logo

Home | Products | Research | Projects | History

We are expanding and migrating into web apps this suite of very fast English text "understanding" tools. Their names (left side) link to functional summaries below.

Each tool uses a fixed codebase for syntax, plus related semantic extensions (right side).  Our core memory manager (yellow) uses a cohesive set of application-specific Topic Maps to define/guide/document both new (green) modules:


English-like Syntax
Semantics via Metadata
TRANSCRIBER Phrase Structure PARSER
  • Grammar Rules
  • Rule Application
  • Syntax Structures
Case Frame THESAURUS
  • Preposition Models
  • Verb Complements
  • Inglish Associations
MODELER Agent Training in WORDS
  • 1-to-1 Web Dialogs
  • Editing of Lexicons
  • Speech Act Scripts
Topics in CONTEXT
  • Discourse Models
  • Semantic Constraints
  • Chart Production
LEXICON Tokenizing SCANNER
  • Lexical Morphology
  • Syntactic Features
  • Multitoken Names
Token INTERPRETER
  • Expected Concepts
  • Driven by a Context
  • Sets Token Meaning

TRANSCRIBER - analyzes text structures in high-speed memory

The PARSER (yours or ours) structurally diagrams sentences and phrases. Ours is fast (semi-deterministic) with a decent grammar - a package designed to find clause structures unambiguously.  Overall parsing complexity rises fast if yours demands plausibility tests of implied semantic interactions, which usually can only prune the set of "final" situation models emitted by our multi-stage analysis.

Our THESAURUS option lets SYNTAX structures of interest be mapped from such surface semantics into DISCOURSE models tracking anaphora.  Logic processes using Common Logic or Owl can use such data to model what was said in a text.  Warning: such output may cost 50 times more than what was discussed, which can be found by simpler tools below.

MODELER - manages topic models within a mid-term memory

This is a listener for WORDS, a topic-modeling language that can express the cyclic nodes found in any syntax analysis. Used in standalone mode, it can also interactively define custom vocabulary extensions needed for your domain.

Either way, CONTEXT models the result per our ontology for Idealized English, which covers a discourse model for handling anaphora, plus a very powerful set of open-source semantic constraint patterns that can formally associate or annotate topics denoting whatever subjects WORDS expressions have declared.

LEXICON - holds needed lexical data within long-term memory files

Our SCANNER utility maps an English paragraph (posted or read) into a stream of world-class lexical data supporting linguistic analysis. Each token shows its root,  inflection if any, part-of-speech, and powerful binary linguistic features that can guide a parser to properly handle each word's expected complement patterns.  Under "MEANS", it assigns as candidate meanings Roget's category options.

Our INTERPRETER culls lexical ambiguity from inputs very early using contextual expectations, optionally aided by curation.  Our standalone Agency framework for semantic text analysis, search, summarization, alerting and word-learning uses this as its core.  Other NLU options above, if added, therefore run deterministically to boost accuracy and functionality without degrading net analysis speed.



Lexikos Corporation
Boston & Knoxville
Email: Dan@Lexikos.com