Inferring High-Level Events from Timestamped Data:...

Inferring High-Level Events from Timestamped Data:... | AI Research

Key Takeaways

This paper introduces a new logic-based framework designed to transform raw, timestamped clinical data into meaningful, high-level events.
In this paper, we develop a novel logic-based approach to detecting high-level temporally extended events from timestamped data and background knowledge.
Our framework employs logical rules to capture existence and termination conditions for simple temporal events and to combine these into meta-events.
As some incorrect events might be inferred, we use constraints to identify incompatible combinations of events and propose a repair mechanism to select preferred consistent sets of events.
While reasoning in the full framework is intractable, we identify relevant restrictions that ensure polynomial-time data complexity.

Paper AbstractExpand

In this paper, we develop a novel logic-based approach to detecting high-level temporally extended events from timestamped data and background knowledge. Our framework employs logical rules to capture existence and termination conditions for simple temporal events and to combine these into meta-events. In the medical domain, for example, disease episodes and therapies are inferred from timestamped clinical observations, such as diagnoses and drug administrations stored in patient records, and can be further combined into higher-level disease events. As some incorrect events might be inferred, we use constraints to identify incompatible combinations of events and propose a repair mechanism to select preferred consistent sets of events. While reasoning in the full framework is intractable, we identify relevant restrictions that ensure polynomial-time data complexity. Our prototype system implements core components of the approach using answer set programming. An evaluation on a lung cancer use case supports the interest of the approach, both in terms of computational feasibility and positive alignment of our results with medical expert opinions. While strongly motivated by the needs of the healthcare domain, our framework is purposely generic, enabling its reuse in other areas.

This paper introduces a new logic-based framework designed to transform raw, timestamped clinical data into meaningful, high-level events. While electronic health records contain vast amounts of data—such as individual drug administrations or test results—they often fail to explicitly document the underlying clinical phenomena, like a specific disease episode or a course of therapy. This framework provides a systematic way to bridge that gap, allowing researchers and clinicians to infer these hidden events and represent them in a way that aligns with medical reasoning.

Detecting Events from Data

The framework categorizes events into two types: "simple events" and "meta-events." Simple events are derived directly from raw data using existence and termination conditions. For example, a series of antibiotic administrations can be grouped to infer an "antibiotic therapy" event. The system distinguishes between "persistent" events, which continue until a specific stop condition is met, and "non-persistent" events, which require regular observations to remain active. Meta-events are then built on top of these simple events, allowing for higher levels of abstraction, such as identifying concurrent treatments or complex disease progressions.

Ensuring Consistency and Reliability

Because automated inference can sometimes produce incorrect or conflicting results, the framework includes a repair mechanism. It uses constraints to identify incompatible event combinations—such as a patient being assigned two mutually exclusive therapies simultaneously—and selects the most consistent set of events. Additionally, the system assigns confidence levels to inferred events based on the rules that generated them, helping users prioritize more reliable information.

Computational Feasibility

The researchers acknowledge that reasoning across the full, complex framework is computationally demanding. However, they identified specific restrictions that allow the system to operate with polynomial-time data complexity, ensuring it remains practical for real-world use. To demonstrate this, they implemented a prototype using answer set programming (ASP).

Medical Application

The approach was evaluated using a lung cancer use case involving actual hospital data. The results showed that the system is computationally feasible and, importantly, the inferred events aligned positively with the assessments of medical experts. While the framework was developed to meet the specific needs of the healthcare domain, its design is intentionally generic, meaning it can be adapted for event detection in other fields beyond medicine.