Communicating and Language
1. The Problem
communication among systems, as speech act (input, processor, output), in a common language
levels in a language: syntax/grammar, semantics, pragmatics
natural vs. artificial languages; natural language processing (NLP): understanding and generating
multi-agent community: overall behavior
2. Proposed Solutions
Symbolic NLP
- equipped with grammar and dictionary
- understanding/comprehension: syntactic analysis → semantic interpretation → incorporation or response
- generation/production: content selecting → content expressing, using templates
Statistical NLP
- start with corpus, i.e., collection of actual text
- learning relevant probability distributions
- language processing as pattern association
NLP-related topics
Multi-agent system: interaction, distribution, common beliefs, social conventions, culture, science
3. Issues
many-to-many relationship between levels
language learning: special-purpose or general-purpose
individual intelligence and collective intelligence: similarity and difference
NLP and AI, Turing Test
4. Reading
Chapters 22, 23, Section 17.6
5. ideas
NLP by general-purpose reasoning:
- treat words/phrases/sentences as (linguistic) concepts
- relations among linguistic concepts: sequential and correlational; relations among native (non-linguistic) concepts: inheritance/similarity/implication/equivalence; relation between the two groups: symbolizing
- symbolizing relation: many-to-many, multi-values, learned from experience, prioritized
- NLP as goal-directed reasoning on linguistic concepts
- mixture of syntax, semantics, and pragmatics; rather than sequential, binary, many-to-one, or reductionist
- incremental, real-time, and context-sensitive processing of linguistic and native concepts; rather than batch processing of corpus
- NLP as part of thinking, rather than modular and stand-alone
Communication and Socialization:
- innate knowledge (instinct) vs. acquired knowledge (intelligence)
- knowledge from communication, directed (talking) and indirected (reading)
- objective knowledge and social convention as overlapping subjective experience
- education and training as goal-directed experience control, implanting: engineering shortcut
- "friendly AI": built vs. taught
Collective intelligence:
- a community as an intelligent system; common goals (values), knowledge (science), skills (technology), operations (tools), and language
- evolution of the items follow the same "logic" as an individual intelligent system