[Paper Review] KBSET -- Knowledge-Based Support for Scholarly Editing and Text Processing

Jana Kittelmann, Christoph Wernhard|arXiv (Cornell University)|Aug 29, 2019

Semantic Web and Ontologies4 citations

TL;DR

KBSET proposes a Prolog-powered, LaTeX-based framework for scholarly editing that integrates semantic markup, named entity recognition, and external knowledge bases like GND and DBpedia. It enables efficient, logic-based text processing with real-time candidate suggestions, achieving entity resolution in ~7 seconds for large corpora, and supports extensible, maintainable scholarly workflows through declarative, query-driven knowledge integration.

ABSTRACT

KBSET supports a practical workflow for scholarly editing, based on using LaTeX with dedicated commands for semantics-oriented markup and a Prolog-implemented core system. Prolog plays there various roles: as query language and access mechanism for large Semantic Web fact bases, as data representation of structured documents and as a workflow model for advanced application tasks. The core system includes a LaTeX parser and a facility for the identification of named entities. We also sketch future perspectives of this approach to scholarly editing based on techniques of computational logic.

Motivation & Objective

• Address the lack of practical, scalable support for scholarly editing that integrates semantic knowledge and external fact bases.
• Develop an economic, extensible workflow for humanities scholars to create, validate, and present annotated scholarly texts.
• Enable seamless integration of external knowledge bases (e.g., GND, DBpedia) with structured, queryable markup in LaTeX.
• Support advanced text processing tasks such as named entity recognition and semantic reasoning using computational logic.
• Provide a flexible, logic-based infrastructure that decouples markup from presentation and enables reusable, composable annotations.

Proposed method

• Uses LaTeX with domain-specific semantic commands for structured, human-readable markup of scholarly texts, especially correspondence editions.
• Employs a Prolog-based core system to parse LaTeX, represent documents and annotations, and manage knowledge integration.
• Integrates external fact bases (GND, GeoNames, DBpedia, YAGO) via Prolog queries for entity resolution and semantic enrichment.
• Applies feature-based ranking of candidate entities using plausibility heuristics such as name similarity, occupation match, and contextual consistency.
• Supports incremental refinement of entity suggestions through user feedback, updating the assistance document and re-evaluating queries.
• Uses Prolog’s query optimization and logical inference to enable efficient, extensible, and maintainable text processing pipelines.

Experimental results

Research questions

• RQ1 How can a scalable, practical workflow for scholarly editing be designed that integrates semantic markup with external knowledge bases?
• RQ2 To what extent can Prolog serve as a unifying substrate for document representation, query processing, and workflow modeling in scholarly editing?
• RQ3 Can feature-based ranking of named entity candidates improve accuracy and performance in large-scale text processing?
• RQ4 How can logic-based systems mediate between automatically generated and manually curated annotations in scholarly editions?
• RQ5 In what ways can computational logic support extensible, composable, and maintainable scholarly text processing beyond traditional XML/TEI models?

Key findings

• KBSET enables named entity recognition on a 10,000+ query workload in approximately 7 seconds on a modern notebook, demonstrating high performance for large-scale scholarly text processing.
• The system successfully supports the edition of a 2,000-page scholarly correspondence project (Sulzer-Bodmer), proving practical viability in real-world digital humanities applications.
• Prolog serves effectively as a query language, data model, and workflow engine, enabling flexible integration of knowledge bases and dynamic query re-evaluation.
• Feature-based plausibility ranking of entity candidates, combined with contextual and semantic features (e.g., occupation match, Wikipedia linkage), significantly improves accuracy and user guidance.
• The system supports incremental refinement of annotations through user feedback, allowing the assistance document to be updated and reloaded for improved future results.
• KBSET’s architecture enables decoupling of markup from presentation and supports extensible, aspect-oriented-style annotation weaving, suggesting strong potential for future extensibility.