Lately there where some blogs on Nepomuk and in the comment section there were also some critic points and I want to touch one myself here, the usability.
We have seen that Nepomuk can store (sic) lots of data, lots of “sentences”, like an object has a tag, a date or whatever. That is pretty nice as it enables to connect all that information together.
Problem
Now where I see the problem is that accessing all that information basically works the same way: nepomuksearch:/hasTag:OpenCL
That’s the problem! I do not care to enter these lines to learn the verbs, neither do other users especially if they do not know that they have to do it this way. And I’m not even sure if hasTag/tag is localised, if it is not yeah fun for the regions in the world where they use different letters/alphabets than that.
As you see the problem is not really Nepomuk, but rather the present ways to interact with it as a user.
Explaining what you want
As I’ve posted in a comment of a blog some weeks ago the way searching in my opinion should work is the way you explain something to others. You talk about “London”, your counterpart thinks you are talking about the city, so you say “No, not the city but …”
- “… a book by Jack London”
- “… our school colleague Jeffrey London”
- “… a person on irc with that nick”
- “… the paper I wrote on London’s homeless people”
- …
It is a trial and error, you throw one term and maybe your counterpart instantly knows what you are talking about and if not you have to be more specific, make connections etc.
The perfect search you should always avoid is “Let’s talk about Jeffrey London, you know who we went to school with [nice so far], who was born on 1/1/1970 in Bram, whose parents were …, whose RNA is.”
Proposal
But this is not to become a troll-entry, so what I propose – actually most likely not do, so drive-by-posting 😉 – is imitating and improving Gnome’s tracker search.
Tracker in Ubuntu 7.10 (two years old)
There you have one line where you enter whatever you want to enter, be it names, tags, parts of a document etc. When in their gui you have categories on the left, like “Documents”, “Images” etc., we should add Contacts, Mail, and other categories we find useful, but there shouldn’t be too many.
None has to care about the semantics of a search and as probably most people don’t have thousands of files with “London” somewhere they should find what they want easily.
If they don’t find what they want they can specify one of the categories I mentioned above, add another term to the search term, or in the KDE case they should also have an options-section where they can fine-tune the search:
- part of filename
- created (on|before|after|between)
- last viewed (on|before|after|between)
- created by (dropdown list with people who were found in the search as author etc.)
- etc.
And that’s not all, there could be an automatically created section with nepomuk categories that were found with this search like:
- Tag
- Contact
- …
essentially this could be merged with the categories mentioned above, though there should be a fallback e.g. if Strigi did not index a file-yet there should still be a category “Images” with it inside.
That should give them results very fast. The target should not be to create the “perfect search” where you get what you want on first try, but rather to have a good starting point where the user can narrow the results down if they want or need to.
GSOC-project
Yes I have read about the gsoc projects for nepomuk and like the changes on the search (I’m not sold on the loading/saving), though judging from screenshots (!) it still rather tries to implement a perfect search rather than adding means to improve the non-perfect result.
Backdraws
Yes with all that you would not be able to use all of Nepomuk features but you will be able to use it, everyone will be.
It would be a starting point where improvments could happen, there is no sense in waiting for KDE 4.4++ or whatever to have a nice desktop search, if the capabilities are there already.
Please make Nepomuk useable, you can make it perfect later on.
mat69's (mfuchs) Blog 
October 3, 2009 at 4:45 pm |
Currently, AFAICR, only Sebastian Trueg works more or less full time on Nepomuk. Help would be definitely welcome. You know the KDE philosophy…
October 3, 2009 at 5:02 pm |
Help is always welcome, nearly everywhere, though not everyone can help and or has the time. Nevertheless should that not stop them from posting their thoughts.
Imo the way you answered is using a killer phrase and not really talking on what I posted on. So if you want to comment on what I posted feel free to do so.
October 3, 2009 at 4:53 pm |
I have asked now a over year from the developers when we get usable UI for the nepomuk. Since first versions of KDE (4.0.y) I have tested and waited something to happend. On 4.2 was great to get it work with folder view, but was still very difficult to get.
And you are now doing actually the same problem by that screenshot from Ubuntu (theme) what the Nepomuk came to solve.
If I type “ubuntu” I do not want after that start searching groups on left side to specify something else. If I am searching a document what I wrote about Ubuntu, I want it to be found easily by asking “Ubuntu document”. Or if I want to find the email where the document was attachment, just saying “Ubuntu email with document” or “emailed ubuntu document”.
I feel the Gnomes Tracker very difficult to use. Just because it does not solve my problem, but just brings too much choises front of me.
And still the problem, what you mentioned too, is the language. I do not want to search with | by | from | yesterday or other words from different language. I want to use that language what user has set to desktop environments default.
