timo honkela: a short introduction to modeling ambiguity and vagueness
TRANSCRIPT
Timo Honkela, Modeling Meaning and Knowledge, 14.3.2016
Timo Honkela
Modeling Meaning and Knowledge14 Mar 2016
A short introduction to
Modeling ambiguity and vagueness
Timo Honkela, Modeling Meaning and Knowledge, 14.3.2016
Discrete and continuous referents
● Discrete set of word senses
→ Word sense disambiguation“He visited the bank for a transaction”“He visited the river bank”“He visited the bank”
● Meaning of a word represented– as a continuous value in some dimension, or
– in a continuous multidimensional vector space
Timo Honkela, Modeling Meaning and Knowledge, 14.3.2016
http://www.cs.toronto.edu/~gh/
ftp://www.cs.toronto.edu/pub/gh/Hirst-semInt-88.pdf
Timo Honkela, Modeling Meaning and Knowledge, 14.3.2016
Wikipedia definition ofambiguity resolution
“In computational linguistics, word-sensedisambiguation (WSD) is an open problemof natural language processing and ontology.”
Wikipedia definition ofword-sense disambiguation
“Ambiguity resolution is used to find the value of a measurement that requires modulo sampling. This is required for pulse-Doppler radar signal processing.”
Timo Honkela, Modeling Meaning and Knowledge, 14.3.2016
Vagueness
● “In analytic philosophy and linguistics, a concept may be considered vague – if its extension is deemed lacking in clarity,
– if there is uncertainty about which objects belong to the concept or which exhibit characteristics that have this predicate (so-called "border-line cases"), or
– if the Sorites paradox appliesto the concept or predicate.”
https://en.wikipedia.org/wiki/Vagueness
https://en.wikipedia.org/wiki/Timothy_Williamson
https://en.wikipedia.org/wiki/Sorites_paradox
Timo Honkela, Modeling Meaning and Knowledge, 14.3.2016
Crisp versus fuzzy sets
● Mary is 30 (40, 50, 60, 70, 80, …) years old.Does she belong to the set of old women?Within traditional set theory, one has to answer either yes or no. There is a millisecondwhen a person turns old. Makes sense?
● According to the fuzzy set theory, items belong to some set according to a degree of membership. Typically this degree is represented by a number between 0 and 1.
Timo Honkela, Modeling Meaning and Knowledge, 14.3.2016
Lotfi Zadeh: Fuzzy sets
Berkeley 2006
https://www.eecs.berkeley.edu/ Faculty/Homepages/zadeh.html
Berkeley 2008
https://www.eecs.berkeley.edu/ Faculty/Homepages/zadeh.html
Timo Honkela, Modeling Meaning and Knowledge, 14.3.2016
Honkela 1997
https://users.ics.aalto.fi/tho/thesis/
Basic fuzzy setFuzzy set in
multiple dimensionsData-driven fuzzy
set formation
Timo Honkela, Modeling Meaning and Knowledge, 14.3.2016
Fuzzy sets and conditional probabilities
● A particular object can be quite (fuzzily) round.Can it be probably round?
● Could conditional probability used to considerif some object is called round (or not) bysome collection of people (in some particularcontext)
● It may be fair to say that fuzzy sets and conditional probabilities serve different purposes (both relevant in linguistics)