clarke@csri.toronto.edu (Jim Clarke) (01/21/89)
(GB = Galbraith Building, 35 St. George Street)
(SF = Sandford Fleming Building, 10 King's College Road)
SUMMARY:
AI SEMINAR - Wed., Feb. 8, 11 a.m. in GB 404 -- Terry Sejnowski
"Stereo Vision of Transparent Surfaces and Distributed Representations"
AI SEMINAR - Thurs., Feb. 9, 11 a.m. in SF 1105 -- Janyce M. Wiebe
"Recognizing Characters' Thoughts and Perceptions"
----------------------
AI SEMINAR - Wednesday, February 8, 11 a.m. in Room GB 404
Terry Sejnowski
Johns Hopkins University
"Stereo Vision of Transparent Surfaces and Distributed Representations"
AI SEMINAR - Thursday, February 9, 11 a.m. in Room SF 1105
Janyce M. Wiebe
Department of Computer Science, SUNY at Buffalo
"Recognizing Characters' Thoughts and Perceptions"
I will discuss the recognition of characters' thoughts and per-
ceptions (subjective sentences) in third-person narrative text.
Subjective sentences reveal characters' judgments, evaluations,
emotions, plans, goals, and beliefs, and so recognizing them is
essential to the reader's comprehension of the story told by the
text. It is also essential to the reader's comprehension of the
text itself, since subjective sentences often contain deictic and
expressive terms that must be understood from the point of view
of the thinking or perceiving character (the subjective charac-
ter). How readers recognize subjective sentences is a signifi-
cant problem because the fact that a sentence is subjective and
who the subjective character is are often not explicitly indicat-
ed in the text. A subjective context is one or more subjective
sentences attributed to the same subjective character. My focus
is the role of linguistic information in the reader's recognition
of subjective contexts. My algorithm uses two kinds of informa-
tion. The first, potential subjective elements, are linguistic
elements that suggest that the sentence is not objective narra-
tion. They include elements that express emotion, evaluation, or
lack of knowledge. The second kind of information used is the
structure of the text. My algorithm is a discourse process that
decides whether the current sentence is subjective, and, if so,
who the subjective character is, given the features of the
current sentence and what came before in the text. An important
kind of sentence is one with a psychological or perceptual term,
which often establishes the beginning of a subjective context.
References in subjective contexts reflect the beliefs of the sub-
jective character, even if the character is mistaken. How the
results of the discourse process are used to understand refer-
ences with respect to an appropriate set of beliefs will be dis-
cussed.
_________________________________________________________
Note: Psychology talk title "Hebb Synapses, Hebb Molecules, and
the Search for Anti-Hebbian Long-Term Depression" February 8, 4
p.m. 4/F lounge of Sidney Smith Hall
--
Jim Clarke -- Dept. of Computer Science, Univ. of Toronto, Canada M5S 1A4
(416) 978-4058
BITNET,CSNET: clarke@csri.toronto.edu CDNNET: clarke@csri.toronto.cdn
UUCP: {allegra,cornell,decvax,linus,utzoo}!utcsri!clarke