|
|
The aim of this project is extraction of temporal, spatial and causal
information from web pages. This information can then be exploited by a
web-based Question-Answering(QA)-system. The system should respond to
natural language question with answers and possibly further questions for
clarification or further specification. The types of questions will focus on
temporal, spatial and causal aspects: On which day was the general
election 2002 in Germany held? What is the name of the chancellor? Who
was chancellor in 1963? Where was the capital of the Federal Republic of
Germany in 1989? Why did the coalition of social-democrats and liberals
end?
|
|
In order to generate high-quality answers, it is very important to capture
the described events and to derive the expressed temporal relations between
these events from the news messages. Previous approaches to QA cannot
deliver this feature of the extracted data. The importance of extracting
such detailed temporal information will be briefly shown at the following
example: The consultations about the coalition started already on
Monday after the general election. The temporal anchoring of
this statement is done via the Prepositional Phrase (PP) after the
general election. Further cues for the temporal information
to be extracted are began, already
and on Monday. In order to derive the meaning of the
sentence and relate this meaning to the overall context, the following
inferences have to be drawn: (1) an event of type general election took
place (2) a temporal anchoring of the consultation is done. After having
extracted this and further information, the following question can
successfully be answered: "When did the consultation about the
coalition of the last general election took place?" |
|
