Presentation is loading. Please wait.

Presentation is loading. Please wait.

IFLA Library Reference Model

Similar presentations


Presentation on theme: "IFLA Library Reference Model"— Presentation transcript:

1 IFLA Library Reference Model
Overview and Discussion The views and opinions presented by myself and Chris Holden are our opinions alone and do not represent the opinions of the Library of Congress. The IFLA Library Reference Model, referred to in the rest of this discussion as LRM, is a “high-level conceptual reference model developed within an enhanced entity-relationship modelling framework” that “aims to make explicit [the] general principles governing the logical structure of bibliographic information, without making presuppositions about how that data might be stored in any particular system or application” (IFLA LRM, p. 7). IFLA Library Reference Model

2 LRM Antecedents Functional Requirements for Bibliographic Records (1997) Functional Requirements for Authority Data (2009) Functional Requirements for Subject Authority Data (2010) Created by the IFLA Functional Requirements for Bibliographic Records Review Group, LRM is a consolidation of the Functional Requirements for Bibliographic Records, or FRBR, approved by IFLA in 1997, Functional Requirements for Authority Data, or FRAD, approved in 2009, and Functional Requirements for Subject Authority Data, or FRSAD, approved in LRM replaces these conceptual models.

3 User Tasks Find: To bring together information about one or more resources by searching on relevant criteria. Identify: To clearly understand the nature of the resources found and to distinguish between similar resources. Select: To determine the suitability of the resources found, and to be enabled to either accept or reject specific resources. Obtain: To access the content of a resource. Explore: To discover resources using the relationships between them and thus place the resources in a context. As with the previous models, LRM defines a set of user tasks which is then used to develop the entities, attributes, and relationships in the model. The user tasks are considered from the perspective of the end-user. This includes traditional library patrons and may also include library staff and others responsible for the creation and maintenance of bibliographic data. However, administrative and rights metadata, while needed for the management of bibliographic and authority data to enable it to meet user needs are considered out of scope for LRM. The following user tasks are identified and defined in LRM: [click] Find: To bring together information about one or more resources by searching on relevant criteria. Information systems must be able to provide effective searching capabilities for users. [click]Identify: To clearly understand the nature of the resources found and to distinguish between similar resources. Information systems must be able to clearly describe the resources it provides. [click] Select: To determine the suitability of the resources found, and to be enabled to either accept or reject specific resources. Information systems must be able to allow and support relevance judgments so that the user may select the most appropriate resource for their needs. [click] Obtain: To access the content of a resource. Information systems must be able to provide access to required resources. [click] Explore: To discover resources using the relationships between them and thus place the resources in a context. Information systems must be able to provide methods for discovery of other resources.

4 Entities, Attributes, and Relationships
Class based Entities inherit the attributes and relationships of the superclass entity One way street Person (LRM-E7) Subclass of Agent (LRM-E6) Contact Information (LRM-E6-A1) Field of Activity (LRM-E6-A2) Because Persons are Agents they also have Contact Information and Field of Activity attributes BUT Profession/Occupation (LRM-E7-A10), attribute of Person, is not applicable to all Agents, which also include Collective Agents (LRM-E8) The entities, attributes, and relationships in LRM, while modeled similarly to the previous models, is different in that the model is classed-based. What that means is that if an entity is a subclass of another entity, it inherits the attributes defined at the higher-level, or if a relationship is declared at a higher level, it is also applicable at the lower-level. [click] For example, the entity Person (LRM-E7) is a sub-class of the entity Agent (LRM-E6). The attributes Contact Information (LRM-E6-A1) and Field of Activity (LRM-E6-A2) are attributes of an Agent. Because Person is a sub-class of Agent, these attributes also apply to the entity Person. [click]However an attribute defined in a sub-class does not mean that the attribute is applicable to the superclass. For example, the attribute Profession/Occupation (LRM-E7-A1) is an attribute of Person. However this attribute is not applicable to the entity Agent, its superclass. Because of the classing inherent in the model, the number of entities, attributes, and relationships defined in the model are on the small-side: only 11 entities, 37 attributes, and 36 relationships are defined. However, I want to stress that LRM is a high-level model. If required, additional entities, attributes, and relationships may be defined by library applications as long as they are compliant with the high-level model.

