Evaluating Semantic Search for 18th-Century Intellectual History via Locke
A new academic paper published on arXiv evaluates the efficacy of semantic search technologies in analyzing 18th-century intellectual history, specifically focusing on the reception of John Locke’s foundational works. The study addresses limitations in current digital humanities methods, which predominantly rely on lexical text reuse detection that captures verbatim quotations but often misses paraphrases and implicit engagements. By employing expert annotations grounded in a semantic taxonomy, the researchers tested whether off-the-shelf semantic search pipelines could identify meaning-level correspondences overlooked by traditional lexical baselines. The results indicate that semantic search successfully retrieves significantly more implicit receptions of Locke's ideas. However, the analysis also reveals a "lexical gatekeeping" effect, where retrieval performance remains partially constrained by surface vocabulary overlap. These findings underscore both the potential and the existing limitations of using semantic retrieval systems to trace the circulation of ideas within large historical corpora. The associated data has been made publicly available to support further research in computational history and natural language processing applications for archival texts.
Wire timeline
Evaluating Semantic Search for 18th-Century Intellectual History via Locke
A new academic paper published on arXiv evaluates the efficacy of semantic search technologies in analyzing 18th-century intellectual history, specifically focusing on the reception of John Locke’s foundational works. The study addresses limitations in current digital humanities methods, which predominantly rely on lexical text reuse detection that captures verbatim quotations but often misses paraphrases and implicit engagements. By employing expert annotations grounded in a semantic taxonomy, the researchers tested whether off-the-shelf semantic search pipelines could identify meaning-level correspondences overlooked by traditional lexical baselines. The results indicate that semantic search successfully retrieves significantly more implicit receptions of Locke's ideas. However, the analysis also reveals a "lexical gatekeeping" effect, where retrieval performance remains partially constrained by surface vocabulary overlap. These findings underscore both the potential and the existing limitations of using semantic retrieval systems to trace the circulation of ideas within large historical corpora. The associated data has been made publicly available to support further research in computational history and natural language processing applications for archival texts.
cs.AI updates on arXiv.org