The lexicon harvested from
FrameNet 1.3 contains 2,527 orthographically distinct, single-word verbs, distributed over 486 senses. This incorporates 3,747 distinct verb-sense pairs, meaning that each single-word orthographic verb has an average of 1.48 senses.
The TRIPS lexicon (17-7-2008) contains 2,175 orthographically distinct, single-word verbs, distributed over 265 senses. This incorporates 2,208 distinct verb-sense pairs, meaning that each single-word orthographic verb has an average of 1.02 senses.
Of the 2,527 orthographically distinct, single-word verbs in the
FrameNet -harvested lexicon:
- 1,344 are also in the TRIPS lexicon
- 1,183 are NOT in the TRIPS lexicon (distributed over 342 senses, corresponding to 1,522 verb-sense pairs)
Of the 2,175 orthographically distinct, single-word verbs in the TRIPS lexicon:
- 1,344 are also in the FrameNet -harvested lexicon
- 831 are NOT in the FrameNet -harvested lexicon
The task now is to take each of the 1,522 verb-sense pairs in the
FrameNet -harvested lexicon which are NOT in the TRIPS lexicon, and assign it to a TRIPS sense.
I created a file (fn2trips-mappings.xml), from the
FrameNet -harvested lexicon and the TRIPS lexicon, with the following format:
frames
> frame* @name
> lf* @name
> intersection
> verb* @name
> complement
> verb @name
> others
> verb* @name -- these orthographic verbs do not appear in TRIPS at all
There are a total of 342 frames (i.e.
FrameNet semantic types) listed, each of which is invoked by at least one single-word verb which is NOT in the TRIPS lexicon (in the 'others' element). For each frame, each TRIPS semantic type (i.e. LF-type) which invokes at least one verb which is also invoked by the
FrameNet frame is listed along with:
- the intersection - those verbs which are common to both the FrameNet frame and the TRIPS LF-type
- the complement - those verbs which are in the TRIPS LF-type but NOT in the FrameNet frame
46 of the 342 frames have no LF-types specified, meaning that none of the verbs they are invoked by are in the TRIPS lexicon. The other 296 frames have the following numbers of candidate LF-types:
Frames LF-types
59 1
44 2
48 3
36 4
25 5
11 6
16 7
8 8
6 9
4 10
8 11
6 12
4 13
4 14
1 15
1 16
3 18
1 19
1 20
3 21
2 22
1 23
1 25
2 26
1 27
The frame which has 27 candidate LF-types is Bringing.
The 46
FrameNet frames which have no mappings at all to TRIPS LF-types (i.e. none of the verbs invoking them are in the TRIPS lexicon at all) involve 106 verb-sense pairs. We will have to ignore these verbs for the moment. I hence divided the mappings up into two files:
fn2trips-mappings-set1.xml [296 frames]
fn2trips-mappings-set2.xml [46 frames]
Relevance order of LF-types?
--
MarkMcConville - 29 Jul 2008