From jcowan@reutershealth.com Wed Mar 07 15:34:57 2001 Return-Path: X-Sender: jcowan@reutershealth.com X-Apparently-To: lojban@yahoogroups.com Received: (EGP: mail-7_0_4); 7 Mar 2001 23:34:57 -0000 Received: (qmail 3301 invoked from network); 7 Mar 2001 23:34:57 -0000 Received: from unknown (10.1.10.26) by l10.egroups.com with QMQP; 7 Mar 2001 23:34:57 -0000 Received: from unknown (HELO mail.reutershealth.com) (204.243.9.36) by mta1 with SMTP; 7 Mar 2001 23:34:56 -0000 Received: from reutershealth.com (IDENT:cowan@[192.168.3.11]) by mail.reutershealth.com (Pro-8.9.3/0.0.0) with ESMTP id SAA10649; Wed, 7 Mar 2001 18:37:21 -0500 (EST) Message-ID: <3AA6C620.9040908@reutershealth.com> Date: Wed, 07 Mar 2001 18:37:04 -0500 User-Agent: Mozilla/5.0 (X11; U; Linux 2.2.16-22 i686; en-US; 0.8) Gecko/20010215 X-Accept-Language: en MIME-Version: 1.0 To: humanmarkup Cc: lojban@yahoogroups.com Subject: Introducing the Logical Language Group Content-Type: text/plain; charset=us-ascii; format=flowed Content-Transfer-Encoding: 7bit From: John Cowan X-Yahoo-Message-Num: 5734 [Prenote for lojban-list members: the Human Markup Language web page is at http://humanmarkup.org] This is to introduce the Logical Language Group to the Human Markup Language effort. For the past 40+ years, the LLG and its predecessors have been engaged in making a *fully explicit* language for human communication. (Don't get hung up on the word "logical" in our title; we *use* predicate logic but we are not limited by it.) Our Web page is at http://www.lojban.org . As part of this effort, we have devised a mechanism for expressing, rather than stating or predicating, human emotions. Essentially, short words formed mostly of vowels only are used to "decorate" plain statements to supply the emotional or discursive context for them. In order to move from an abstract design to a fully functional language, we have had to create a taxonomy of emotions. It seems to me (speaking for myself, not the LLG) that adopting this taxonomy would make a useful starting point for the HumanML effort, at least for the following fields of expression: attitudes/emotions (both propositional and "pure") evidentials ("how the speaker know") discourse functions speech acts We have nothing special to provide on gestures. Our general design identified 39 emotional or attitudinal scales. Each scale has end-points which can be thought of as opposing emotions/attitudes, such as "surprise...expectation". If you are surprised at something, you are not at all expecting it, and vice versa. There is also a zero point on each scale, which in this case can be labeled "no surprise": something which is neither surprising nor expected. Just 39 emotional scales are obviously not enough for completeness. Starting with a much longer list, we were able to subcategorize our 39 emotions by employing 6 general categories and 8 more restricted ones. The general categories can be applied to subdivide almost any emotion. For example, "physical" is a general category and so is "social": we can distinguish between physical caution (keeps you from jumping off the roof) and social caution (keeps you from making rude comments in a loud voice in public). The restricted categories can formally be applied to any of the emotional scales, but may only be useful with some. Somewhat arbitrarily, we subdivide each scale into 7 points: very negative, negative, slightly negative, neutral, slightly positive, positive, very positive. For example, the scale "pride...shame" has the points "very proud", "proud", "slightly proud", "modest", "slightly ashamed", "ashamed", and "very ashamed". The English words we use to describe the scales don't always exactly fit the pure notion of a scale, which is intended to take precedence. The general model therefore is that one's emotional state is a set of tuples {E x C x S}, where E is a principal emotion, C is a general or restricted category, and S is a point on its scale. By expressing the complete set of tuples, one can give one's entire emotional state, but of course it will be typical to express just a few. For completeness, we augment E with the neutral or unspecified attitude/emotion, which can be used to express a category or scale-point by itself when the emotion is obvious. Grammatically, we allow the attachment of an emotion to any amount of text or speech, from a single word to a whole document or discussion. This agrees well with hierarchical markup of documents. I don't wish to overwhelm the HumanML (I favor this form, BTW) community with too much detail all at once. The writeup at http://www.lojban.org/files/reference-grammar/chap13.html may not be very comprehensible out of its context, and is focused on the specifics of the Lojban language rather than the design. I will therefore enter into specifics about emotions and the other categories mentioned above in future postings, assuming the HumanML response is not too negative. (I am quite prepared to be told that my efforts are premature or off the point.) -- There is / one art || John Cowan no more / no less || http://www.reutershealth.com to do / all things || http://www.ccil.org/~cowan with art- / lessness \\ -- Piet Hein