Upload
angela-scott
View
215
Download
1
Embed Size (px)
Citation preview
Understanding Semantic Relationships
By Veda C. Storey
CMPT 455/826 - Week 5, Day 1
Sept-Dec 2009 – w5d1 1
Today is full of questions
• While many of them could make good critique items, it is important for all of us to
Sept-Dec 2009 – w5d1 2
Focus
• What is the main concept of this paper
Sept-Dec 2009 – w5d1 3
Focus
• Why is this paper included in the course at this point
Sept-Dec 2009 – w5d1 4
Focus
• How does this paper relate to the other concepts we has studied so far?
Sept-Dec 2009 – w5d1 5
Relationships
• Discuss the differences between– types of semantic relationships– instances of these relationships
Sept-Dec 2009 – w5d1 6
Relationships
• How should we use types of semantic relationships in design?
Sept-Dec 2009 – w5d1 7
Relationships
• How should we use instances of semantic relationships in design?
Sept-Dec 2009 – w5d1 8
Relationships
• Do semantic relationships hold for all instances?
Sept-Dec 2009 – w5d1 9
Relationships
• What are some reasons why semantic relationships should hold for all instances?
Sept-Dec 2009 – w5d1 10
Relationships
• What are some reasons why semantic relationships might not hold for all instances?
Sept-Dec 2009 – w5d1 11
Relationships
• What should we do about semantic relationships that do not hold for all instances?
• How can we work with this?
Sept-Dec 2009 – w5d1 12
Groups of Relationships
• It seems that some of the distinctions made in this paper are somewhat blurred
• (e.g. Member-Group vs Member-Collection)
• Discuss whether or not these blurred distinctions will have a significant effect on the resulting database and whether or not they should be ignored if they cannot be better described
Sept-Dec 2009 – w5d1 13
Groups of Relationships
• The paper identifies a number of types and subtypes of semantic relationships.
– Are they mutually exclusive
– Do they have to be mutually exclusive?
– How can we work with uncertainty in deciding which one to use?
Sept-Dec 2009 – w5d1 14
Groups of Relationships
• In specifying semantics more precisely, there might be a possibility that the semantics might change over time.
• Discuss whether or not changes might occur in the resulting structure and any possible impact they might have
Sept-Dec 2009 – w5d1 15
Groups of Relationships
• While the paper introduces a number of new semantics, it does not propose any way of distinguishing them once they are used to develop an E-R model.
• Discuss the possible problems of losing this information and waht can be done about it.
Sept-Dec 2009 – w5d1 16
Groups of Relationships
• Discuss how the ideas in this paper could lead to identifying even more semantic relationships
Sept-Dec 2009 – w5d1 17
Groups of Relationships
• In various places the paper suggests that some types of semantic relationships do not / might not have a direct impact on the structure of the final database design.
• Discuss whether or not this is true and why the author might say this.
Sept-Dec 2009 – w5d1 18
Individual Relationships
• The paper seems to assume that each term has a single unambiguous meaning – (except for terms like “has”, which it is trying
to replace with more specific terms)
• Discuss the challenges and opportunities involved in dealing with terms that can have multiple meanings.
Sept-Dec 2009 – w5d1 19
Individual Relationships
• The paper deals with inclusion as a unique, exclusive relationship. However there are many types of inclusion which need not be exclusive – (e.g. you can be a student in various classes)
• Discuss the range of inclusions that are possible
Sept-Dec 2009 – w5d1 20
Individual Relationships
• Discuss how “possession” differs from other types of relationships.
• Discuss what some of the unique challenges it introduces for modeling.
Sept-Dec 2009 – w5d1 21
Individual Relationships
• The paper identified the possible existence of similar “sibling” terms, but just recommended choosing one and deleting the other.
• Discuss the problems with ignoring synonyms and suggest a better approach to dealing with them.
Sept-Dec 2009 – w5d1 22
Individual Relationships
• The paper identified “siblings” and “antonyms” as two types of semantic relationships, – but it did not consider that mutually exclusive
siblings might be not be antonyms.
• Discuss how similarities and differences should be dealt with to ensure that all possibilities are covered.
Sept-Dec 2009 – w5d1 23
Individual Relationships
• The paper discusses “antonyms” as 2 separate mutually exclusive relationships.
• Is there a simpler, more straightforward way of dealing with mutually exclusive relationships?
Sept-Dec 2009 – w5d1 24
Individual Relationships
• While the paper discussed some temporal issues, it didn’t go into all possible types of temporal relationships.
• Discuss the range of possible temporal relationships.
Sept-Dec 2009 – w5d1 25
Individual Relationships
• The paper primarily discusses simple relationships, based on a single semantic.
• Discuss the challenges of dealing with relationships that could involve compounds of multiple different semantics (e.g the problem of antonyms).
Sept-Dec 2009 – w5d1 26
Applying this
• While the paper has some potentially interesting ideas, there does not seem to be any strong motivation for developers to make use of them.
• Discuss how and why a developer could be encouraged to make use of these ideas.
Sept-Dec 2009 – w5d1 27
Applying this
• The paper suggests that there are three possible sources of inputs to design:– end users– developers– a design tool (that is designed to help)
• Discuss the strengths and limitations of each.
Sept-Dec 2009 – w5d1 28