5 Entities Res (LRM-E1) Work (LRM-E2) Expression (LRM-E3)
Any entity in the universe of discourse Superclass of all other entities in LRM Allows for extensibility of LRM Work (LRM-E2) The intellectual or artistic content of a distinct creation Expression (LRM-E3) A distinct combination of signs conveying intellectual or artistic content. Manifestation (LRM-E4) A set of all carriers that are assumed to share the same characteristics as to intellectual or artistic content and aspects of physical form. Item (LRM-E5) An object or objects carrying signs intended to convey intellectual or artistic content. As I mentioned earlier, LRM defines 11 entities. The top-level entity is Res (LRM-E1) (Latin for “thing”) and is defined as any entity in the universe of discourse. It is the superclass of not only the other 10 entities in the model, but is also the superclass of other entities that have not yet been labeled. In order for LRM to be extensible, and in order for it be used as the basis of cataloging rules, this superclass is essential. The subclasses of Res include entities with which most of you are familiar from earlier FR models. They are: [click] Work (LRM-E2): The intellectual or artistic content of a distinct creation. The Work is a conceptual object, meaning that no single material object can be identified as the work. A work can be identified by the commonality of content between Expressions of Works. [click] Expression (LRM-E3): A distinct combination of signs conveying intellectual or artistic content. An expression is an abstract object, separate from the carriers used to record it. [click] Manifestation (LRM-E4): A set of all carriers that are assumed to share the same characteristics as to intellectual or artistic content and aspects of physical form. [click] Item (LRM-E5): An object or objects carrying signs intended to convey intellectual or artistic content.

6 Entities Agent (LRM-E6) Person (LRM-E7) Collective Agent (LRM-E8)
An entity capable of deliberate actions, of being granted rights, and of being held accountable for its actions. It is strictly equivalent to its subclasses, Person and Collective Agent. Person (LRM-E7) An individual human being. Restricted to real persons who are living or who are assumed to have lived. Collective Agent (LRM-E8) A gathering or organization of persons bearing a particular name and capable of acting as a unit. Excludes national, religious, cultural or ethnic groups. Agent (LRM-E6): An entity capable of deliberate actions, of being granted rights, and of being held accountable for its actions. It is strictly equivalent to its subclasses, Person and Collective Agent. [click] Person (LRM-E7): An individual human being. Restricted to real persons who are living or who are assumed to have lived. [click] Collective Agent (LRM-E8): A gathering or organization of persons bearing a particular name and capable of acting as a unit. Excludes national, religious, cultural or ethnic groups.

7 Entities Nomen (LRM-E9) Place (LRM-E10) Time-span (LRM-E11)
An association between an entity and a designation that refers to it. Place (LRM-E10) A given extent of space. Time-span (LRM-E11) A temporal extent having a beginning, an end and a duration. Nomen (LRM-E9): An association between an entity and a designation that refers to it. For example, I am a Person. My Nomens include Damian (given name), Damian Iseminger (given name and surname), Damian Scott Iseminger, and Iseminger, Damian, (authorized access point). [click] Place (LRM-E10): A given extent of space. [click] Time-span (LRM-E11): A temporal extent having a beginning, an end and a duration. All entities, except for those that are a subclass of another entity, are considered disjoint, that is something cannot be two or more entities at the same time.

8 Attributes Declared when the target of a potential attribute was not an instance of any other entities in the model Relationships declared when the target is another LRM entity Title of work Attribute in FRBR Relationship in LRM Res (LRM E-1) has appellation (LRM-R13) Nomen (LRM E-9) Medium of Performance (LRM-E3-A8) and Key (LRM-E3-A7) Target not an LRM entity, so modeled as attributes of Expressions With the exception of Collective Agent, all of the entities have properties, called attributes, associated with them. The criteria for establishing whether a given attribute should be defined for an entity was determined by whether the target of the potential attribute was not an instance of any of the other entities in the model. If the target of the potential attribute was one of the LRM entities, then a relationship was used. [click]For example, in LRM there is no attribute for the title of a work. Because the entity Nomen exists, the work title is instead modeled as a relationship between the Work entity and the Nomen entity, specifically using the “has appellation” relationship (LRM-R13). [click]By contrast, the Expression entity has the attributes of Medium of performance (LRM-E3-A8) and Key (LRM-E3-A7). These are not modelled as LRM entities, so therefore they are given as an attributes.

