Perl 6 - the future is here, just unevenly distributed

IRC log for #marpa, 2016-02-02

| Channels | #marpa index | Today | | Search | Google Search | Plain-Text | summary

All times shown according to UTC.

Time Nick Message
05:03 koo7 joined #marpa
10:51 koo8 joined #marpa
12:02 kaare__ joined #marpa
12:35 btyler_ joined #marpa
16:21 koo8 joined #marpa
16:57 Idiosyncrat joined #marpa
17:27 Idiosyncrat joined #marpa
18:18 Idiosyncrat http://web.stanford.edu/group/cslipublications/csl​ipublications/koskenniemi-festschrift/11-nivre.pdf
18:18 Idiosyncrat http://www.linguistics.fi/julkai​sut/SKY2006_1/1.6.6.%20NIVRE.pdf
18:20 Idiosyncrat Two interesting papers, which explore the relationship between "grammar parsing" -- things like the Marpa approach == and ...
18:20 Idiosyncrat what Nivre calls "text parsing", which means the kind of parsing you have to do in NLP, but which is also ...
18:21 Idiosyncrat interesting in some other computer applications.
18:21 Idiosyncrat Nivre gives a nice theoretical structure for some of the ideas I have about directions going forward.
18:22 Idiosyncrat Nivre starts by pointing out that the two suggest very different approaches, but that in fact ...
18:23 Idiosyncrat grammar parsing (Marpa-like parsing) can be very useful in text parsing.
18:23 Idiosyncrat Perhaps looking at these papers will suggest applications for and extensions of Marpa.
18:25 Idiosyncrat Also, did someone on this list point me at a Nivre paper before?
18:26 Idiosyncrat I just ran across this, but I have the feeling I've seen something else by him.
18:31 Idiosyncrat AFK for now
18:34 koo8 joined #marpa
19:15 ernimril joined #marpa
19:58 VsyachePuz https://en.wikipedia.org/wiki/CiteSeer should help, it it's main purpose to find other works...
20:00 VsyachePuz i.e. http://citeseerx.ist.psu.edu/search?q​=author%3AJoakim+Nivre&sort=cite
20:20 koo8 joined #marpa
21:30 Idiosyncrat joined #marpa
21:31 Idiosyncrat VsyachePuz: actually the 2nd of the links is to his own sites list of his papers, which is even handier ...
21:31 Idiosyncrat I was being lazy and rather than read thru them all myself for what's relevant was hoping for hints from others on the channel ...
21:32 Idiosyncrat but read thru them all myself is what I might do someday.
21:45 VsyachePuz http://stp.lingfil.uu.se/~nivre/bibliography.html
21:47 VsyachePuz no references to nivre in my log since 2015-11-14, only last 24 hours
21:56 VsyachePuz MaltParser .nuget for IKVM - https://github.com/sergey-tihon/MaltParser.NET
22:01 choroba joined #marpa
22:04 VsyachePuz "The MaltParser is the implementation of all the algorithms published by Nivre et al. So whenever experimental data is reported for the Nivre algorithm(s), it is probably the result of running the MaltParser."
22:09 ernimril joined #marpa
22:29 Idiosyncrat Yes, I've backlogged and can't find anything that looks like a reference to Nivre's papers.
22:29 Idiosyncrat From time to time I do Google keyword searches and look at interesting-sounding PDF's that pop up ...
22:30 Idiosyncrat so that may be why I remembered his name.
22:34 choroba I worked at a NLP department several years ago. Nivre is a big name in NLP.
22:45 Idiosyncrat The two papers show a lot of deep thinking.  They make a lot of points that I've been trying to drive at, but nowhere near so coherently.  And they're short!
22:46 Idiosyncrat Just to be clear -- in the above I'm trying to say that I have not managed to be as coherent as Nivre has.
22:51 choroba but natural language parsing is much dirtier than formal grammars. It involves hacks and statistics :)
22:53 VsyachePuz in my impression, this is result of the following process: 1) create as much text as possible, explore all ambiguilities, explain and document them; 2) find key observations, make generalizations, check them against facts; 3) reduce the text as much as possible, to retain only the essential and new things.
22:53 VsyachePuz i see no way to jump directly to the point 3 without step 1 first
22:57 VsyachePuz ---
22:57 VsyachePuz the major russian search company (yandex.ru) uses GLR for parsing texts, not Earley
22:57 VsyachePuz https://github.com/yandex/tomita-parser/
22:58 VsyachePuz (it's C++ + lua, as in your religion)
23:00 choroba Where I worked, we used https://metacpan.org/pod/Treex
23:02 Idiosyncrat Btw, I've a degree in AI from 1978.  Some progress has been made, but for almost 4 decades amazingly little.
23:03 Idiosyncrat I mean, imagine how relevant what you were taught in 1978 would be if your degree was in EE?
23:03 Idiosyncrat Anyway, I thought Nivre framed the issues very, very well.
23:04 Idiosyncrat Both in showing the relevance of work of my type to NLP ...
23:04 Idiosyncrat but also it's limits.
23:19 VsyachePuz If I see 2 ISO standarts - one for EBNF, and one for CLI (common language infrastructure), how to create a new standart for mapping between them? they don't give details - http://www.iso.org/iso/home​/standards_development.htm
23:24 VsyachePuz http://www.iso.org/iso/how-to-write-standards.pdf
23:26 VsyachePuz they explains everything, just need to read

| Channels | #marpa index | Today | | Search | Google Search | Plain-Text | summary