week 2/3

Question 1

RDF Graph =

Answer 1

set of triples

Question 2

a triple:

Answer 2

subject, predicate, object

Question 3

RDF graph representation

Answer 3

node and directed arc (labeled) diagram

Question 4

semantic web is based on…

Answer 4

triples

Question 5

… are the cornerstone of the semantic web

Answer 5

ontologies

Question 6

components of ontologies

Answer 6

classes, relations, axioms, instances

Question 7

OWL =

Answer 7

ontology web language

Question 8

One of specializations of OWL is

Answer 8

RDF/XML

Question 9

owl: =

Answer 9

comes from owl name space and is used to define types in ontology

Question 10

rdf:ID=”” where ID is the

Answer 10

class

Question 11

symbol in serialized owl means

Answer 11

refer to this resource somewhere else in document

Question 12

rdfs: domain means

Answer 12

allowed to use property only on certain instances

Question 13

kb: in serialized owl means

Answer 13

knowledge base, defined by user

Question 14

hermes framework input

Answer 14

news items from rss feesds, domain ontology linked to a semantic lexicon, user query

Question 15

hermes framework outpu

Answer 15

news items as answer to the user query

Question 16

4 phases HF

Answer 16

1. News classification: relate news to ontology concepts
2. Knowledge Based Updating: update knowledge base with info
3. News querying: allow user to express concepts of interest and the temporal constraints
4. Result presentation: present the news items that match user’s query

Question 17

3 types of users domain ontology

Answer 17

casual user, domain expert, knowledge engineer

Question 18

ontology development

Answer 18

incremental middle-out approach, anchor concepts to their corresponding wordnet synsets, validated using the ontoclean methodology

Question 19

preprocessing of news

Answer 19

filter out duplicates news (heuristic based on title)

Question 20

multiple lexical representations for same concept:

Answer 20

ontology synonyms and semantic lexicon synonyms

Question 21

concepts w/o subclasses or instances:

Answer 21

semantic lexicon hypernyms

Question 22

a longer match in news classification subsedes ..

Answer 22

a shorter match

Question 23

WSD:

Answer 23

interconnection based on word context and a semantic lexicon

Question 24

synset:

Answer 24

set of synonyms

Question 25

sense:

Answer 25

word with specific meaning

Question 26

synset relations:

Answer 26

hypo/hypernyms, syno/antonyms, holo/meronyms

Question 27

KBU, detect ...

Answer 27

events in news items

Question 28

events in KBU have associated...

Answer 28

lexico sem patterns for their detection and actions for KBU

Question 29

KBU:variables are used to..

Answer 29

link extracted info with update information

Question 30

KBU: events are stored in...

Answer 30

ontology

Question 31

query formulation. Present domain knowledfe as...

Answer 31

directed labeled multigraph

Question 32

query formulation. User selects...

Answer 32

direct concepts of interest in conceptual graph

Question 33

QF: user is able to add to its selection

Answer 33

indirect concepts of interest

Question 34

news is time

Answer 34

stampled

Question 35

query execution:

Answer 35

generate query in semantic query language and execute the semantic query

Question 36

sort news items based on..

Answer 36

their relevance degree to the query

Question 37

results presentation, per each news item show a...

Answer 37

summary

Question 38

Hermes news portal HNP is an implementation of..

Answer 38

the hermes framework

Question 39

precision=

Answer 39

concepts correctly identified in news items/# of concepts identified in news items

Question 40

recall=

Answer 40

concepts correctly identified in news items/ # of concepts that should have been identified in the news items

Question 41

Hermes framework presents ..

Answer 41

news items that match user interests

Question 42

hermes cornerstone:

Answer 42

domain ontology anchored in a semantic lexicon

Question 43

HNP based on

Answer 43

wordnet semantic lexicon, owl ontology,, sparql queries etc.

Question 44

event =

Answer 44

rela between 2 entities in text

Question 45

application of event extraction in..

Answer 45

news classification and KBU

Question 46

event extraction languages:

Answer 46

lexico syntatic(uses lexical and syntatic info to extract events), lexico semantic (also uses semantic info from domain ontology)

Question 47

why hermes phases have to happen in certain order?

Answer 47

in order to update knowledge base, it's important to know which concepts from the knowledge base have been already found in the news. In order to query the news, its important to have the most up to date knowledge base view on the real world. As news querying uses a boolean query, in the results representation its needed to rank the query output based on the query generalized disjuncitve semantics and present the results in a user visually appealing manner

Question 48

3 owl features needed for modelling the domain ontology

Answer 48

``` cardinality restrictions (hasceo has cardinality 1) property symmetry (competitor is a symmetric property) property inverse (inverse of hasceo is isceoooof) ```

Question 49

why hermes uses a beta is 0.3 in F_beta

Answer 49

in hermes rule learning its given priority to precision over recall when learning 1 rule, as the group of learned rules per event will ensure a high recall

Question 50

difference lexico syntatic and lexico semantic pattern language to info extraction?

Answer 50

lexico sem pattern utilizes concepts from domain ontology which makes it easier to build and maintain patterns due to their increased expressivity and conciseness compared to their lexico syntactic counterparts

Question 51

performance of lexico syntactic and lexico semantic rules is evaluated based on..

Answer 51

rule creation times, precision, recall and F1 measure

Question 52

f1 measure =

Answer 52

harmonic mean within precision and recall

Question 53

groups of rules extract...

Answer 53

specific events

Question 54

rule learning using genetic programming termination criteria

Answer 54

rule: fixed number of its (keep best rule) | rule group: max number of its didnt improve rule group

Question 55

fitness of a rule has values

Answer 55

alpha and beta

Question 56

by beta smaller than 1 we give more importance to..

Answer 56

precision