9 Relationships Res (LRM-E1) is associated with (LRM-R1) Res (LRM-E1)
All other relationships are refinements of the high-level relationship Relationships in LRM are not exhaustive The relationships are the glue that holds together LRM. At its heart, there is actually only 1 high-level relationship in the model: Res is associated with Res (LRM-R1). What this statement means is that any entity present in LRM may have a relationship with another entity. All of the other relationships present in the model are refinements of this relationship. As with the entities and attributes, the relationships in LRM are not exhaustive. Implementation scenarios may create other relationships that are needed. All of the entities and their relationships in LRM can be summarized by consulting the following diagrams, which come directly from the draft IFLA LRM report. In the diagrams, and the ones that follow it, the entities are given in rectangles and the relationships are illustrated through lines connecting the entities. A single arrow indicates the cardinality for the entity is 1; a double arrow indicates the cardinality is Many.

10 The WEMI Stack The core relationships in the model are between Works, Expressions, Manifestations, and Items. A Work is realized through one or many Expressions, but an Expression may only realize 1 Work; an Expression may be embodied in multiple Manifestations; likewise a Manifestation may embody multiple Expressions; a Manifestation may be exemplified by many Items, but an Item only exemplifies 1 Manifestation.

11 Responsibility Responsibility for Works, Expressions, Manifestations, and Items are represented by this diagram. The gist of this diagram is that Agents are responsible for Works, Expressions, Manifestations, and Items.

12 Subjects Appellation The subject relationship in LRM is modeled as 1 or more Works having as subjects 1 or more Res. Since all entities are sub-classes of Res, this means that Works may have as a subject any of the LRM entities or any object of intellectual discourse, which is covered by the Res entity. [click] The appellation relationship is modelled by this diagram. This means that a single Res may have many Nomen appellations. The circle at the end of the Nomen entity indicates that Nomens are composed of other Nomens.

13 Agents This diagram illustrates the relationships between Agents. Here we see that Persons and Collective Agents are types of Agents, and that Collective Agents have members that are Agents. The cardinality of the is member of relationship is not provided in this diagram.

14 Putting all the diagrams together, plus adding in the Place and Time-span entities, yields us this diagram, which illustrates all the entities and relationships in the model. This, in a nutshell, is LRM. We will now examine some portions of the model that may be of interest to music catalogers.

15 The Work in LRM The work entity in LRM only has two attributes.
Category A broad attribute used to characterize the creative domain or form/genre of a work Creative domain: Is the work literature? Music? A motion picture? Form/genre: Is the work a symphony? A novel? Representative expression attribute “An attribute which is deemed essential in characterizing the work and whose values are taken from a representative or canonical expression of the work.” [As Damian mentioned,] the work is so abstract under LRM that it only has two attributes specifically associated with it - category, and representative expression attribute. Category is a broad attribute used to characterize a work with regard to creative domain and form/genre. We'd use category to state that a work was a piece of music, or a motion picture, or in the form of a symphony or a concerto or a novel. The representative expression attribute is defined in LRM as ““An attribute which is deemed essential in characterizing the work and whose values are taken from a representative or canonical expression of the work.”

16 The Work in LRM Attributes previously associated with the work have been relegated to the expression level. Medium of performance Key Language continues to be an expression level attribute. But what does this mean, and why do we need a representative expression attribute? Attributes previously associated with the work in the original FRBR, such as medium of performance and key, have been relegated to the expression level. As with the original FRBR, language continues to be an expression level attribute as well. None of these attributes apply to the work.

17 The Work in LRM But without these attributes on the work level, we get some strange results. This is the authority record for Schubert’s piano sonata D. 959 in A major. Under LRM, we can’t say that the work represented by this authority record is a work for piano, or a work in A major. These two attributes – [click] medium of performance and [click] key - are no longer work level attributes. All we can say is that the work is a piece of music in sonata form, but we can't say that the work has a medium of performance or a key. The expression for piano in A major is just one expression among many.

