Abstract
The keyword-based searches perform unsatisfactorily because of their poor semantic identification, isolated output information and non-uniformed output format, which lead to hard manual screening. In order to overcome these drawbacks, under the challenges that the ambiguity of legal language, the deficiency of three-stage inference for court decisions and the limited role of cases in China, a circular ontology between normative documents and judicial cases is proposed in order to contribute open-textured legal concepts and improve the retrieval accuracy. Taking spiritual compensation in medical disputes as an example, we discuss the correlativeness of normative documents and the similarity of judicial precedents, and the paper offers a practical framework of ontology-based legal information retrieval.
Introduction
In the legal domain, retrieving precise information is critical. Traditional keyword-based search engines often struggle with the inherent ambiguity and "fuzziness" of legal language. A simple keyword search might overlook relevant cases that use synonymous phrasing or fail to account for the complex hierarchical relationships between different laws.
Within the Chinese legal system, a statute-based (civil law) jurisdiction, there is a growing need to bridge the gap between high-level legislation (normative documents) and their practical application in court (judicial cases). Currently, these two areas often exist as isolated information silos, placing a significant manual burden on legal professionals to find relevant precedents. To address this, researchers have developed a "circular ontology" framework that achieved an 80% retrieval hit rate—a 20% improvement over traditional methods.
Methodology
The core of this approach is a circular ontology framework that integrates two distinct types of legal knowledge:
- Normative Ontology (Statutes): Constructed using a "bottom-up" approach, this ontology extracts elements from legal principles, acts, regulations, and judicial interpretations. It maps the hierarchy of Chinese law, recognizing the varying levels of authority and practical "executive force."
- Case Ontology (Precedents): Developed via a "top-down" approach, it categorizes judicial precedents based on feature spaces such as "Basic Information" (person, time, place), "Legal Acts," "Factual Information" (infringement method, disability degree), and the final "Verdict."
The system utilizes ICTCLAS for high-precision semantic segmentation of Chinese text and Jena as the core ontology search engine. To determine case similarity, a Nearest-Neighbor Algorithm calculates a weighted sum of various features, allowing the system to identify legally similar cases even when wording differs. The "circular" nature of the framework refers to the feedback loop where retrieved rules inform case selection, and precedents, in turn, clarify the application of those rules.
Key Insights: The Hierarchy of Law
A significant observation in this research is the relationship between legal authority and practical application. While the Constitution sits at the top of the legal hierarchy, lower-level statutes and directives are often more "operable and foreseeable," carrying greater "executive force" in daily judicial practice.
By treating law as a web of relationships rather than isolated keywords, the ontology approach allows for semantic understanding. For instance, the system can recognize that a specific medical regulation is a subset of broader civil law, facilitating more accurate retrieval.
Experimental Results
The framework was tested on a sub-domain of spiritual compensation in medical disputes, utilizing a corpus of 256 verdict judgments. Comparison tests against traditional keyword search revealed:
- Hit Rate: The ontology-based approach achieved an 80% hit rate, compared to 60% for keyword-based methods.
- Accuracy: This represents a 20% increase in retrieval accuracy.
- Efficiency: The system demonstrated a reduced need for manual screening by providing semantically linked, highly relevant results.
Conclusion and Future Work
While the circular ontology framework shows great promise in improving legal information retrieval, challenges remain. The current model is domain-specific to medical disputes; scaling it to the entire legal system would require a massive expansion of the ontology. Additionally, the system still faces difficulties with highly generalized legal terms and complex reasoning. Future work will focus on expanding the database and implementing more advanced knowledge-based indexing methods to maintain performance at scale.
Summary updated