RTS <- SHRDLU NLP + speech recognition

bishop_pass · 2002-03-17T15:46:57

I think this would be great. Take off the shelf (university GNU parsing software using semantics and syntax), combine it with off the shelf speech recognition software (look into Sphinx) and command your troops using this technology. Has anyone ever read the transcripts produced by SHRDLU? SHRDLU dates to 1970 and used 64k. It is very achievable intelligence when confined to a domain such as SHRDLU''s. RTSs have domains which are perfectly suitable to such engines. Command your troops via speech recognition: YOU: "Royce, move your troops over the hill and wait." COMMANDER: "Roger that..." YOU: "Jensen, What is the situation with you guys?" JENSEN: "Sir, we''re holed up under enemy fire. We need reinforcements." YOU: "Baker, status?" BAKER: "The north drainage is providing cover for us at the moment. We''re moving north, ETA to the pass is 1 hour." YOU: "Why are you in the north drainage?" BAKER: "Sir, the south drainage was too exposed." YOU: "Johnson, I need air coverage for the 5th battalion at cooridnates... Enemy approaching from the west." ------ Details: Let''s look at the first example. You say: "Royce, move your troops over the hill and wait." The parser identifies Royce as interprets you message to be directed to Royce. ''Move'' is a verb and the parser interprets it as a command. ''Your troops'' is a noun phrase, ''over'' is a preposition, ''the hill'' is a noun phrase (a direct object) and ''wait is a successive command. ''The hill'' is ambiguous, but context should identify which hill is being discussed. Look into SHRDLU. You will be amazed at what is possible when limited domains are used. Remember, SHRDLU was programmed in 1970 and had a conceptual understanding of what was being discussed, and had knowledge and memory of discourse. Use Google.

Game Design and Theory Game Design

Started by bishop_pass March 14, 2002 11:55 PM

12 comments, last by bishop_pass 22 years, 10 months ago

bishop_pass

Author

112

March 15, 2002 11:55 AM

Correction: SHRDLU used 100 to 140k 36 bit words. See the manual for SHRDLU.

_______________________________
"To understand the horse you'll find that you're going to be working on yourself. The horse will give you the answers and he will question you to see if you are sure or not."
- Ray Hunt, in Think Harmony With Horses
ALU - SHRDLU - WORDNET - CYC - SWALE - AM - CD - J.M. - K.S. | CAA - BCHA - AQHA - APHA - R.H. - T.D. | 395 - SPS - GORDIE - SCMA - R.M. - G.R. - V.C. - C.F.

a person

118

March 15, 2002 07:52 PM

so ibm was correct, you only need 640k. heh.

bishop_pass

Author

112

March 16, 2002 03:24 PM

quote: Original post by Oluseyi
Care to port it from MacLisp?

A while ago I read about a project that was porting it from MacLisp. But I wasn''t necessarily advocating using SHRDLU for parsing, but merely showing what has been done. There are sophisticated parsers out there, but they are useless without domain knowledge.

SHRDLU integrated parsing with domain knowledge, discourse knowledge, and memory of past events, making it so effecitve.

Dauntless

314

March 17, 2002 03:46 PM

Like you said, I think one of the important concepts for something like this is totally identifying the problem domain and the knowledge concepts. Knowing what kinds of problems the computer will have to deal with, and the knowledge "library" the computer has is crucial to its problem solving capabilities. I think these two aspects are the most crucial to implementing something like this.

However, it still relies somewhat on hardwired rules. The computer can not solve problems that it does not have the rules (the knowledge concepts) for, or if it comes across a problem that it has never seen before (there is something missing in the problem domain). I''d like to see an AI implementation that can try to extrapolate or interpolate information on "things" it has not been exposed to before. Thought I suppose this is somewhat the holy grail of AI

The world has achieved brilliance without wisdom, power without conscience. Ours is a world of nuclear giants and ethical infants. We know more about war than we know about peace, more about killing than we know about living. We have grasped the mystery of the atom and rejected the Sermon on the Mount." - General Omar Bradley

RTS <- SHRDLU NLP + speech recognition

This topic is closed to new replies.

Popular Topics

Recommended Tutorials

RTS <- SHRDLU NLP + speech recognition

This topic is closed to new replies.

Popular Topics

Recommended Tutorials

Reticulating splines