18 Schubert, Franz, 1797-1828. Sonatas, piano, D.959, A major
The Work in LRM All expressions of a work are equal in the eyes of LRM. An expression of this Schubert work may be performed on piano in A major, just as Schubert composed it. Another expression of the work could be arranged for violin, and in C major. WORK: Schubert, Franz, Sonatas, piano, D.959, A major This is because all expressions of a work are equal in the eyes of LRM. So an expression of this work could have piano as the medium of performance, and may be in A major. However, another expression of this work could just as easily be for violin, and be transposed to C major.  Technically, these two expressions are equal realizations of the work in LRM, which states that “in a strict formal sense, within the model all expressions of a work are equal realizations of the work.” EXPRESSION 1 EXPRESSION 2 MoP: piano key: A major key: C major MoP: violin

19 Representative expressions
LRM recognizes that most users are going to assume that one of these expressions represents Schubert’s intentions better than the other. The representative expression attribute is a shortcut that allows us to apply certain “canonical” expression level attributes to the work entity. Users can now search for “works for piano” or “works in A major” This can get confusing, especially for users not versed in bibliographic modeling. But that’s okay! The model recognizes that this is not intuitive, and that most users are going to assume that the work of Schubert’s piano sonata is going to be a work for piano. The representative expression attribute allows us to choose attributes from expressions that best represent Schubert’s original intention, and apply them to the work. So, a user looking for Schubert's piano sonatas isn’t going to have to ask for something super complicated, like "Do you a particular expression of a work in which the attributes of said expression best represent the intentions of the creator of the work?" Instead, the “representative expression” attribute is a shortcut that allows us to discuss expression-level attributes as though they are associated with the work entity. This allows a cataloger to choose certain “canonical” characteristics of an expression to be “transferred” to a work to be used for work identification. This would allow users to search for “works for piano” easily, without having to articulate that they’re really looking for “works in which the canonical expression is for piano.”

20 Schubert, Franz, 1797-1828. Sonatas, piano, D.959, A major
MoP: piano WORK: Schubert, Franz, Sonatas, piano, D.959, A major Representative expression attributes: key: A major EXPRESSION 1 EXPRESSION 2 So here’s that chart again of two expressions of Schubert’s piano sonata. Again, these expressions are each given equal authority in the model. However, if we decide that the expression for piano in A major is the “canonical” expression, [click] we can transfer its attributes to the work as representative expression attributes. So while the work technically still has no medium of performance and key, the “representative expression attributes” allow us to discuss the work as if it were for piano, and in A major. MoP: piano key: A major MoP: violin key: C major

21 Representative expression attribute:
WORK: Wagner, Richard, Ring des Nibelung. Rheingold. Language: German EXPRESSION 1 EXPRESSION 2 To use an another example, Wagner’s Das Rheingold is a work that technically has no language in the LRM model. This is because the work entity is an abstract concept that has no language. Language is an expression-level attribute. Here, we see the work of Das Rheingold and two expressions of the work, each in a different language. One is in German and one is in English. Neither takes precedence over the other. However, we could designate the language attribute of the German expression of Das Rheingold as a “representative expression attribute” and [click] port this language attribute to the work. This would allow users to search for operas composed in German without having to grapple with the intricacies of a bibliographic model. So while technically Das Rheingold has no associated language in LRM, it allows us to state that the most canonical expression of the opera is in German, and we can associate this language attribute with the work entity to assist users in searching for operas in German. Language: English Language: German

22 Representative expressions
Catalogers have a lot of latitude in determining representative expression attributes. Unlike the previous draft, one no longer has to designate a specific expression as representative. It is not even necessary to indicate which expression from which the representative expression attribute is taken. LRM gives catalogers a lot of latitude in determining which expression, if any, is the “representative” one - this is something that LRM says is to be “operationalized by the relevant cataloging practice.” Unlike the previous draft, one no longer has to choose a specific expression as the representative expression. In the new LRM iteration, we only choose expression attributes to be associated with the work, rather than an entire expression to be canonized as "representative.” It is not even necessary to indicate a specific expression from which the representative expression attribute is taken.

