23 Entity-Event Ontology Construction by Conceptualization …
347
time and place, it is possible to analyze statements like “John said yesterday that
Sandy is going to be in Portland tomorrow”. If opinion was published July 4,
then first event is “Sandy visiting Portland on July 4”, and the second one is
quotation event published on July 4 and containing John’s opinion as of July 3
and a reference to the first event.
• Temporal analysis requires culture detection because depending on language,
hemisphere, and other contextual data, it is difficult to interpret dates, time of
year, the first day of a week, and so on [7].
• When working with constantly changing data, it is important to preserve fragments
of text with mentions of event. For example, mainstream media sources often edit
news articles—and being able to see this is useful for analysts.
• Sentiment analysis is not related to ontologies, but maintaining author sentiment
for every mention is convenient for analysts. For example, it enables analysis of
general sentiment regarding authors from specific countries or specific types of
media.
• Sometimes there is no rule or event type available for linking adjacent entities
in text, but for practical reasons, this kind of links should also be preserved and
presented to users. It can be done with a special attribute “related entities” in every
event.
• Ontologies and knowledge bases are often viewed as something static, but as they
have the power to be the ultimate source of data in knowledge-based systems, it
is wise to continuously update them as new data arrive and populate them with
new facts. For instance, ontologies can easily be a data source for an interactive
dashboard or search system.
• In practice, it is often better to implement knowledge bases on top of conventional technologies like document-oriented or relational databases and encapsulate them through API. For some reason, software development engineers are
sometimes biased toward technologies like RDF, OWL, SPARQL, logic programming, Prolog, first-order logic, and try to avoid them. As these technologies
are seldom mainstream, they obstruct widespread adoption of ontologies. Thus,
researchers need to communicate that knowledge-based systems are not tied to
research community tools and can make use of conventional technologies as well.
23.4 Implementing Ontology Construction for the Internet
Internet and other sources nowadays are acting as a mirror of real-world events: Every
second, countless people send and describe things that they experience, companies
upload and produce tons of documents, armies of journalists and bloggers interpret
and follow events, conduct citizen investigations, and analyze different sources. For
example, every minute, Twitter users send more than 511,000 tweets, Tumblr users
publish 92,000 posts, 188,000,000 million emails are sent, and 277,000 Instagram
stories are posted [8].
347
time and place, it is possible to analyze statements like “John said yesterday that
Sandy is going to be in Portland tomorrow”. If opinion was published July 4,
then first event is “Sandy visiting Portland on July 4”, and the second one is
quotation event published on July 4 and containing John’s opinion as of July 3
and a reference to the first event.
• Temporal analysis requires culture detection because depending on language,
hemisphere, and other contextual data, it is difficult to interpret dates, time of
year, the first day of a week, and so on [7].
• When working with constantly changing data, it is important to preserve fragments
of text with mentions of event. For example, mainstream media sources often edit
news articles—and being able to see this is useful for analysts.
• Sentiment analysis is not related to ontologies, but maintaining author sentiment
for every mention is convenient for analysts. For example, it enables analysis of
general sentiment regarding authors from specific countries or specific types of
media.
• Sometimes there is no rule or event type available for linking adjacent entities
in text, but for practical reasons, this kind of links should also be preserved and
presented to users. It can be done with a special attribute “related entities” in every
event.
• Ontologies and knowledge bases are often viewed as something static, but as they
have the power to be the ultimate source of data in knowledge-based systems, it
is wise to continuously update them as new data arrive and populate them with
new facts. For instance, ontologies can easily be a data source for an interactive
dashboard or search system.
• In practice, it is often better to implement knowledge bases on top of conventional technologies like document-oriented or relational databases and encapsulate them through API. For some reason, software development engineers are
sometimes biased toward technologies like RDF, OWL, SPARQL, logic programming, Prolog, first-order logic, and try to avoid them. As these technologies
are seldom mainstream, they obstruct widespread adoption of ontologies. Thus,
researchers need to communicate that knowledge-based systems are not tied to
research community tools and can make use of conventional technologies as well.
23.4 Implementing Ontology Construction for the Internet
Internet and other sources nowadays are acting as a mirror of real-world events: Every
second, countless people send and describe things that they experience, companies
upload and produce tons of documents, armies of journalists and bloggers interpret
and follow events, conduct citizen investigations, and analyze different sources. For
example, every minute, Twitter users send more than 511,000 tweets, Tumblr users
publish 92,000 posts, 188,000,000 million emails are sent, and 277,000 Instagram
stories are posted [8].