Howdy, Stranger!

It looks like you're new here. If you want to get involved, click one of these buttons!


edited February 2016 in Features

Bearing in mind that your motto is "Remember your stuff" and that presumably people will want to remember field-based stuff like contacts, will you be providing a way of having templates? This happens to be well-suited to JSON and when people import or export data then having templates would presumably work very well. Additionally, it would be a shame if one could not retain the field type when exporting data, and field types might help in the future when the application becomes more intelligent with searching using voice commands, such as "What street does John live in?". If John lived in a street with a name like 'Gasslands' then without a field association the app would not know that this word was a street name.

Incidently, one of the things that I have been looking for is a very simple customer relationship management application which would be a little more than an extended contact system to allow me to take the details of customers at shows and to record certain types of information about them like 'Last Contact Date'. The trouble is that all of the apps I have seen are desperate to grab your data and keep it in a way that prevents you from getting at it easily and of course the fact that they have the data means that it's not very secure. Being able to complete a form when entering a note would allow me to create my own CRM system and keep my data safe.

I guess what I am coming round to saying is that I think it would be great if one could define a note type and that that note type would have specific fields. When one was entering a note that was formatted, one would be prompted to enter specific fields.

Irrespective of whether or not you decide to employ templates, would you envisage every address to be a separate note or would you expect all addresses to appear in the same note?


  • This is a very interesting idea.

    Many other note taking apps have the concept of a note type. You choose the type and, from that, you have an automatic template so you can fill in the fields. There's usually also a general type, which is an unstructured note. I don't know if you can define new structures in these other apps, maybe you can in a few.

    Whatever I do, i have to reconcile it with the AI part of the project that is planned after the desktop version. I've not talked much about this yet, but the idea is that the AI will automatically label concepts within notes and use this meta data for search and collation.

    To that end, the note itself need not have any special structure. For example, dates and street addresses could be parsed so that their component parts be identified without having specific fields predefined in the note.

    Notes and documents like to acquire formatting of some kind. Most systems like to embed the formatting within the document (eg HTML). My plan is to separate all meta data into, conceptually, a separate document. The meta data could be semantic (such as template labels) or graphic (visual formatting such as font size, bold, colour etc).

    I think this idea is quite novel and has several advantages, the main one being a solution to semantic markup.

    Getting back to the idea of templates, instead of defining structure, you'd define a "how to recognise" semantic fields. Not sure how this will manifest as yet. Could be access to a rules engine or something funkier.

  • edited February 2016

    I agree that it would be important for concepts to be labelled and I like the idea that there need not be any special structure to a note, provided those concepts 'tags' can be attached to each item.

    However, I worry about the success of automatic parsing because an item may not be easily recognisable, that's why I picked that street name. If everything was easy, e.g. an address like:

    2, Acacia Avenue,
    AB1 2CD

    ...then the parsing might be easy. But if it was:

    Near Norwich

    ...then it might not be recognised as an address as easily, just a list of names. Not sure how to fix this.

    Is there a database somewhere which might attempt, or help the user to label words? For example, if you looked up "Florence", you would want something that told you that it was most likely to be a woman's name or the place in Italy but it might be other things. Potentially the system could offer suggestions and help you to label them.

    This all sounds a bit like Dr Chandra speaking to SAL in the movie 2010 - 1' 50" onward...

    When the system asks you "Will I dream?" then you'll kow you're getting somewhere :-)

  • Good clip. but I don't think machines will dream. I think dreaming is the process by which the human brain garbage collects things that are irrelevant. The notion that human memory is extensive but recall is limited is wrong. That would be inefficient and life is never inefficient.

    Machines will not dream because their garbage collection (ie the process of determining the differences between patterns), will happen concurrently with knowledge acquisition.

    It is only that way with humans because it's convenient to have this "offline phase" concurrent with physical rest and machine will not need physical rest.

    Back to your address example; the second case, although not a postal address, is nevertheless some kind of location. This is exactly why semantic markup will be superior, because it can accommodate incomplete or approximate facts.

    I have already the design for knowledge representation and the code to implement it. There are a few unfinished bits. But the plan will be to put a "version 1" of this into Lexiy once the desktop is done.

    Regarding words, yes a word dictionary will map onto an ontology. A given word can have multiple meanings and the semantics of any given word or sub-phrase is understood in the context of others around it. So when you say, "I'm going to Florence", it will know you mean the place. The same applies to the parsing of addresses, for example.

  • edited February 2016

    Whilst the omission of a postcode in that second example may mean that it isn't an officially complete postal address it is nevertheless a real one belonging to someone I know and letters sent to it do arrive properly - I only changed the house name on the first line for reasons of confidentiality. I think this is an example of why templates could be helpful to the system in that it would know what kind of data it was receiving. Placed into the context of using a voice activated wearable device, I wonder if it would be useful if there was a preliminary template command to tell the system what kind of data it was about to be given. For example you might say "take an address" or "reminder" or "recipe". The problem with this kind of thing of course is that the user has to remember the 'commands' so I'm not sure if it would be acceptable. It will be interesting to see how this tagging process works out in the application.

  • Imagine you could define a template in words. That would be the idea way. However, then it's not a "template" as such, but an explanation. That's how you would tell a person. There's no reason why it couldn't work like that with an app.

    Your voice example is a step in that direction.

  • I just wanted to make a note that the reason a template might be useful is because it will prompt the user in a way which will prevent them from forgetting to enter a particular field of data.

  • I do think prompting is a very important point. Suppose a template was defined by word rules, there would have to be a way to query the missing data. This is the same problem as a language ambiguity, you have to ask the user to add more information to make the meaning specific.

Sign In or Register to comment.