And what is very missing feature still, is the Dolphin and save/open dialoges functions. I have tested the save/open dialoge and I must say it has some nice features, but it just push files to somewhere on my homefolder, without me knowing where. And then I should only use nepomuksearch:/ to find them.
October 3, 2009 at 5:44 pm |
What I meant was not that you have to use the categories but rather that you could. Personally I also prefer to type something than rather have to use the mouse — maybe should have emphasized that more above.
So if you preferred to enter “John mail” the result should be something useful. In fact in the example above it would also contains documents that have “John” and “mail” inside them, not sure if that could be solved with a ranking system.
The Gnome Tracker was an example and it is by far not perfect — that is why they came up with diferent things like Zeitgeist — but you can use it. [1] That is also why I meant to improve its concepts.
Still the best backend is of no use if you don’t use it for whatever reason.
[1] In my experience the backdraw of it actually is that there are so many results that contain the phrase you were searching in their body.
October 3, 2009 at 5:11 pm |
@mat69 – My remark was mostly on addressing why it’s not progressing as it should. Secondly, I second some of Fri13’s concerns (it’s the same in nepomuksearch:// if you omit the tag: or hasTag: part) – and using the categories would be – IMO -awkward.
The best would be some natural language processing (NLP) thrown in, but you know the difficulty of handling that over *all* the languages KDE is translated in.
What would help Nepomuk being more used and tested, IMO, would be moving the nepomuk-kde component out of playground and somewhere else, in the main KDE modules or in extragear: it adds a LOT of functionality, but few distros package it (IIRC, Mandriva).
BTW, Mandriva is actually pushing Nepomuk quite a bit: take a look at http://doc4.mandriva.org/bin/view/labs/Nepomuk-mdv2010-RC (currently does not load, sadly).
October 3, 2009 at 5:14 pm |
There are really two different useful styles for search, both of which I’d like to see in KDE and neither of which are revolutionary (though there is no reason KDE shouldn’t go beyond what already exists)
(1) A simple input field with an *incremental* search (search as you type) with some sort of sensible preview of the information. This should “just work” so if you type “food” documents tagged “food” or with names containing “food” should be near the top and documents indexed as containing the word “food” should be listed lower down. This could be enhanced by allowing OPTIONAL specifiers and syntax such as “hastTag:fruit AND NOT hasTag:apple”
(2) A graphical compound query creator. eg like:
(this could be translated into a string with proper search syntax to allow the user to learn how to do complex searches without all the drop-down boxes)
It would be great if the two of these could be one interface.
October 3, 2009 at 5:21 pm |
I have been using nepomuksearch:/ quite a lot lately and IMHO it lacks search in the results.
October 3, 2009 at 5:34 pm |
there’s a reason interfaces like tracker haven’t taken off despite being available for years and years now. they aren’t what people want. people don’t want a search tool. they want the environment and their apps to show them relevant information. it’s a fine difference. so while a tracker style interface would be nice and would make the integration into dolphin nice, it really shouldn’t ever be the ultimate goal imho.
October 3, 2009 at 5:52 pm |
No, in fact it should not be the ultimate goal but a starting point. And so far I think that the interface is easy to understand and use, the results might not be ideal or outright bad in some cases but none said that we should copy the results.
I think this interface makes it easy — with extensions above — to graphically narrow down the search results if needed. It is also an representation of the results, so like when you search for a tag [1] you might not have known that you also tagged a video with this tag, though with the categories you know that instantly.
Though I have to disagree partially as in many occasions I want a search tool.
[1]imagine there was an auto-competition that offered tag (if that is possible as in fast enough)
October 3, 2009 at 6:56 pm
as i said, while such an interface would be nice for many, focusing on such a thing as a “starting point” or even as what’s obviously needed to get nepomuk further along is, imho, a mistake. it’s a nice feature to have on the side and maybe one of the people commenting here (or writing this blog enter? 😉 would like to write such a thing (99% of the work is already done for you).
it won’t, however, make nepomuk “take off” or even get it to a sizable number of users out there. if that was the case, then these beagle / tracker type interfaces would be what everyone uses all the time.
they are absolutely useful for specific use cases, but it’s not even near what could make it real world usable to the majority of people.
spotlight from Apple gets a bit closer, btw, and purely through interface choices.
btw, have any of you tried using krunner with the nepomuk (desktop search) runner? 🙂
October 3, 2009 at 7:37 pm
Aaron, I have used the nepomuk runner and it does work, which is great. It works well for a quick search, if I know exactly what I am looking for and if there are few hits. However, I have thousands (I am not exaggerating) of PDFs and ODFs and the searches I make return way too many results for KRunner to display them.
