;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; UNFINISHED DRAFT AS OF 2008-01-30T02:25 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; Maybe start with Use Case 1a... Alexander is compiling a list of emotion words and wants to annotate, for each word or multi-word expression, the emotional connotation assigned to it. In view of automatic emotion classification of texts, he is primarily interested in annotating the valence of the emotion (positive vs. negative), but needs a 'degree' value associated with the valence. In the future, he is hoping to use a more sophisticated model, so already now in addition to valence, he wants to annotate emotion categories (joy, sadness, surprise, ...), along with their intensities. However, given the fact that he is not a trained psychologist, he is uncertain which set of emotion categories to use. ...(The above is quoted from this at http://www.w3.org/2005/Incubator/emotion/XGR-emotion/#usecase1a) I also found a nice distillation of the use case here... http://www.w3.org/2005/Incubator/emotion/wiki/UseCases .... (ii) Requirements: scope of the emotion annotation: * refer to one or several words of text emotion description: * emotion dimension: valence dimension, a scale from -1 to 1 * emotion categories: set of labels, with intensity (from 0 to 1) other: * advice: which set of emotion categories should be used? ....Okay, from the get go, I will not (here at least) address this requirement... "advice: which set of emotion categories should be used?". ...Instead, I think it gives weight to the claim that we as emotion markup language designers be forced to pick one or another paradigm. ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; So, first of all one requirement is this... The language should not only annotate emergent emotions, i.e. emotions in the strong sense (such as anger, joy, sadness, fear, etc.), but also different types of emotion-related states. ...and a second requirement is this... The emotion markup should provide a way of indicating which of these (or similar) types of emotion-related/affective phenomena is being annotated. ...Okay now suppose I am a party line Sherer person completely in love with Scherer (2000). Well then I would want to use this taxonomy... AffectiveState Emotions Moods InterpersonalStance PreferenceOrAttitude AffectDisposition ...In owl that would look like so..... But, of course, the ontology of affect space is an area of hot debate. A fan of Douglas-Cowie et al. (2006) would insist on this taxonomy... AffectiveState Attitudes Established emotion EmergentEmotion-FullBlown EmergentEmotion-Suppressed Mood PartialEemotion-TopicShifting PartialEmotion-Ssimmering StanceTowardsPerson StanceTowardsObjectOrSituation InterpersonalBond AlteredStateOfArousal AlteredStateOfControl AlteredStateOfSeriousness EmotionlessState ....A third person might like this except for EmotionlessState. They may say it does not make sense to have a subclass of AffectiveState that is without affect. How in the world will we deal with all these differences of opinion? With changes in the field as affective science evolves....The key to me seems to be to (a) allow people to continue to debate what is the correct taxoomony and yet (b) still let people who need to use some term set at least make some headway and leverage *something* without having to wait for the people in (a) to reach conssensus. KEY DISCUSION TOPIC: One thing I feel I need input on to help me understand the requirements is to discuss is what is the groups opinion on this point? ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; Okay, I have provisionally assumed that allowing for multiple taxonomies (or even more complex -- multiple theories) is a requirement. So, how do we do this? Well, one idea would be to have separate namespaces. You could have annotation ontology in one namespace and then you would have a separate namespaces. . TECHNICAL QUESTION: Do the other annotation options (e.g. XML or RDF) allow for namespaces? That said, on a practical level I am not sure how much annotations will really buy us. From my moderate experience with protege, you have to do more work to make assertions across two namespaces. As a praticcal "brass tacks" point, when you have 100 bits of text to get annotated this can get really tedious fast. (Then again, maybe y'all will stay far afield from protege and instead write your own custom tools that will take care of the nitty gritty details of namespaces and such. TECHNICAL QUESTION: Any sense as to what kinds of tools might actually do that markup? Something pre-existing? protege? Something custom created?) Given the potential ugliness of namespaces perhaps we just need to have everything be in one namespace but use different owl files. If we chose this route it will seem important that each emotion class have a unique name. For example, if Scherer has a class called Happy and Cowie has a class called Happy, life will be a lot less confusing and error prone if we adopt a unique naming convention. E.g. can prefix each SCH2000 and each Douglas-Cowie term as DC2006 or something. THus we'd have SCH2000-Happy and DC2006-Happy. TECHNICAL QUESTION: Can we import between xml files? Can we import between rdf files? ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; So that is one set of issues to get lost in. Another set of issues have to do with the actual use cases. So, how to ontologize Alexander's use case? As a reminder these are his requirements: Quoting again... emotion description: * emotion dimension: valence dimension, a scale from -1 to 1 * emotion categories: set of labels, with intensity (from 0 to 1) ...So, first of all I am not crystal clear on what the above means. Suppose we had two labels on a single emotion description. Suppose these labels were "Happiness" and "Shame". Is that possible? I suppose this would be a case of mixed emotion, right? I suppose any time you have two or more labels you have a case of mixed emotion, right? Now, suppose the intensity of the hapiness lavbel was 0.9 and the intensity of the the shame label was 0.9. Then, what would the valence be -- 0 perhaps because Shame is negative and Happiness is postive? Maybe how to compute an overall valence for a mixed emotion is itself something that should be parameterizable because subject to years of debate and research. Yes no maybe? ;;;;;;;;;;;;;;;; At first blush, my approach is to have a representation of that object that we wish to makup. Let me consider a specific real world case that I hope is a decent exemplar of Alexander's use case? Specifically I consider the point in Macbeth where Lady M says "Out damn spot!" How might I go aobut ontologizing this. I will first do this diagramatically to give as easily grokable description. Then, hopefully, if time, will spell out the owl -- which I lament is not very human readable. Now, some diagramatic notation conventions: Slots or relations are have a name that begins with a lowercase character (e.g. annotationTextIs) Individuals (as opposed to classes) have a name that begins with a * followed by an uppercase character, e.g. *Alexander. Primitive Data types (e.g. strings, time instants, numbers, booleans) are put in quotations. *"Out damn spot!" ^ | annotationTextIs | *Annotation24601 | |--- annotationAuthor: *Alexander | |--- annotationTime: "2008-01-23T13:36" | |-- annotatedEmotion: *FearInstance35 | |--instanceOf: *DC-2006-Fear | |--valence: "-0.8" | |--