Comments on: Rolling out the Roles

By: Adding Your Language to Ubiquity Parser 2

Adding Your Language to Ubiquity Parser 2 — Wed, 29 Apr 2009 12:04:18 +0000

[…] property is the inventory of semantic roles and their corresponding delimiters. Each entry has a role from the inventory of semantic roles and a corresponding delimiter. Note that this mapping can be many-to-many, i.e., each role can have multiple po […]

By: A Demonstration of Ubiquity Parser 2

A Demonstration of Ubiquity Parser 2 — Fri, 24 Apr 2009 08:29:58 +0000

[…] Rolling out the Roles […]

By: Attachment Ambiguity—or—when is the gyudon cheap?

Attachment Ambiguity—or—when is the gyudon cheap? — Wed, 15 Apr 2009 06:21:04 +0000

[…] “April 15th” is an object argument, but it will only check its noun type, not put it in the correct semantic role (position). Perhaps this is something to think about in the […]

By: Zach Lym

Zach Lym — Tue, 14 Apr 2009 04:53:48 +0000

Why not just have share as the command with on being the modifier and auto suggest the social networking site based on history, cookies, usage, etc?

By: Gordon P. Hemsley

Gordon P. Hemsley — Mon, 13 Apr 2009 00:31:57 +0000

I think it's worth noting that I'm probably not the only one who would "save this page as PDF".

By: mitcho

mitcho — Fri, 10 Apr 2009 02:47:58 +0000

Thanks for your comments Walter. Healthy skepticism is welcome and encouraged!

Perhaps I didn't make the goals for this approach clear… although I've called these "semantic roles" there isn't any actual semantic analysis going on… these are just "roles" which tend to have clear morphosyntactic markings in various languages. For example, an "instrument" role is normally marked by something like English "with" or "using"… a "goal" is marked by "to" or "toward", etc, and these two will not get mixed up. In many other languages it would be this way. The goal is to make use of these abstract notions of "roles" of arguments in the commands to help semi-automatically translate the argument structure of commands. The original post which explains this approach can be found here.

I hope that helps explain what we're trying to do, but I don't think it's as big of an endeavor as the post would have you think.

By: Walter K

Walter K — Thu, 09 Apr 2009 20:45:46 +0000

Sounds like you are shooting for semantic language-dependent command parsing and full-blown inference engine... Is this the direction Ubiquity is taking? What happened to the initial idea of (fairly) short list of predictable commands that a user is able to master which speed up common everyday tasks? This new direction sounds like too big of a bite to take. I think many large companies spent decades of research on language processing and we have not seen a lot of results from that research yet.

Don't take this the wrong way. I'd like this new engine to work, but this sounds like a very large project with many unknowns. Language differences are enormous. For instance, valid English command constructs might not make sense in another language. But if it just works for English commands, it might be good enough for now!

By: Alejandro Moreno

Alejandro Moreno — Thu, 09 Apr 2009 17:31:54 +0000

"share-on-delicious as blah with tags x, y" could be used. Also "… using tags x, y". So maybe the tags could be an instrument? An instrument for categorization

share-on-delicious {object} {as:alias} {using, with:instrument} and clean arguments for instrument, removing "tags"

I don't know what I'm doing, but Ubiquity TNG is exciting!

By: Ran

Ran — Thu, 09 Apr 2009 08:21:08 +0000

For share-on-delicious, the title is the alias (name) and the tags are probably positions ("in category x"). If not, then tags probably are a role not yet specified, but they are not names.