Common sense knowledge and language
Language highlights and connects
concepts and relationships that people know from their experiences.
- The police arrested the demonstrators because they feared violence.
- The police arrested the demonstrators because they advocated violence.
- They drank two cups of tea because they were warm.
- They drank two cups of tea because they were cold.
- The box is in the pen.
- The pen is in the box.
- Time flies like an arrow.
- Fruit flies like an apple.
- Colorless green ideas sleep furiously.
- She saw the man in the park with a dog.
- She saw the man in the park with a statue.
- She saw the man in the park with a telescope.
These examples show that individual words are not so important for
understanding. A sentence evokes a mental image because people have
knowledge about the world. Language provides hints for constructing
mental images but does not provide a complete description of
- He opened the ... with the ... and walked inside.
- He ... the door with the key and ... inside.
- ... opened the door ... the key and walked ...
Therefore natural language understanding and processing requires
common sense knowledge and context.
What is "context"?
Can common sense knowledge and context be stored in a (formal) language?
If "yes", it can be stored in ontologies such as CYC.
- concepts and assertions
- context and micro-theories (eg. theories of space, temporal reasoning)
- CYC-L (logical inferences)
CYC's applications (as reported in 1995)
- information retrieval (eg. detailed user model)
- (semi-automatically) linking heterogeneous information sources
(eg. interface between two different databases)
- processing structured data (eg identify inconsistencies in databases)
- assist in word processing (spell checking, grammar, vocabulary,
word and sentence completion)
- assist in speech recognition
- simulations (eg. in adventure games)
- image retrieval using image captions
- machine translation