About SEMAFOR

[fidel Aseh]

 Hi Guys, 

I am presently trying to manually annotate over 1000 sentences using FrameNet frames; I mostly use LU's (lexical units in FrameNet) to label the frames but this is quite tough and confusing; some LU are targets to more than one frame and it's hard to manually determine which frame that will be. Any suggestion from anyone please?

[Nathan Schneider]

Hi,

You are not alone in appreciating the difficulty of FrameNet-style annotation. In my view, the richness and complexity of FrameNet is part of what makes it interesting and potentially very powerful for applications. But it also makes it intimidating.

I have never attempted FrameNet annotation myself. But I have talked to the Berkeley FrameNet creators about their annotation practices, and it is quite an intense undertaking. The way they approach annotation there, it involves several weeks of training people who have already studied syntax. The training covers much of the material documented in the 130-page reference manual (https://framenet2.icsi.berkeley.edu/docs/r1.7/book.pdf) on the syntactic and semantic criteria for LUs and FEs. Even after that, it is not hard to find errors in the annotations. Their process for creating new frames, and to a lesser extent adding new LUs to existing frames, requires an order of magnitude more expertise and consideration.

Of course, it's possible that parts of the annotation process could be simplified. The annotation process used by Johannsen et al. in http://aclweb.org/anthology/D15-1245 was apparently simpler, though it is unclear what this means for the quality of their data relative to data produced at Berkeley. (Some of the FrameNets in other languages also have adapted the annotation process to their needs, but I take it that you're interested in annotating English.)

Best,

Nathan

On Fri, Apr 7, 2017 at 12:05 AM, fidel Aseh <asfid...@gmail.com> wrote:

 Hi Guys, 

I am presently trying to manually annotate over 1000 sentences using FrameNet frames; I mostly use LU's (lexical units in FrameNet) to label the frames but this is quite tough and confusing; some LU are targets to more than one frame and it's hard to manually determine which frame that will be. Any suggestion from anyone please?

--
You received this message because you are subscribed to the Google Groups "semafor-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to semafor-users+unsubscribe@googlegroups.com.
To post to this group, send email to semafo...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/semafor-users/ad794cbb-bbcb-42e6-b8dd-3cc548e2af53%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

--

Nathan Schneider

Assistant Professor of Linguistics and Computer Science, Georgetown University

http://nathan.cl