weak slot filler

Upload: nitesh-mishr

Post on 09-Apr-2018

221 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/8/2019 Weak Slot Filler

    1/16

    10/1/2010

    Weak Slot & Filler Structure

    Artificial Intelligence

    Monotonic Inheritance can be performed substantially more efficientlywith these structures than with pure logic, and non monotonicinheritance is also easily supported.The reason that make Inheritance easy is that the knowledge in slotand filler systems is structured as a set of entities and their attributes.

    o It indexes assertions by the entities they describe. As a result,retrieving the value for an attribute of an entity is fast.

    o It make it easy to describe properties of relations. To do this in apurely logical system requires higher-order mechanisms.

    o It is a form of object-oriented programming and has theadvantages that such systems normally have, includingmodularity and ease of viewing by people.

    o We will discuss 2 things - Semantic Nets & Frames

  • 8/8/2019 Weak Slot Filler

    2/16

    10/1/2010

    Weak Slot & Filler Structure

    Artificial Intelligence

    Semantic NetsThe main idea is that the meaning of the concept comes from theyways in which it is connected to other concepts.In a semantic net, information is represented as a set of nodesconnected to each other by a set of labeled arcs, which represent

    relationships among the nodes.Is aHas -partinstanceuniform color team

    we can use inheritance to derive additional relations

    Mammal

    Person Nose

    Blue Sachin India

  • 8/8/2019 Weak Slot Filler

    3/16

    10/1/2010

    Weak Slot & Filler Structure

    Artificial Intelligence

    Semantic NetsIntersection Search :- We try to find relationships among objects byspreading activation out from each of two nodes and seeing wherethe activation meets. Using this we can answer the questions like what is the relation between India and Blue.

    It takes advantage of entity based organization of knowledge that slotand filler representation provide.Represent Non-binary Predicates : Simple binary predicates likeisa(Person,Mammal) can be represented easily by semantic nets butother non binary predicates can also be represented by usinggeneral-purpose predicates such as isa and instance.

  • 8/8/2019 Weak Slot Filler

    4/16

    10/1/2010

    Weak Slot & Filler Structure

    Artificial Intelligence

    Semantic NetsThree or even more place predicates can also be converted to abinary form by creating one new object representing the entirepredicate statement and then introducing binary predicates todescribe relationship to this new object.

    Example score(Australia, India, 250-275)isaVisiting team scoresHome-team

    Game

    G5 275-250 Australia

    India

  • 8/8/2019 Weak Slot Filler

    5/16

    10/1/2010

    Weak Slot & Filler Structure

    Artificial Intelligence

    Semantic NetsMaking important distinctions

    height

    height height

    greater-thanWe can add some more information.

    John 72

    BillJohn

    H2H1

  • 8/8/2019 Weak Slot Filler

    6/16

    10/1/2010

    Weak Slot & Filler Structure

    Artificial Intelligence

    Semantic NetsPartitioned Semantic NetsWe want to represent simple quantified expressions in semantic nets.It can be done with the help of partitioning the semantic net into ahierarchical set of spaces, each of which corresponds to the scope of

    one or more variable.Eg The dog bit the mail carrier. [ this is simple and can be easilyrepresented]Every dog has bitten a mail carrier.Every dog has bitten every mail carrier.

    [To be read from the book]

  • 8/8/2019 Weak Slot Filler

    7/16

    10/1/2010

    Weak Slot & Filler Structure

    Artificial Intelligence

    Semantic NetsEvolution of Frames

    As seen in the previous problem, there are problems which aredifficult to solve with Semantic Nets. Although there is no clear distinction between a semantic net and frame system, then more

    structure the system has, the more likely it is to be termed a framesystem.

  • 8/8/2019 Weak Slot Filler

    8/16

    10/1/2010

    Weak Slot & Filler Structure

    Artificial Intelligence

    Frames A frame is a collection of attributes (called slots) and associatedvalues that describe some entity in the world.Sometimes a frame describes an entity in some absolute sense,sometimes it represents the entity from a particular point of view only.

    A single frame taken alone is rarely useful, we build frame systemsout of collections of frames that are connected to each other by virtueof the fact that the value of an attribute of one frame may be another frame.

  • 8/8/2019 Weak Slot Filler

    9/16

    10/1/2010

    Weak Slot & Filler Structure

    Artificial Intelligence

    Frames as Sets and InstancesThe set theory is a good basis for understanding frame systems.Each frame represents either a class ( a set) or an instance ( anelement of class)Considering the Cricket example Person, Adult Male, Bowler, Team

    are all classes. Sachin and India are entities.Both isa and instance relations have inverse attributes, which we callsubclasses & all-instances.

