ATOM: AdapTive and OptiMized dynamic temporal knowledge graph construction using LLMs

TL;DR

ATOM tackles the challenge of building dynamic temporal knowledge graphs from unstructured text by introducing atomic fact decomposition and parallel merging with dual-time modeling. The framework splits input into atomic facts, extracts 5-tuples in parallel, and merges atomic TKGs in parallel to produce a DTKG, mitigating forgetting and improving exhaustivity and stability while achieving substantial latency reductions. Evaluations on the 2020-COVID-NYT dataset show improved temporal and factual exhaustivity and strong DTKG consistency, with significant scalability benefits over state-of-the-art zero-/few-shot methods. The approach offers a practical, scalable pathway for real-time, dynamic knowledge accumulation and temporal reasoning in large-scale text streams, with potential future gains from LLM fine-tuning and supervised entity/relation resolution.

Abstract

In today's rapidly expanding data landscape, knowledge extraction from unstructured text is vital for real-time analytics, temporal inference, and dynamic memory frameworks. However, traditional static knowledge graph (KG) construction often overlooks the dynamic and time-sensitive nature of real-world data, limiting adaptability to continuous changes. Moreover, recent zero- or few-shot approaches that avoid domain-specific fine-tuning or reliance on prebuilt ontologies often suffer from instability across multiple runs, as well as incomplete coverage of key facts. To address these challenges, we introduce ATOM (AdapTive and OptiMized), a few-shot and scalable approach that builds and continuously updates Temporal Knowledge Graphs (TKGs) from unstructured texts. ATOM splits input documents into minimal, self-contained "atomic" facts, improving extraction exhaustivity and stability. Then, it constructs atomic TKGs from these facts while employing a dual-time modeling that distinguishes when information is observed from when it is valid. The resulting atomic TKGs are subsequently merged in parallel. Empirical evaluations demonstrate that ATOM achieves ~18% higher exhaustivity, ~17% better stability, and over 90% latency reduction compared to baseline methods, demonstrating a strong scalability potential for dynamic TKG construction.

Figures (7)

• Figure 1: ATOM’s architecture, running in parallel, ensuring scalability, speed, and continuous updates. Unstructured texts observed at time $t$ are denoted by $D_t$, the $i$-th temporal atomic fact observed at time $t$ is denoted by $f_{t,i}$, the $i$-th atomic TKG observed at time $t$ is denoted by $\mathcal{G}^t_i$, the TKG snapshot observed at time $t$ is denoted by $\mathcal{G}_{s}^t$, and the updated TKG at time $t$ is denoted by $\mathcal{G}^t$.
• Figure 2: Exhaustivity vs. token count as context for (a) the atomic fact decomposition (b) 5-tuples extraction.
• Figure 3: Latency comparison of the baseline methods as a function of the number of atomic facts as input.
• Figure F.1: Example overview of ATOM's pipeline. It begins with atomic fact decomposition, followed by the extraction of atomic TKGs from these facts, which are then merged in parallel. When an incoming update arrives, ATOM handles the temporal resolution by transforming the end action into the affirmative part while modifying only the $t_{end}$, then merges the resulting atomic TKG with the existing DTKG.
• Figure F.2: Two DTKGs constructed using ATOM and Graphiti from 09-01-2020 (in UNIX, 1578524400) to 23-01-2020 (in UNIX, 1579734000) from 2020-COVID-NYT dataset. Left (ATOM): Preserves observation times ($t_{obs}$) separately from validity periods, with timestamps encoded in UNIX format to eliminate overhead associated with string parsing operations and timezone conversion calculations. Right (Graphiti): Treats observation time as validity start time. $valid\_at$ corresponds to $t_{start}$ in Graphiti's time modeling. The highlighted fact "The mysterious respiratory virus spread to at least 10 other countries" is observed on 23-01-2020, but this does not guarantee the spread occurred at that time. ATOM's dual-time modeling prevents such temporal misattribution.

Theorems & Definitions (3)

• Definition 1: Dynamic-Temporal KG with Dual-Time Modeling
• Definition 2: Temporal Atomic Fact with Dual-Time Modeling
• Definition 3: Atomic Temporal KG