It seems clear to me (after thinking about & studying both learning and learning systems recently – as well as participating in several, both online & through distance learning) that a proper learning system – as I envisage it & will create it – follows this basic process [stated as description of ALGORITHM(S) – note that this is in very rough note form only!]:
1. READ unstructured textual information & conduct CONCEPT or KEYWORD ANALYSIS > ie HIGHLIGHT key ideas/concepts as you go > this is dependent upon what is deemed RELEVANT at that TIME > this can come from a database of CONCEPTS’ aka keywords & key phrases, RANKED for RELEVANCY (needs to be seeded ie pre-populated by HUMAN but later can be BUILT UPON by AI thru RECURSIVE LEARNING); text is parsed, OCCURENCE of keywords & phrases counted & recorded, CONTEXT of key word/phrase occurrence is noted > basically the web search algo but applied to a specific knowledge domain
2. REPEAT stage 1 but replace Keyword Analysis with RELATIONAL or GRAMMATICAL ANALYSIS > strip out NOUNS, VERBS, DESCRIPTORS & FILLER words > further analyse NOUNS into SUBJECT & OBJECT (as linked by some ACTION (verb) > establish & record all CONNECTIONS aka RELATIONSHIPS between words in text > eg what subjects (noun1) are doing what ACTIONS (verbs) to what objects (noun2); what DESCRIPTORS are being applied by whom to what subjects, objects & actions > store in growing database
3. Pull out relevant PHRASES to delineate IDEAS in CONTEXT > how this is done algorithmically by AI is not yet known to me > cld be that text is ANALYSED TWO WAYS (as per steps 1 & 2 > Keyword Analysis + Grammatical Analysis) ways, COMPARED & recombined in a SIMPLIFIED & condensed STRUCTURE (equiv to process of creating a MIND MAP)
4. Repackage & PRESENT simplified & structured text back to learner for human review/processing/understanding
5. Algorithmic process may be repeated by any number of RECURSIONS or iterations > as AI goes deeper into analysis of text, more CONCEPTS (keywords/phrases) and RELATIONSHIPS (subject-action-object constructs) will be identified > these new concepts & relationships can be added to the database by the AI (with or without human review/approval)
6. Extensions:
- knowledge & understanding can be strengthened by CROSS-REFERENCING both CONCEPTS & RELATIONSHIPS across multiple knowledge DOMAINS to build up a stronger RELATIONAL UNDERSTANDING
- with access to CONCEPTS within one KNOWLEDGE DOMAIN, AI can search across that domain for EXAMPLES of APPLICATIONS of said CONCEPTS & possibilities for extension a given RELATIONSHIP between subject-object by EXPLORING/EXAMINING OTHER likewise RELATIONSHIPS across multiple knowledge domains