    As a class is represents a set, there are 2 kinds of attributes that canbe associated with it.Its own attributes & Attributes that are to be inherited by each

    element of the set.(* - This is how they are represented.)

  • 8/8/2019 Weak Slot Filler

    10/16

    10/1/2010

    Weak Slot & Filler Structure

    Artificial Intelligence

    Frames as Sets and InstancesSometimes, the difference between a set and an individual instancemay not be clear. Example Team India is an instance of class of Cricket Teams and can also be thought of as set of players.Now the problem is if we represent Team India as a sub class of

    Cricket teams, then indian players automatically become part of allthe teams, which is not true. We have to do something to stop this.Instead we can make Team India a sub class of class called CricketPlayers.To do this we need to differentiate between regular classes and metaclasses.

    Regular Classes are those whose elements are individual entitieswhereas Metaclasses are those special classes whos elements arethemselves classes.

  • 8/8/2019 Weak Slot Filler

    11/16

    10/1/2010

    Weak Slot & Filler Structure

    Artificial Intelligence

    Frames as Sets and InstancesThe most basic meta class is the class CLASS. It represents the setof all classes.

    All classes are instances of it, either directly or through one of itssubclasses.

    The class CLASS introduces the attribute cardinality, which is to beinherited by all instances of CLASS.Cardinality stands for number.Every Class is a set but every set is not a Class.

    [Refer Fig.9.5 from the text book]

  • 8/8/2019 Weak Slot Filler

    12/16

    10/1/2010

    Weak Slot & Filler Structure

    Artificial Intelligence

    Other ways of Relating Classes to Each Other We have discussed that

    A class1 can be a subset of class2.If Class2 is a meta class then Class1 can be an instance of Class2.

    Another way is - mutually-disjoint-with relationship, which relates a

    class to one or more other classes that are guaranteed to have noelements in common with it. Another one is is-covered-by :- which relates a class to a set of subclasses, the union of which is equal to it.If a class is-covered-by a set S of mutually disjoint classes, then S iscalled a partition of the class.

  • 8/8/2019 Weak Slot Filler

    13/16

    10/1/2010

    Weak Slot & Filler Structure

    Artificial Intelligence

    Slots as Full-Fledged Objects(Frames)Till now we have used attributes as slots, but now we will like torepresent attributes explicitly and describe their properties.Some of the properties we would like to be able to represent and usein reasoning include :

    The class to which the attribute can be attached?Constraints on either the type or the value of the attribute. A value that all instances of a class must have by the definition of theclass.

    A default value for the attribute.Rules for inheriting values for the attribute.

    Rules for computing a value separately from inheritance.Whether the slot is single-valued or multi-valued?

  • 8/8/2019 Weak Slot Filler

    14/16

    10/1/2010

    Weak Slot & Filler Structure

    Artificial Intelligence

    Slots as Full-Fledged Objects(Frames)To be able to represent these attributes of attributes, we need todescribe attributes(slots) as frames.These frames will be organized into an isa hierarchy, just as anyother frames are, and that hierarchy can then be used to supportinheritance of values for attributes of slots.

    Now let is formalize what is a slot. A slot here is talked about as a relation . It maps from elements of itsdomain (the classes for which it makes sense) to elements of itsrange(its possible values). A relation is a set of ordered pairs. Thus itmakes sense to say that relation R1 is a subset of another R2.In that case R1 is a specialization of R2.

    Since a slot is a set , the set of all slots, which we will call SLOT, is ameta class. Its instances are slots, which may have sub slots. Fig-9.10

  • 8/8/2019 Weak Slot Filler

    15/16

    10/1/2010

    Weak Slot & Filler Structure

    Artificial Intelligence

    Slots Values as ObjectsIn the last section, we have used slots as explicit object that we couldmake assertions about.Here, we will discuss about using slot values as objects.Lets take the following example

    John :height : 72Bill :height :The only information we have here is that John is taller than Bill

  • 8/8/2019 Weak Slot Filler

    16/16

    10/1/2010

    Weak Slot & Filler Structure

    Artificial Intelligence

    Slots Values as ObjectsWe need to make an statement about the value of a slot withoutknowing what the value is. To do that, we need to view the slot andits value as an object.We will expand our representation language to allow the value of slot

    to be stated as either or both of : A value of the type required by the slot. A logical constraint on the value. The constraint may relate the slotsvalue to the values of other slots or to domain constants.

    Eg : View Fig : 9.14.