1.
Diagram of
Chapter 38
Natural Language Understanding
/^^^^^^^^\ Ghost AI Natural Language Understanding /^^^^^^\ / EYE \ / EAR \ / \ ____________ / ________ \ | | | | / \ | / \ | | _______ | | | / \-->-->--( AudListen)| | / old \ | | | / \ | \________/ | | / image \--|----+ ( AudInput ) | ______ | | \ engram / | a|c|f \ / | / \ | | \_______/ | b|o|i \ /-->-->-->( AudMem ) | | | s|n|b \____________/ | \______/ | | visual | t|c|e | | | | | | | r|e|r _______V__ __V_______ | ___V____ | | memory | a|p|s / \ / \ | / \ | | | c|t| ( OldConcept )( NewConcept ) |( AudRecog )| | channel | t| | \__________/ \__________/ | \________/ | | | | | \ __________ / | | | ______ | | | \/ EnParser \/ | "JOHN" | | / \ | | | \__________/ | "GIVES" | | / new \ | | | | | "THE" | | / image \ | |_|_ _____V_______ | "BOY" | | \ engram / | / \ / \ | "A" | | \ / | \ Psy /<--( InStantiate ) | "BOOK" | | \______/ | \___/ \_____________/ | |
2. Process of
Natural Language Understanding
The AudListen module is alert for auditory input.
Listening for input can be passive, as in the ghost.pl AI and MindForth for robots, or event-driven, as in the JavaScript Ghost.html AI.
AudInput sends phonemic characters to the AudMem module.
AudRecog recognizes concepts already known to the AI.
AudInput sends recognized concepts to OldConcept and unrecognized concepts to NewConcept.
EnParser calls InStantiate to create a new node of a concept.
InStantiate creates a time-bound instance of a concept.
EnParser retroactively creates associative tags.
3. Proof of
Natural Language Understanding
Once a ghost-class AI Mind has understood a sentence of input in natural language, the asking of questions about the knowledge in the sentence and the discussion of the idea in the sentence can ddemonstrate proof of Natural Language Understanding. The SpreadAct module for spreading activation makes the AI able to respond properly to various formats of questions about an understood sentence.
4. Problem of
Natural Language Understanding
According to Wikipedia, Natural Language Understanding is an AI-hard problem.
http://en.wikipedia.org/wiki/Natural-language_understanding
http://cyborg.blogspot.com/2017/09/pmpj0926.html -- Using NLU to answer questions
http://mc.ai/first-working-artificial-intelligence-solves-nlu
NLU problem-article:
https://www.theatlantic.com/technology/archive/2018/01/the-shallowness-of-google-translate/551570/