Research Tool Materials

Hi YC!

Here are some materials demonstrating various aspects of Research Tool, a new kind of analytics platform. We wanted to demonstrate the scope and maturity of our software as well as provide some insight into our query language, Narrative Dynamics Language (NDL), and the associated ontology.

Update August 11: We've added new analytics capabilities to NDL and the ontology, adding two new analytic arcs. This gives us these arc definitions:

1. Analytic Arc: Models the dynamics of a single phenomenon (What is changing? How fast is it changing?).
2. Correlative Arc: Models the relationship between different phenomena (What is changing together?).
3. Diagnostic Arc: Models the evolution of understanding about a phenomenon (Why did our interpretation of it change?).

These new arcs enable queries that are impossible in other systems because they are queries about the reasoning process itself.

• Causal Analysis: Consider the query, "Find all instances where a symptom was re-classified... immediately following a severity increase..." An analyst can now find all the points where a quantitative change led to a qualitative shift in understanding.
• Ontology Evolution: Consider the query "Show me how many 'Market Risks' were re-classified as 'Regulatory Risks' after a specific law was passed..." This enables an analyst to measure the impact of a real-world event on the conceptual framework used by the organization.
• Predictive Modeling: The system can now learn the quantitative precursors to a qualitative shift. By training a model on verified Diagnostic Arcs, the platform could learn to predict, for example: "When a company's debt-to-equity ratio (quantitative metric) increases by more than 40% in two quarters (dynamic on a Thematic Spine), there is a 75% probability that human analysts will re-classify its risk profile from 'Stable' to 'Volatile' within the next 30 days."

The docs are lagging because we are building and testing the code. I'll update them this week, I hope.

Files

Here I've included presentation that introduces the platform and its concepts, as well as our query language specification and associated ontology. The goal is to demonstrate the depth of the project.

Software overview

A brief presentation showing the overall structure of Research Tool's microservices.

Inside Neo4j

A look inside our graph database. We parsed two books in our tests, a Biology Textbook and "Origin of the Species." This gives you an ideas of how the structure of a document is reflected in the graph. Sorry for my whisper-y voice, I recorded it while my family was napping.

Inside ElasticSearch

Our tool reflects the structure of documents in a graph database and associated text in ElasticSearch. They are connected by a unique_id.

Inside Blob Storage

We use GCS to store documents to be parsed and placed in the system. The parser also stores images from the parsed documents in GCS.