NLP

NLP
NLP

NLP is a difficult challenge, as it demands the intimate interaction of many different techniques - lexing, collocations, parsing, semantics, structure building - all using the same knowledge representation (obviously what humans do), and all happening together. These processes cannot function effectively in isolation, so a technology that can span and represent them all is necessary. The following material shows how Active Structure can achieve that. The crucial difference between this approach and all others is that here the structure is "soft" enough (and visible enough) to be modified by what it reads - the structure is actualised, not an algorithm or a rule.

pearls.jpg (232090 bytes)

Collocated Verbs: Different ways that collocated verbs are handled.
Is Mining of Knowledge Possible?: The reasons why Data Mining is inappropriate for knowledge-rich text.
Large parsing model: Text of model used to analyse legal documents
Extension to Active Structure to Handle Free Text: How much needed to be changed
Mathematics of the Law: Reading legal text and creating a valid logical/mathematical structure
Contract Driven Enterprises: Turning contract documents into active knowledge
Contract Clause Example: Converting a clause into Active Structure
General Legal: Beginning of representation of legislation
Outlining: Looking further afield, over the heads of adjacent symbols
Above the Fray: Eliminating ambiguity using constraint reasoning as the final step
The Semantic Octopus: Making relations do more of the work.
PipeLine to Failure in BioInformatics: What is wrong with pipelining
More on Prepositional Maps: Details of maps, and their uses
Relation Structures: Structures build into patterns.
Embedded Lists: Handling list structures embedded in a sentence using indirection.
Dynamic Insertion: Inserting symbols in the parse chain that are implied by their surroundings.
Active Structure Grammar: The difference between a static view of grammar as an end in itself, and active structure grammar as a system component with intermingling of many simultaneous operations, including changing the grammatical structure
Active Structure in Law: Applying active structure to the representation of legal text.
Building Structure: Shows how structures built from words are manipulated to reveal the sentence structure.
Constraint Reasoning in Parsing: Dynamic building and solving of a constraint problem as a means of NLP.
Cutting and Healing: Putting the pieces back together in a more meaningful way after cutting them apart.
Prepositional Chains: Chains as a way that dense scientific or legal text builds a microcontext.
Searching: Searching structures transformed from text using search structures transformed from search text.

Presentations

	Searching the Internet
	Searching Free Text
	Accessing Knowledge in Free Text
	Is Mining of Knowledge Possible?
	Words To Knowledge - an overview
	Prepositional Chains
	Conjunctions

	Connections and Meanings - the same word can have many meanings
	Relations - treating Relations like Entities
	Wiring Up - creating the structure underlying the text
	Object Groups - properties from a group of objects

Design Notes

Network Components

Some operators for handling free text:

NLP Specific Operators

Shells

Some operators for handling the structures that free text builds:

Hierarchy Specific Operators

Related Areas

Graphical Structure Building Tools

Orion KM

Searching an Active Ontology: Using Active Structure as a dynamic ontology, and then searching through it for information

loom.jpg (266229 bytes)

The diagram is modelled after the Jacquard loom, and attempts to show how a structure can read words and build structure, which becomes part of the structure used to read more words.