antonyms in member class

[jeffrschneider]

Could someone clarify the use of antonyms, opposites or converses inside a member class, or a sub-class? For example, psych-verbs-31.2 {admire, adore, ...} + {detest, distrust, ...}

[amosleokim]

VerbNet is based on Levin-style classes, where the driving intuition is that "the range of syntactic variations for a class of verbs is a reflection of the underlying semantics" (Palmer). Put simply, verb senses that share an identical "range of syntactic variations" tend to also share some underlying semantic content. The semantics captured by the class you've given as an example are working exactly as intended based on this mechanism. That is, "synonymy/antonymy" aren't really ideas that are intended to be explicitly represented in VerbNet. Please refer to the background literature under the Publications tab at [https://verbs.colorado.edu/verbnet/] for a more rigorous explanation.

[MarthaSPalmer]

Having said that, we are in the process of adding verb-specfic features to class-members to capture exactly these type of systematic semantic differences. Haven’t gotten to this class yet.

Martha

On Aug 15, 2018, at 4:22 PM, A. Leo Kim notifications@github.com<mailto:notifications@github.com> wrote:

VerbNet is based on Levin-style classes, where the driving intuition is that "the range of syntactic variations for a class of verbs is a reflection of the underlying semantics" (Palmer). Put simply, verb senses that share an identical "range of syntactic variations" tend to also share some underlying semantic content. The semantics captured by the class you've given as an example are working exactly as intended based on this mechanism. That is, "synonymy/antonymy" aren't really ideas that are intended to be explicitly represented in VerbNet. Please refer to the background literature under the Publications tab at [https://verbs.colorado.edu/verbnet/] for a more rigorous explanation.

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHubhttps://github.com/kevincstowe/verbnet/issues/14#issuecomment-413357572, or mute the threadhttps://github.com/notifications/unsubscribe-auth/AEO9fiqRPm2-kcK8ADDvsFtPT9DU-O2Nks5uRJ8jgaJpZM4V-uuZ.

[jeffrschneider]

Thank you.

I'm heading down a similar path: adding semantic features to the VerbNet verbs. Although there are many use-cases for VerbNet, two prominent cases are: 1. Word Sense Disambiguation, 2. Semantic distinction and similarity. I'm interested in the latter.

In general, using attributes to describe semantic content, scales, valences, etc. is fairly straightforward. However, in VerbNet it is unclear which attributes are common between all members in a class. For example, some semantic attributes have been documented for run-51.3.2. One might assume that the all the members in the class share some semantic content, e.g., {living-entity-location-movement + pedal}, and from that base, additional properties are accumulated:

This begs the question, how are the base (common) attributes of a class documented? If the answer is, they're not, then I'd question the value of adding partial semantic content to an unknown base. If the answer is, each class will get a set of common attributes, then I'd suggest that you will need a standardized manner to identify concepts such as negation, converses, reverses & scale.

[MarthaSPalmer]

The common attributes are captured by the class level semantic predicate argument representations, as described in the publications.

Martha

On Aug 16, 2018, at 8:10 AM, Jeff Schneider notifications@github.com<mailto:notifications@github.com> wrote:

Thank you.

I'm heading down a similar path: adding semantic features to the VerbNet verbs. Although there are many use-cases for VerbNet, two prominent cases are: 1. Word Sense Disambiguation, 2. Semantic distinction and similarity. I'm interested in the latter.

In general, using attributes to describe semantic content, scales, valences, etc. is fairly straightforward. However, in VerbNet it is unclear which attributes are common between all members in a class. For example, some semantic attributes have been documented for run-51.3.2. One might assume that the all the members in the class share some semantic content, e.g., {living-entity-location-movement + pedal}, and from that base, additional properties are accumulated: [image]https://user-images.githubusercontent.com/1353365/44212331-a98ea480-a130-11e8-9966-c0609397a04a.png

This begs the question, how are the base (common) attributes of a class documented? If the answer is, they're not, then I'd question the value of adding partial semantic content to an unknown base. If the answer is, each class will get a set of common attributes, then I'd suggest that you will need a standardized manner to identify concepts such as negation, converses, reverses & scale.

— You are receiving this because you commented. Reply to this email directly, view it on GitHubhttps://github.com/kevincstowe/verbnet/issues/14#issuecomment-413558690, or mute the threadhttps://github.com/notifications/unsubscribe-auth/AEO9flXSjGm7XMkHJWC9nXBy5PF7yZkHks5uRX1fgaJpZM4V-uuZ.

[jeffrschneider]

Sorry, I remain confused. What might be helpful is to show an example of the 'class level semantic predicate argument representations' for "run-51.3.2".

