Skip to content
AL | Apatheia Labs
T

Temporal Parser

Timeline Construction & Analysis

Constructs unified timelines from scattered date references. Identifies gaps, overlaps, and impossible sequences that indicate error or manipulation.

Date ExtractionTimeline ConstructionGap DetectionSequence ValidationDeadline Tracking

The Problem

Dates are scattered across hundreds of pages in inconsistent formats. A report dated 15 March references an assessment from 22 March — a temporal impossibility buried in volume. Manual timeline construction is slow, error-prone, and collapses under the weight of large document sets. Critical gaps and impossible sequences go undetected.

How It Works

  1. 1Extract all temporal expressions with context
  2. 2Normalize to ISO 8601 format with confidence scores
  3. 3Link events to their temporal anchors
  4. 4Build event sequence graphs
  5. 5Validate against logical constraints

Inputs

  • • Document corpus
  • • Temporal expressions
  • • Event references

Outputs

  • • Unified timeline
  • • Gap analysis report
  • • Anomaly findings

What You Get

TEMPORAL CONTRADICTION DETECTED

Event A: Assessment report authored
  Date: 15 March 2023 | Source: Doc 12, p.3 | Author: Dr. J. Smith

Event B: Assessment session conducted
  Date: 22 March 2023 | Source: Doc 14, p.1 | Participants: Dr. J. Smith, Client A

Finding: Report (Event A) predates the assessment it describes (Event B) by 7 days.
CASCADE type: TEMPORAL | Severity: HIGH

Gap detected: 14 June 2023 – 9 September 2023
  No recorded activity for 87 days
  Statutory deadline (45-day review) exceeded
  Last actor: Social Worker B

Works With

EEntity Resolution

Provides the identity layer so the Temporal Parser can build per-person timelines and track who did what when.

KContradiction

Consumes timeline data to identify TEMPORAL type contradictions — events that couldn’t have happened in the stated order.

LAccountability

Uses deadline tracking to identify where statutory timeframes were breached and which actors were responsible.

Use Cases

Police investigation review

Reconstructing the sequence of witness statements, forensic results, and investigative decisions to identify where the timeline diverges from what was reported to the court.

Contract dispute analysis

Tracking amendment dates, notice periods, and performance milestones across hundreds of communications to identify breached deadlines.

Medical record sequencing

Building a unified patient timeline from GP notes, hospital records, specialist referrals, and pharmacy dispensing logs to identify treatment gaps or contradictory clinical accounts.

Technical Approach

  • Date extraction combines regex patterns for structured dates with NLP models for natural language temporal expressions
  • ISO 8601 normalization resolves locale ambiguity using document metadata, institutional conventions, and cross-referencing against anchor dates
  • Event-graph construction using directed acyclic graphs with temporal constraint edges, enabling both forward and backward sequence traversal
  • Anomaly detection via constraint satisfaction — the engine identifies configurations where no valid temporal ordering exists