Worse, I had to turn off Strigi indexing altogether as I complain in my posting below, since its heavy i/o usage prevents me from using my computer in any meaningful way. I have a modern system with a dual-core cpu and lots of ram, but still need to wait sometimes for minutes because my computer becomes sluggish with Strigi enabled. 😦
Anyways, thank you for your awesome work, it is really, really appreciated. 🙂
October 5, 2009 at 3:56 pm |
OSX integrates spotlight very well. That should be KDEs _starting_ point.
And no, people want search. They just want it integrated into the file dialog etc.
October 3, 2009 at 6:19 pm |
Let me start out by saying that I understand the promises of Nepomuk and believe in them. I devour every news about the project and check out playground from time to time, just to see how it is going.
However, while I agree that the tracker search gui is not perfect, it does work. Searches show results based on all kinds of information (tags, file names, file content), which is not perfect, but useable. Being able to refine it by choosing categories is a nice and helpful addition. More options would be great, sure! But so far it is more powerful than the nepomuk search in Dolphin. Also, the tracker daemon is more kind to my harddrive (less i/o), while Strigi keeps blocking my computer if I copy lots of large files or right after KDE boots up. Basically, the Strigi file indexer is not useable for me right now, seriously crippling nepomuk. For this issue, see https://bugs.kde.org/show_bug.cgi?id=196402
I am a big fan of what KDE is doing in this area, but, unfortunately, so far tracker is more helpful to me.
What we need is a) a better search gui and b) better i/o management for Strigi.
October 3, 2009 at 7:02 pm |
I have to agree with mutlu and mat69.
I am very interested in nepomuk and technologies like akonadi. I thing those things will defintely improve our workflow.
However, normally I am a average user, I guess. While working I do not care about the technology behind. I just want a working and fast way to to search for a certain file or information. I just want to put in a name or a simple search phrase like “mail john” as stated above or even just parts of it like “mail jo”. And I want to appear the results instantly.
And maybe because I like spotlight, I want my search to be up to date, so it has to run in the background as a daemon or such. This without interfering my workflow.
At the moment, KDE doesn’t offer any usable way of searching for files or searching files itself. Nepomuk isn’t very integrated in the Desktop ATM a. I don’t want to type “nepomuksearch” in front of every seach I do. Also Strigi is often consuming many CPU cycles and is very harsh to my harddrive in terms of disk usage. So this combination just doesn’t work as it should now. And when I want to verify if I am the only one with a certain problem in IRC I usual get the answer of “this just doesn’t work at the moment, sorry”.
I personally think the first step is to get a usable file searching tool first. I don’t want to care about Redland or sesame or a buggy strigiclient. I just want to open Krunner or click on the binokular in my panel and enter the search phrase – whatever this might be. Tracker is far from beeing perfect, but it is at least somehow usable. So mat69 is right when he says: make nepomuk usable and care about making it perfect later on.
October 3, 2009 at 10:04 pm |
well, I guess the question then becomes how willing are you to port Kerry to use Nepomuk? 🙂
I mean, it’s there, it’s stable, it is KDE-based and it does what you (and me fwiw) seem to want from desktop search…
also another tool that could be useful would be an “explorer” of the information stored in one’s machine in the form of a moving around a (3D?) graph. This has the advantage of browsing the existing information (aiding discoverability and not requiring to remember stuff) and showing things that may be related but is not immediately apparent.
October 4, 2009 at 9:29 am |
I completely have forgotten Kerry, it’s ages when I used it the last time, but I won’t be the one porting it for several reasons — time amongst them, fear of ending up as maintainer.
Anyway it is great that you commented and pointed me to Kerry again!
In general it is fantastic that I got so many responses with ideas on this topic, even if I won’t realise them (at least not most) they widened my perspective on the whole topic and maybe someone else has also good use for them.
October 5, 2009 at 6:44 pm
well, I wasn’t directly referring to you in order to port Kerry, it was more of a rhetorical question 🙂 But the issue remains: KDE already has a desktop search frontend, designed for another backend. Is it very hard to port it to use Nepomuk? Obviously we shouldn’t treat it as the final word on search, but atm we are really badly missing the functionality!
October 5, 2009 at 8:27 pm
Have a look at http://websvn.kde.org/trunk/playground/base/nepomuk-kde/knepsearchclient/
October 4, 2009 at 1:23 am |
I think a more natural way to search for things is using textual based interfaces, like Ubiquity in Firefox or Enso in Windows. The great advantage of this approach is that it uses verbs to express the main action, and imho that’s the way we humans try to interact with computers: “google for firefox” or “find electron in the wikipedia” or “mail Jhon with my last report”, things like those. In some ways this has been compared to a revival of the cli design but with a human design suited for the modern times.
The disadvantage of the gui’s is that they force you to focus on the objects. After you find the object you are looking for then you perform the action you were trying to do from the very beginning. But while you find the object you probably get distracted and start to do something else. I think that’s the reason the Enso devs forced you to maintain the Caps key pressed while you were doing some action.
