Chrome extension to display contextual knowledge

The new Web clipper chrome extension is a very good first step to Fibery integrated in the place where a lot of the knowledge discovery happen: the browser.
It is great for inputing new knowledge in Fibery, that then will be linked, and used in our workflow

We are missing the opposite, an extension (could be packaged as the same chrome extension) displaying contextual information from Fibery when browsing the web.
Few uses cases:

  1. Linkedin profiles: I would like to see my Contact Entity (with info if the person is/was a candidate, my app user, partner, etc) in a side bar. Mapping key would be the linkedin url field. I already have that for my ATS (Lever.com) chrome extension for candidates, or with LinkedMatch for Pipedrive for prospect/client contacts, etc.
  2. news/article: I could see that a colleague already import this relevant news, and linked it to other entities.

For 1. and 2. We could generalize the matching: display any Entity that have a url field that match the current page url (they are caveat with url parameters, 301 redirects, etc, but you get the idea)

  1. Entities recognition: highlight in a page the words that match an Entity name, and on hover or click display context from Fibery. Could be configurable to decide which Types should be part of this automatic annotation, to avoid clutter.
    Maybe some Named Entities Recognition algos could be used to improve the matching quality.

Basically Fibery knowledge would be much more useful if they are used to contextualize the outside information.

For reference, the Web clipper thread:

3 Likes

I think this is a pretty clever idea, which could be helped by changing how the current extension works a bit. Instead of just dumping text into one of your entities, instead it would help to model at the very least the source of the text better.

This would require some extra setup or the extension would need the option to do it, but you could have a Domain and Page type. Each new “thing” you capture from the extension would need to add the Page and Domain if they don’t exist yet, then add the thing you are capturing with a reference the the Page. So, you could then lookup data by Page and Domain to show relevant information associated.

I definitely see how it could be possible to do this, but the general modeling approach would need some thought. The main issue may just be that you’d need relationships between everything an the Page/Domain types so you can surface it contextually.

1 Like

Sounds a bit like polymorphic relations
:slight_smile:

1 Like

Always glad to see any bump of Polymorphic, although I know you @Chr1sG , @Oshyan and many other regulars are fully behind this. @rothnic hoping you will join our ranks with unfettered support fo this soon, maybe we can move the needle! The lack of Polymorphic continues to plague my team daily in multiple ways!

2 Likes

To be fair, polymorphism would be useful for me, but it’s low down my wish-list :-/

3 Likes

As I told @Polina_Zenevich and @vadim, polymorphism is high on my wish-list. Otherwise it is hard to use Fibery on a large scale.

@rothnic just a url field could do the trick to start for 1. and 2. . Basically what I want is not only have a “write” extension but also a “read” feature that display the knowledge in fibery on the page I’m browsing.
A bit like annotation tools like Hypothes.is or Diigo but with organized data instead of lousy annotations.

For 3. it has to by a dynamic matching, since the point is not to link a page/url with an entity in Fibery, but match words in the page being seen with entities in our knowledge base. So let’s say I’m viewing a Bloomberg article that discuss the new service launched by Company X, I would like to see “Company X” highlighted an on hover/click see information from my relevant Fibery entities. And then I can click to go to Fibery and start exploring more related entities if needed. I would be very powerful to share knowledge with new joiners, and also faster to add new information to existing entities.

1 Like

Hey @JMaynier good to see you would highly prioritize Polymorphic, it remains about at the top of my list. @mdubakov has also posted (I can’t locate it now) saying he didn’t think it was a lot of work for the team to implement, just not prioritized.

In order to help the cause, could you go ahead and “heart” the original post here? I saw you haven’t done so yet. We need all the support we can get!

Thanks!

1 Like

Yeah, the issue is just what terms to search for. If using text, then you are forced into a pretty difficult search problem. However, if you utilize semi-structured or structured data, like URLs, page metadata, and Schema.org data, then you have something more concrete to utilize. An article about some company is likely to link to that company at some point.

