[Nepomuk] RFC: The grammar of the new Nepomuk query parser

Mario Fux KDE ML kde-ml at unormal.org
Tue Jun 4 10:32:08 UTC 2013


Am Mittwoch 29 Mai 2013, 14.06:37 schrieb Denis Steckelmacher:
> Hi,

Morning

> As my GSoC proposal, « A new query parser and auto-completed input field
> for Nepomuk » was accepted, I think is is time to discuss the grammar
> that will be used by this parser.
> 
> This mail proposes a grammar that is somewhat complex but should be very
> human-friendly. I tried to keep it implementable, and I hope it is, but
> future reflexion may remove some difficult parts of this grammar to make
> it more computer-friendly.
> 
> (this is a very long mail, sorry)

[snip]

> Ambiguity and syntax coloring
> -----------------------------
> 
> There is no way to syntax-highlight a text when it can have more than
> one meaning, except if the text can be colored, underlined in other
> colors, then rendered in an italic font, etc.
> 
> One idea could be to show an arbitrary highlighting, for example the one
> with the biggest tokens (so that the user sees that he or she has to put
> commas to end strings at the correct places). Another possibility is to
> show ambiguous parts grayed-out, with a tool-tip showing the ambiguous
> ASTs and how the user can disambiguate the query ("put quotes around the
> string" for instance).

When I thought myself about such a system I always thought it was a good idea 
to return to the user (additionally to the results) the extracted query (in 
natural language form).

This way s/he sees where the possible misinterpretation happened and s/he can 
change her/his query accordingly and try it again.

Although with your system the problem of ambiguity should/could be solved even 
before you get the results of the parsed query.

On the other hand to above feedback system could substitute the syntax 
highlighting and auto-completion (which are complex for natural language).

[snip]

Some years ago I got access to some semantic web/ontology/parser software in 
our CS institute:
https://files.ifi.uzh.ch/ddis/oldweb/ddis/research/talking-to-the-semantic-
web/ginseng/index.html

I even got the source code for some of these projects and can probably get you 
access too.

Thx for your work
Mario


More information about the Nepomuk mailing list