Lecture 06: Folksonomies and Uncertainty

Question 1

What are the seven problems associated with tagging?

Answer 1

1. Homonymy - same word having different meanings
2. Synonymy - relation between two different words that hold the same meaning
3. Granularity - level of detail considered
4. Aspect - the way in which something may be viewed or regarded
5. Not tight control of tagging - free text vs. controlled vocabulary
6. Seemingly chaotic

Question 2

What are the three reasons tagging is contributing to success?

Answer 2

1. Low barriers to entry - not required sophisticated knowledge
2. Feedback and Asymmetric Communications - negotiate meanings of tags, locate other related resources from tags of a certain resource

Question 3

What are desire lines?

Answer 3

The concept used to identify the relation between emergent semantics from tags. Desire lines is a path developed by erosion caused by human footfall, which provide the shortest or most easily navigated route.

Question 4

What are the social aspects of tagging behaviour?

Answer 4

• If a web resource is tagged by many users, it has more chance to be seen by other users.
• If a web resource is seen by many users, it has more chance to be tagged by other users.
• This creates a positive feedback loop and leads to exponential growth

Question 5

What is self-normalisation in regards to folksonomies?

Answer 5

The idea that controlled vocabularies become more consistent without any external control over time.
Self-normalisation can be promoted by shared information retrieval spaces.
The folksonomies tend to show a power law curve and long tail effect.

Question 6

What is Power Law?

Answer 6

1. Very small number of events that have a very high probability of appearing
2. Very large number of events that have a very low probability of appearing

Question 7

What are the two types of folksonomies?

Answer 7

1. Broad folksonomy

2. Narrow folksonomy

Question 8

What is a Broad folksonomy?

Answer 8

A broad folksonomy is whereby many people tag the same object, it follows the power law and there is an agreement on using a few popular tags. Broad folksonomies can be used to select preferred terms or extract a controlled vocabulary.

Question 9

What is a Narrow folksonomy?

Answer 9

A narrow folksonomy is whereby a few people tag the same object, as a result it loses the richness of the masses. Aids in tagging objects which are not easily findable. Specific target audiences will tag objects with specific personal tags to improve future retrieval.

Question 10

What is a tag cloud?

Answer 10

A tag cloud is a visualisation of the tags in a folksonomy with tags being weighted based on their frequency of occurrence. The more important tags are bigger and louder in the illustration. It provides an instant illustration, providing precise and specific orientation of the content. The advantage of a tag cloud is it highlights the most important popular subjects dynamically.

Question 11

What are the four types of tag clouds?

Answer 11

1. Sorted Alphabetically - font size, weight, colour
2. Sorted by Frequency - most important tags may be highlighted
3. Not sorted at all - most important are highlighted
4. Sorted according to similarity - similar terms appear close to each other

Question 12

What is a tag index?

Answer 12

A tag index is an alternative to a tag cloud in that it provides an index of tags, in a listed manner which is the best solution for precise content presentation.

Question 13

What are the advantages of folksonomies?

Answer 13

1. Serendipity - information discovery (browsing vs. finding)
2. Reflect the population’s conceptual model
3. Accomodate Diversity
4. Self-moderating
5. Low cost alternative
6. Enable emergence of social groups

Question 14

What are the disadvantages of folksonomies?

Answer 14

1. No synonym control
2. Lack of precision
3. Lack of hierarchy
4. Low findability quotient
5. Problems with scaling
6. Susceptible gaming

Question 15

What are the two goals of automatic tag analysis?

Answer 15

1. Map between Synonyms

2. Link related terms

Question 16

What are the three problems of automatic tag analysis?

Answer 16

1. Typos
2. Spelling variations
3. Morphological variations

Question 17

How can typos be dealt with and what are the two types of spell checking algorithms?

Answer 17

Typos can be tackled by spell checking algorithms, in particular:

1. Phonetic Mastering Algorithms - words that sound alike
2. String Similarity Algorithms - words that look alike

Question 18

How can morphological variations be tackled?

Answer 18

1. Stemming - reducing inflected words to their stem, base or root form : e.g. fishing, fished, fish, fishes, fisher all relate to fish.
2. Lemmatisation - determining the lemma of a given word, the canonical, dictionary or citation form of a word : e.g. fishing fished fish relate to fish
   : e.g. fisher, fishers relate to fisher

Question 19

What is synonym detection via dictionary?

Answer 19

Databases such as Wordnet are used which are comprised of a large lexical database of content words, and these words are grouped into synsets. The contents of the dictionary links all the lexical relations between words:

• Antonym (opposite in meaning)
• Hyponym (narrow term, Hypernym broader term)
• Meronym ( part of , Holonym the whole )

Question 20

What is data mining?

Answer 20

Data mining is a variety of techniques to sort through data, identifying patterns and establishing probabilistic relationships.

Question 21

What is the Dung Argumentation framework?

Answer 21

The dung argumentation framework (AF) is a pair >
Where A is a set of arguments, and –> is a binary relation on A.
It can be represented as a direct graph, where nodes are arguments and edges are attacks.
It helps solves errors in presenting data and nomenclature.

Question 22

What are the semantic properties in the Dung Argumentation framework which arguments must follow?

Answer 22

1. Conflict-freeness : an attacking argument and an attacked argument cannot stay together
2. Admissibility - extension should be able to defend itself
3. Strong Admissibility - base your defence on unattacked arguments.