I really would like to see some of the nepomuk metadata exposed through an interface like Enso. I think it would be extremely easy to use and very intuitive. The problem with krunner is that it’s the main door to a hell lot of things, from konqueror and konsole profiles to window management and finding your files with strigi. So again you cannot specify the action you want krunner execute and you have to find the object you want between the multiple suggestions it brings. That’s why I only have activated a couple of krunner plugins.
Enso is open source now, so probably it could be adapted to be used in KDE in the not so distant future, although its distributed under a BSD license for now.
Just my two cents.
October 4, 2009 at 7:28 am |
Firstly, sorry for my English.
I would asks about Nepomuk/Dolphin integration. Why don’t you create filter pane/sidebar for Dolphin? We can have there treeview of Nepomuk topologies, but limited to files inside current directory.
I thought also about path edit(above Dolphin window). It can been related too. We can drag filters from filter into path. Difference is simple – when we change directory(clicking on icon, the path edit is reset and sets to path to selected folder, but filters in filter pane aren’t reset). Additionally there’s usability difference – we can click only on “+”, “-” to add/remove search criteria. Treeview would been longer, so it’s not too quick, but we can show in them more information.
I think path to directory can be represented as search criteria.
October 4, 2009 at 9:35 am |
Nice idea! I’m not familiar with the dolphin code but maybe I’ll give it a try. 🙂
October 4, 2009 at 9:37 pm
If you would like to add a sidebar to Dophin to browse ontologies and to create a more powerful search function, you should have a look at at least two of the nepomuk programs in playrgound: “Nepomukshell” and the “Nepomuk Search Client”. They are in http://websvn.kde.org/trunk/playground/base/nepomuk-kde and part of the Mandriva Beta CD, see http://doc4.mandriva.org/bin/view/labs/Nepomuk-mdv2010-RC
Obviously, some things are happening, and we would all appreciate if you helped Trüg in his efforts.
October 4, 2009 at 3:26 pm |
“Why don’t you create filter pane/sidebar for Dolphin?”
I have dreamed about sidepanel for dolphin now a over year. It should be used to search and filter data.
There just place tha + and – signs to add something perimeters, calender view to drag a date area (multiple directions and dates) and mayby even later a map like on digiKam where you can just drag a box to select country or area.
If later we are having GPS’s on laptops, it would be awesome to get such information to stored on files as well when saving or opening them. Then you could even just find data by remembering “I was on Zürich on that time when I edited that file”.
October 5, 2009 at 7:02 am
Somebody have suggested to integrate Plasma with Dolphin. It’s not good idea, in my opinion, but it allows to better select criteria. Simply select search by date/time and Plasma calendar would appear. We can also drag date from system calendar. It would be nice feature to use whole desktop for selecting search criteria. This makes search more related to our current work.
October 5, 2009 at 9:31 am
I would not take the plasma calender to pop-up when selecting “date/time” search function. User would need to move mouse from dolphin to calender view and it would not be understand right away. The calender view should be already on the search itself. Isn’t it possible to do without plasma at all? I can not see why plasma is needed for that anyway.
November 5, 2009 at 4:25 pm |
I think that A sidebar would be less useful than having instant virtual folders in the Dolphin interface much like Audiocd:/ does. You could have the top 10 items not placed in any directories so you can easily see/preview them and anything that is deeper is automatically dropped into a directory so you can quickly jump to contacts or Time to refine your search visually
October 6, 2009 at 12:36 am |
I’ve been looking forward to having the much vaunted Nepomuk usable since 4.0.
Its been years now and there is still little to show for one of the major selling points of the KDE4 series – at least to the end user, which is where it really matters.
March 19, 2014 at 4:56 pm |
Many thanks for sharing this story. I’m definitely tired of struggling
to find relevant and smart commentary on this subject.
Everyone these days seem to go to extremes to either drive home their viewpoint or suggest that everyone else in the globe is wrong.
thank you for your brief as well as relevant information.
May 3, 2014 at 2:32 pm |
After exploring a few of the articles on your website,
I truly appreciate your technique of blogging. I saved as a favorite it to my bookmark
website list and will be checking back in the near future.
Please check out my website as well and let me know how you feel.
November 22, 2014 at 1:16 am |
Еxcellent waу of describing, and nice post to get facts cοncerning my presentation topic, which i am going tο present in college.
November 24, 2014 at 10:27 am |
Thanks fߋr every other informative blog. Where else could I ǥet that kind of info written in such an ideal approɑϲҺ?
I have a prօject thzt I am simply now operating on, and I’ve
been at the looҝ out for such information.
March 11, 2016 at 11:27 am |
Hello! I’ve been reading your website for a while now
and finally got the courage to go ahead and give you a shout out from Lubbock Tx!
Just wanted to mention keep up the good work!