I’ve had needs come up as well. Without it, you are forced to try to fit anything that is similar into one Type or setting up a ton of relationships between Types

4 Likes

Ok good to hear from you @rothnic in this, and what you stated above is a very simple, precise way to describe one of the biggest reasons Polymorphic would be a huge help in day-to-day in Fibery. Thanks!

1 Like

That is a really cool concept, but I suspect would be difficult and resource-intensive to implement. There are definitely companies trying to tackle that kind of thing, but they seem to be more dedicated to that area of functionality (e.g. “match saved content to active browsing content”). I don’t know if this is something Fibery can tackle directly, but perhaps it could come through other means, either via API, or external to Fibery, like a tool that can index and match content across any web-based system. Sort of similar to Promnesia, except with content matching I guess:

2 Likes

I did not know Promnesia, seems interesting, but clearly a solution for devs/hackers. I cannot ask my team to install and run a local server, crontab, etc.
Could be a first step, but I really believe that is should be a core feature.
My view is that Fibery needs to be good in the 3 following areas to be very effective in the context of a team of more than 100 people :

  • Knowledge management: rich linked entities and documents to hold our collective brain, for organized objects (Companies, Contacts, Features, User Story, Candidates, Competitor, etc) and less organize information (article, news, books) that you want to relate to the organized entity. That the core value of Fibery.

  • Tasks management: the thing is the tasks don’t exist out of nowhere, they are usually linked to the main entities of our domain (see above). Fibery lacks better tasks management features, Clickup better in that respect, like reminders (linked to notifications), recurring tasks, improve calendar views, maybe integrate with more personal todo list (so you can display on your personal todo all your tasks)

  • Enrich context in other tools: once you have lots of knowledge and tasks in Fibery, you want to have those accessible in your other daily tools and actions: when writing/receiving emails, when browsing news/potential new knowledge, when sourcing candidates, etc. Chrome extension is clearly the easiest/best way to go for that.

Beam app and other similar projects are taking the approach of backing that directly in the browser. There is a convergence.

I agree with you on all this in principle. And Fibery is directly aiming to tackle points 1 and 2. Those are both potentially quite complex on their own.

Yes, this is definitely something that would be ideal. I’m just saying this seems out of the scope of what Fibery has at least suggested they are trying to do, and in itself is also extremely challenging (as evidenced by the lack of any other good tools that actually manage it very well). I would love to have this, it’s definitely part of the “grand future ideal vision”, but I don’t see it as being something the team will have the resources to tackle any time soon.

That said, maybe someone can build on the API to attempt something like this…

2 Likes

Yeah, I agree. Not likely a core Fibery problem for any foreseeable future, but is definitely doable. I think a slightly simplified version of something like what Promnesia is doing wouldn’t be that difficult for a developer with time. I’m just not sure if you want every pageview and visit recorded in Fibery :slight_smile: . The hard part for me was just getting the basic extension framework setup. Otherwise, it would just be doing basic CRUD stuff through the API.

I had to stop working on the fibery extension I was playing around with so I could get some other things done, but could demonstrate something like that when I get a little more time on it. I was already saving the URL out in a separate field when adding the clipped page. You just then need to as a starting point lookup anything associated with that domain or URL. Later on, you could leverage other data from the page, but URLs are a good starting point.

2 Likes

Yes, I was actually thinking of you and your Chrome extension in what I wrote. :grin:

@rothnic Maybe you could share your repo so other could contribute ? I could have a look (but time is hard to find!)

1 Like

Yeah, that is where I was at last with it was cleaning up the repo so it would be a good starting point. I had a bunch of messy paths I went down to get it all working. I am trying to get Fibery in a usable starting point for broader use in our org, so had to put that on the backburner for a bit.

I just shared the unofficial js library modifications I made as a starting point.

2 Likes

This reflection on the complexity and resources needed to build something like this was my first thought when reading the post.
I like the idea however, this is an entire product in and of itself. It seems like a great opportunity for a 3rd party ecosystem.

1 Like