Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

ORACLE: Time-Dependent Recursive Summary Graphs for Foresight on News Data Using LLMs

Lev Kharlashkin, Eiaki Morooka, Yehor Tereshchenko, Mika Hämäläinen

TL;DR

ORACLE presents a production-ready pipeline that converts daily Finnish news into week-over-week, decision-ready foresight for a Finnish University of Applied Sciences. The core idea is a two-level Time-Dependent Recursive Summary Graph (TRSG) that accumulates long-term knowledge, uses semantic clustering, and employs recursive summarization to keep outputs compact and informative. A week-to-week change detector groups deltas into themes under PESTEL perspectives, enabling auditable traceability and actionable recommendations. The curriculum-intelligence use case demonstrates practical relevance and outlines an evaluation plan and future extensions to multilingual sources and policy–science–industry link analysis.

Abstract

ORACLE turns daily news into week-over-week, decision-ready insights for one of the Finnish University of Applied Sciences. The platform crawls and versions news, applies University-specific relevance filtering, embeds content, classifies items into PESTEL dimensions and builds a concise Time-Dependent Recursive Summary Graph (TRSG): two clustering layers summarized by an LLM and recomputed weekly. A lightweight change detector highlights what is new, removed or changed, then groups differences into themes for PESTEL-aware analysis. We detail the pipeline, discuss concrete design choices that make the system stable in production and present a curriculum-intelligence use case with an evaluation plan.

ORACLE: Time-Dependent Recursive Summary Graphs for Foresight on News Data Using LLMs

TL;DR

ORACLE presents a production-ready pipeline that converts daily Finnish news into week-over-week, decision-ready foresight for a Finnish University of Applied Sciences. The core idea is a two-level Time-Dependent Recursive Summary Graph (TRSG) that accumulates long-term knowledge, uses semantic clustering, and employs recursive summarization to keep outputs compact and informative. A week-to-week change detector groups deltas into themes under PESTEL perspectives, enabling auditable traceability and actionable recommendations. The curriculum-intelligence use case demonstrates practical relevance and outlines an evaluation plan and future extensions to multilingual sources and policy–science–industry link analysis.

Abstract

ORACLE turns daily news into week-over-week, decision-ready insights for one of the Finnish University of Applied Sciences. The platform crawls and versions news, applies University-specific relevance filtering, embeds content, classifies items into PESTEL dimensions and builds a concise Time-Dependent Recursive Summary Graph (TRSG): two clustering layers summarized by an LLM and recomputed weekly. A lightweight change detector highlights what is new, removed or changed, then groups differences into themes for PESTEL-aware analysis. We detail the pipeline, discuss concrete design choices that make the system stable in production and present a curriculum-intelligence use case with an evaluation plan.

Paper Structure

This paper contains 30 sections, 2 figures, 1 table.

Table of Contents

  1. Introduction
  2. Related Work
  3. Semantic representations and retrieval.
  4. Clustering and community detection.
  5. Summarization and abstraction.
  6. Temporal modeling and change detection.
  7. LLMs in foresight and strategic analysis.
  8. Oracle Platform
  9. Data Ingestion and Versioning
  10. Sources.
  11. Hashing.
  12. Relevance Filtering for University
  13. Embeddings, Storage and PESTEL
  14. Time-Dependent Recursive Summary Graph (TRSG)
  15. Cumulative Knowledge Base.
  16. ...and 15 more sections

Figures (2)

  • Figure 1: Oracle workflow overview.
  • Figure 2: Example of the L1 layer in the TRSG. Each node represents a cluster of semantically related news items and edges indicate similarity links. The summarization step condenses each cluster into a factual thematic report, forming a connected graph of emerging narratives.