(I've read the publications, the annotators guide, the Levin book, etc - yet this eludes me)

On Fri, Aug 17, 2018 at 8:56 AM MarthaSPalmer notifications@github.com wrote:

The common attributes are captured by the class level semantic predicate argument representations, as described in the publications.

Martha

On Aug 16, 2018, at 8:10 AM, Jeff Schneider <notifications@github.com mailto:notifications@github.com> wrote:

Thank you.

I'm heading down a similar path: adding semantic features to the VerbNet verbs. Although there are many use-cases for VerbNet, two prominent cases are: 1. Word Sense Disambiguation, 2. Semantic distinction and similarity. I'm interested in the latter.

In general, using attributes to describe semantic content, scales, valences, etc. is fairly straightforward. However, in VerbNet it is unclear which attributes are common between all members in a class. For example, some semantic attributes have been documented for run-51.3.2. One might assume that the all the members in the class share some semantic content, e.g., {living-entity-location-movement + pedal}, and from that base, additional properties are accumulated: [image]< https://user-images.githubusercontent.com/1353365/44212331-a98ea480-a130-11e8-9966-c0609397a04a.png>

This begs the question, how are the base (common) attributes of a class documented? If the answer is, they're not, then I'd question the value of adding partial semantic content to an unknown base. If the answer is, each class will get a set of common attributes, then I'd suggest that you will need a standardized manner to identify concepts such as negation, converses, reverses & scale.

— You are receiving this because you commented. Reply to this email directly, view it on GitHub< https://github.com/kevincstowe/verbnet/issues/14#issuecomment-413558690>, or mute the thread< https://github.com/notifications/unsubscribe-auth/AEO9flXSjGm7XMkHJWC9nXBy5PF7yZkHks5uRX1fgaJpZM4V-uuZ>.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/kevincstowe/verbnet/issues/14#issuecomment-413873967, or mute the thread https://github.com/notifications/unsubscribe-auth/ABSmleBZ242f7tscKa8CBX33Z5UmmNWTks5uRsuAgaJpZM4V-uuZ .

[MarthaSPalmer]

They’re on this web page.

https://verbs.colorado.edu/verb-index/vn/run-51.3.2.php

Martha

On Aug 17, 2018, at 8:48 AM, Jeff Schneider notifications@github.com<mailto:notifications@github.com> wrote:

Sorry, I remain confused. What might be helpful is to show an example of the 'class level semantic predicate argument representations' for "run-51.3.2".

(I've read the publications, the annotators guide, the Levin book, etc - yet this eludes me)

On Fri, Aug 17, 2018 at 8:56 AM MarthaSPalmer notifications@github.com<mailto:notifications@github.com> wrote:

The common attributes are captured by the class level semantic predicate argument representations, as described in the publications.

Martha

On Aug 16, 2018, at 8:10 AM, Jeff Schneider notifications@github.com<mailto:notifications@github.com mailto:notifications@github.com> wrote:

Thank you.

I'm heading down a similar path: adding semantic features to the VerbNet verbs. Although there are many use-cases for VerbNet, two prominent cases are: 1. Word Sense Disambiguation, 2. Semantic distinction and similarity. I'm interested in the latter.

In general, using attributes to describe semantic content, scales, valences, etc. is fairly straightforward. However, in VerbNet it is unclear which attributes are common between all members in a class. For example, some semantic attributes have been documented for run-51.3.2. One might assume that the all the members in the class share some semantic content, e.g., {living-entity-location-movement + pedal}, and from that base, additional properties are accumulated: [image]< https://user-images.githubusercontent.com/1353365/44212331-a98ea480-a130-11e8-9966-c0609397a04a.png>

This begs the question, how are the base (common) attributes of a class documented? If the answer is, they're not, then I'd question the value of adding partial semantic content to an unknown base. If the answer is, each class will get a set of common attributes, then I'd suggest that you will need a standardized manner to identify concepts such as negation, converses, reverses & scale.

— You are receiving this because you commented. Reply to this email directly, view it on GitHub< https://github.com/kevincstowe/verbnet/issues/14#issuecomment-413558690>, or mute the thread< https://github.com/notifications/unsubscribe-auth/AEO9flXSjGm7XMkHJWC9nXBy5PF7yZkHks5uRX1fgaJpZM4V-uuZ>.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/kevincstowe/verbnet/issues/14#issuecomment-413873967, or mute the thread https://github.com/notifications/unsubscribe-auth/ABSmleBZ242f7tscKa8CBX33Z5UmmNWTks5uRsuAgaJpZM4V-uuZ .

— You are receiving this because you commented. Reply to this email directly, view it on GitHubhttps://github.com/kevincstowe/verbnet/issues/14#issuecomment-413889852, or mute the threadhttps://github.com/notifications/unsubscribe-auth/AEO9fiw2RqYgnCogdcbHfnb41EJmJINpks5uRtfAgaJpZM4V-uuZ.