This folder contains the Verbs With Attributes dataset, as well as some other goodies (namely, definitions for all of the verbs).
This is the data split that I used for my experiments. Although we collected annotations on the verb template level (consisting of a verb + an optional particle, ex. "put on", "put off", "put"), we split the dataset on the verb level.
The file attributes_split.csv
has an additional column, in_imsitu
which is True if that verb is in the imSitu database. To get the imSitu index, filter out all rows where in_imsitu is false, and then use the corresponding 0-index on the verbs. (If you don't want to do the filtering, you can also use the file imsitu_verbs.txt
).
This file contains definitions for our verb templates that I scraped from wordnik. I filtered out all definitions that weren't labeled as verbs, so you
can pretty much ignore the POS
column. Note that some verb templates don't have definitions (so you'll have to backoff to verbs in those cases).
This contains all of the attributes we mined.
intrans
: 1 if the verb can be used intransitively, 0 otherwisetrans_pers
: 1 if the verb can be used in the form " someone", 0 otherwisetrans_obj
: 1 if the verb can be used in the form " something", 0 otherwiseatomicity
: Contains annotation of verbal aspect.- 0: Accomplishment
- 1: unclear without context
- 2: Activity
- 3: Achievement
- 4: State
energy
: How much motion does the verb require- 0: unclear without context
- 1: No motion
- 2: Low motion
- 3: Medium motion
- 4: High motion
solitary
: Whether the verb tends to be done in a solitary or social context- 0: solitary
- 1: likely solitary
- 2: solitary or social
- 3: likely social
- 4: social
time
: how much time it takes to do that action- 0: N/A
- 1: On the order of seconds
- 2: On the order of minutes
- 3: On the order of hours
- 4: On the order of days
bodyparts_Arms
: 1 if arms are usedbodyparts_Head
: 1 if head is usedbodyparts_Legs
: 1 if legs are usedbodyparts_Torso
: 1 if torso is usedbodyparts_other
: 1 if another body part is usedemot_anger
: 1 if verb is highly associated with angeremot_another
: 1 if verb is highly associated with another emotionemot_fear
: 1 if verb is highly associated with fearemot_happy
: 1 if verb is highly associated with happinessemot_sad
: 1 if verb is highly associated with sadnessemot_surp
: 1 if verb is highly associated with surprisalintrans_effect_0
: 1 if the subject moves somewhereintrans_effect_1
: 1 if the external world changesintrans_effect_2
: 1 if the subject's state changesintrans_effect_3
: 1 if nothing changestrans_obj_effect_0
: 1 if the object (a thing) moves somewheretrans_obj_effect_1
: 1 if the external world changestrans_obj_effect_2
: 1 if the object's state changestrans_obj_effect_3
: 1 if nothing changestrans_pers_effect_0
: 1 if the object (a person) moves somewheretrans_pers_effect_1
: 1 if the external world changestrans_pers_effect_2
: 1 if the object's state changestrans_pers_effect_3
: 1 if nothing changes