23 Representative expressions
The representative expression attributes can come from: The original expression (such as a painting) A network of similar expressions (such as all expressions of a musical work with the same medium of performance) The expression considered canonical by a majority of users. (A specific performance of a popular music work might qualify) LRM does not stipulate criteria for determining a representative expression attribute beyond assisting end-users in finding works. So where do the representative expression attributes come from? [click] The original expression can be seen as the representative one, such as an original painting or artwork. [click] Other times, a network of similar expressions may together possess “representative” attributes - users know Schubert’s Sonata D. 959 is for piano not from examining the original manuscript, but from experiencing a multitude of published scores and recorded performances, all of which contain the “representative” attributes of being for piano, and in A major. However, some cases are not as clear-cut, [click] and users may identify “canonical” traits that are not part of the original expression at all. LRM doesn’t elaborate with any specific examples of this, but one might consider a specific performance of a piece of popular music –most users might think of a certain performance as “canonical” even if it differs from the expression created by the original songwriters. LRM does not stipulate criteria for determining which expression is “representative” beyond assisting end-users in finding works. One also has to identify which attributes of representative expressions are considered important for identification of the work. Language, for example, would be important for textual works, but not maps. 

24 Representative expressions
But why? To smooth over the complexities of bibliographic modelling to make things easier for the user. This allows one to ask for an opera in German or a sonata for piano without having to understand the work/expression distinction. It will be up to the music cataloging community to identify representative expression attributes for specific works. The whole idea is to smooth over a lot of the complexities of the LRM model in order to create a more seamless experience for the end user, who may intuitively ask for an opera in German or sonata for piano without wanting to distinguish between the attributes of works and expressions. It will be up to the music cataloging community to determine how a “representative” expression attribute is identified and which ones are considered important to port to the main work. There are no current guidelines about how any of this is going to work for music; it's up to us to figure it out!

25 Aggregates Defined as “a manifestation embodying multiple distinct expressions.” Three types of aggregates are listed. LRM also has a sizable section discussing aggregates. The original FRBR report from 1998 only briefly touched on aggregates and how they work, but here LRM goes into more greater specificity, drawing from work done by the IFLA Working Group on Aggregates. The model defines an aggregate as “a manifestation embodying multiple distinct expressions,” and lists three types of aggregates.

26 Aggregate collection of expressions
Occurs when multiple independently-created expressions are published together in a single manifestation. Includes selections, anthologies, and series. An aggregate collection of expressions occurs when multiple independently-created expressions are published together in a single manifestation. This includes selections, anthologies, and series. So for example, this CD of R.E.M.’s greatest hits would be considered an aggregate collection of expressions - the manifestation of this CD contains the expressions of several different, independently-created songs.

27 Aggregate resulting from augmentation
Occurs when a single independent work is supplemented by one or more dependent works. An aggregate resulting from augmentation occurs when a single independent work is supplemented by one or more dependent works. An edition of Beethoven’s Sixth Symphony, with an introduction written by the editor, would be an aggregate resulting from augmentation. The additional supplementary material is not integral to the original work, and does not significantly change the original expression. The cataloger can decide when the augmenting material is important enough to merit separate bibliographic identification.

28 Aggregate of parallel expressions
Occurs when a single manifestation embodies multiple parallel expressions of the same work. An aggregate of parallel expressions occurs when a single manifestation embodies multiple parallel expressions of the same work. For example, a DVD containing a motion picture with several audio tracks in different languages would qualify. Another example would a vocal score of Das Rheingold with German and English text. These are two separate expressions of the work, aggregating together here in this aggregate of parallel expressions.

29 You can see the model for aggregates here
You can see the model for aggregates here. In LRM, expressions are the entities that are aggregated, rather than works. When multiple expressions are embodied in a single manifestation, the person who selected and assembled these has created an aggregating work. The creative effort of this aggregation is what creates a new work. This is not the same as a whole/part relationship, which is a completely different relationship between two works. An aggregating work is not a work created with separate parts; it is a work created through the assembly of distinct expressions. 

30 Aggregates Aggregates are not the same as works with a whole/part relationship. LRM stresses that we may not need to model aggregates with such specificity. The model is flexible. If one decides not to model an aggregate work, one can return and do so at a later time. LRM stresses that, even though the model maps these aggregating works very specifically, it may not be necessary to model them at such an intricate level. The model is designed to be deliberately flexible; one may decide not to create distinct bibliographic information for the aggregating work, but one can also return and create a description later, if such a description is decided to be warranted. If an supplementary introduction becomes famous in its own right, one can return and create bibliographic information for an aggregate resulting from augmentation - but we wouldn't necessarily have to do this every time an introduction is published with a work. The emphasis in on flexibility, and on the ability to use these complicated models only when necessary for bibliographic identification.

