLP in the same way that text annotation can/may-one-day-be-able-to?
12/12
12/12
