Why is Natural Language Processing hard?
Many people ask why computers cannot understand what they want to do.
Also, junior programmers may think upon venturing into natural language
that it should be easy. From my experiences with this project, here
are some reasons why processing natural language has proven to be
- Humans only say the least amount possible to get the point across.
Therefore humans may leave large parts of the idea
to the listener to fill in. See Ellipsis module. Also, humans have a
large pool of similar experiences that the speaker
uses to use less language in a conversation.
- Humans have vast experiences with other forms of information input,
such as hearing, sight, touch, ... Computers at this time do not have
this outside connection which means the computer must be told everything
about the outside, phsical world in order to make common sense judgments.
- Words can have multiple meanings.