31 Fictitious Entities Fictional authors, places, and time-spans are objects of human contemplation Covered by Res (LRM-E1) BUT Fictional author ≠ Person (LRM-E7) Fictional place ≠ Place (LRM-E10) Fictional time-span ≠ Time-span (LRM-E11) Perhaps no part of the LRM model has occasioned more consternation in the United States library community than the issues surrounding LRM’s treatment of fictitious entities. In short, while fictitious entities, such as fictional authors, fictional places, and fictional time-spans can be considered objects of human contemplation, and thus covered by the entity Res, they cannot be considered real world objects in LRM: a fictional author is not a Person (LRM-E7), a fictional place is not a Place (LRM-E10), a fictional time-span is not a Time-span (LRM-E11).

32 Fictitious Entities Functional Requirements for Authority Data
Persons and Corporate Bodies may be both real and imaginary Resource Description and Access Chapter 9, Identifying Persons, aligned with FRAD definition Why change things? Much of the uproar concerning fictional entities comes from the fact that Persons and Corporate Bodies in Functional Requirements for Authority Data did cover real and imaginary persons. In response to this, RDA was developed so that the instructions in chapter 9 for identifying persons covered fictional or legendary persons. So why did LRM decide to change things?

33 Fictitious Entities FRBRoo
Extension for library materials to the CIDOC Conceptual Reference Model (CIDOC-CRM) for museum information Museum objects are made by someone Therefore impossible for someone “not real” to produce an artifact Compatibility between FRBRoo and CIDOC-CRM required modification of FRAD definition LRM retains the modification The biggest reason was probably the development of the object-oriented model of FRBR, known as FRBRoo. It was developed as an extension to the CIDOC Conceptual Reference Model (CIDOC CRM) for museum information. Museums deal with real world objects, such as artifacts. It is illogical to make a claim that a fictitious entity produced an object. In fact, it is absurd on its face. So in CIDOC-CRM this concept does not exist. Therefore in order for FRBRoo to be compatible with CIDOC-CRM, the FRAD definition of what constituted a person had to be modified to only include real persons. To maintain compatibility, LRM has retained the requirement that Persons must be real.

34 Fictitious Entities Providing access to a resource through a NAME is NOT the same thing as saying that the fictional entity associated with that name is actually the author of the resource Before You Leap: A Frog's Eye View of Life's Greatest Lessons by Kermit the Frog Name access point: Kermit, The Frog BUT The fictional character known as Kermit the Frog DID NOT author the resource. A human entity using the name Jim Lewis is responsible This does not mean that the name “Jim Lewis” should be used to access this resource Catalogers in the USA have objected to the reality requirement, because it is seen as possibly endangering access to materials which are “authored” by fictional persons. This concern is understandable. Users should be able to find resources by the way in which a resource represents itself. [click] But, and this is very important, providing access to a resource through a NAME is NOT the same thing as saying that the fictional entity associated with that name is actually the author of the resource. [click] To use an example, the name given as the author of the book “Before You Leap: A Frog's Eye View of Life's Greatest Lessons” is Kermit the Frog. Catalogers are correct that an appropriate name access point would be Kermit The Frog. [click] However, using this as an access point does not necessarily mean that the character created by Jim Henson known as Kermit The Frog actually wrote the book! In fact, to make such a claim would be nonsense. Of course the character Kermit the Frog cannot write anything because he isn’t real! [click] Only a human using that name could have written the book. In fact, we do know that a Person who uses the name Jim Lewis is responsible for the book. However this does not mean we should necessarily use that NAME as an access point for this book. It is still appropriate to use Kermit the Frog as an access point. What is not appropriate is to say that the fictional entity Kermit the Frog wrote the book.

35 Fictitious Entities Access methods will probably remain the same
Change will occur in how we model this situation in a post-MARC environment All of this is to say that the current methods of providing access to materials purporting to be authored by a fictional entity will more than likely stay the same. So let’s all exhale and settle down. But what will change is the way in which we think about these kinds of situations and how we will model such situations in a post-MARC environment.

36 Thank You! Contact Information Damian Iseminger: Christopher Holden: Views expressed in this presentation are the views of the authors and not necessarily the views of the Library of Congress That brings us to the end of our discussion. If there are any questions, please save them for the end of the Town Hall. Thank you!


Download ppt "IFLA Library Reference Model"

Similar presentations


Ads by Google