Scholarlink

<p>Juries play a critical role in determining trial verdicts. One major challenge jurors face is interpreting legal texts that include polysemous words: terms that have multiple meaning senses. Not all polysemous terms are equally problematic. Corpus-based methods offer a principled means of identifying which terms genuinely risk misinterpretation, and which carry a stable enough meaning to remain accessible to lay people. Building on recent advances in legal corpus linguistics [<CitationRef CitationID="CR1">1</CitationRef>], the present study applies corpus-based sense-tagging to jury instructions. The noun <i>fault</i> was selected as the unit of analysis due to its polysemous nature across legal and general registers. All word forms and inflections of <i>fault</i> were identified, and their frequencies were recorded in both the Model Utah Jury Instructions (MUJI) [<CitationRef CitationID="CR2">2</CitationRef>] corpus and the Corpus of Contemporary American English (COCA). Psycholinguistic recognition data (reaction times from a lexical decision task) were also consulted to assess the accessibility of <i>fault</i> to lay people. The study triangulates frequency data, dictionary meaning senses, and manually tagged concordance lines to identify dominant meaning senses and determine where overlap or divergence occurs. (Inter-rater agreement was assessed to ensure consistency across two human raters.)Unlike extensionalist/reference-based approaches that rely on co-occurrence frequencies alone, this method places human raters at the center of meaning classification, providing a transparent and replicable ground truth for legal word meaning. The analysis shows that the noun <i>fault</i> maintains a consistent meaning across legal and non-legal registers. Overall, this study contributes to the growing field of legal corpus linguistics by demonstrating how corpus-based sense-tagging can help identify whether a given legal term poses a genuine interpretive risk, and by showing that in the case of <i>fault</i>, its dominant meaning is sufficiently stable to be accessible to lay people. The human-validated methodology proposed here also offers a potential ground truth benchmark for evaluating computational models of legal meaning [<CitationRef CitationID="CR3">3</CitationRef>]. It should be noted that direct juror testing remains a necessary next step.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Sense-Tagging Legal Language: Corpus-Based Disambiguation in Jury Instructions

  • 期刊论文

  • 发表日期：

    2026年7月31日

查看全文
  • Mary Akbary

摘要

Juries play a critical role in determining trial verdicts. One major challenge jurors face is interpreting legal texts that include polysemous words: terms that have multiple meaning senses. Not all polysemous terms are equally problematic. Corpus-based methods offer a principled means of identifying which terms genuinely risk misinterpretation, and which carry a stable enough meaning to remain accessible to lay people. Building on recent advances in legal corpus linguistics [1], the present study applies corpus-based sense-tagging to jury instructions. The noun fault was selected as the unit of analysis due to its polysemous nature across legal and general registers. All word forms and inflections of fault were identified, and their frequencies were recorded in both the Model Utah Jury Instructions (MUJI) [2] corpus and the Corpus of Contemporary American English (COCA). Psycholinguistic recognition data (reaction times from a lexical decision task) were also consulted to assess the accessibility of fault to lay people. The study triangulates frequency data, dictionary meaning senses, and manually tagged concordance lines to identify dominant meaning senses and determine where overlap or divergence occurs. (Inter-rater agreement was assessed to ensure consistency across two human raters.)Unlike extensionalist/reference-based approaches that rely on co-occurrence frequencies alone, this method places human raters at the center of meaning classification, providing a transparent and replicable ground truth for legal word meaning. The analysis shows that the noun fault maintains a consistent meaning across legal and non-legal registers. Overall, this study contributes to the growing field of legal corpus linguistics by demonstrating how corpus-based sense-tagging can help identify whether a given legal term poses a genuine interpretive risk, and by showing that in the case of fault, its dominant meaning is sufficiently stable to be accessible to lay people. The human-validated methodology proposed here also offers a potential ground truth benchmark for evaluating computational models of legal meaning [3]. It should be noted that direct juror testing remains a necessary next step.