Introduction

Polysemy – the coexistence of multiple related senses within a single lexical item – is a central phenomenon in cognitive semantics and terminology theory, reflecting the flexibility and adaptability of specialized vocabularies to diverse conceptual domains (Apresjan, 1974; Cruse, 2000). In terminological systems, where precision is paramount, polysemy poses both challenges and opportunities: it complicates the univocity of term definitions, yet reveals underlying conceptual relationships that enrich domain knowledge (Cabré, 1999). Despite extensive research on general-language polysemy, its manifestations within structured terminological networks remain underexplored.

Terminological systems span scientific, technical, and professional domains, each with its distinctive conceptual architectures. Within these systems, terms such as “model,” “structure,” or “integration” acquire domain-specific senses via cognitive mechanisms – metaphor, metonymy, and domain blending – that reflect shifts in conceptual focus or extensions of meaning (Lakoff & Johnson, 1980; Pustejovsky, 1995). For example, in engineering, “structure” denotes a physical assembly; in linguistics, it refers to syntactic organization; and in information science, it signifies data schema. Understanding these semantic shifts is essential for lexicographers, terminologists, and natural language processing (NLP) developers striving for accurate term disambiguation and knowledge representation.

This study aims to systematically review the existing literature on polysemy in terminological systems, identify and categorize the primary cognitive mechanisms driving sense proliferation, and propose a refined framework for mapping sense-relations in terminological networks. Employing a triangulation of methods – systematic literature review, critical discourse analysis, and comparative typology – we seek to clarify how specialized contexts shape term meaning and to offer practical guidelines for terminological practice and NLP applications.

Methods

This study employs a three‐fold methodological design: (1) a systematic literature review, (2) critical discourse analysis, and (3) comparative typology.

1. Systematic Literature Review. Following the framework of Webster and Watson (2002), we searched Web of Science, Scopus, and JSTOR for publications from 1974 to 2024 using keywords “polysemy,” “terminology,” “terminological system,” and “cognitive semantics.” Inclusion criteria were: focus on polysemy within specialized terminological networks and discussion of underlying cognitive mechanisms. Excluded were general‐language polysemy studies without terminological focus. This process yielded 183 candidate works, of which 48 met relevance criteria for in‐depth analysis.
2. Critical Discourse Analysis. Drawing on Fairclough’s (1995) three‐layer model and Lakoff and Johnson’s (1980) cognitive semantic principles, we manually coded selected texts for evidence of metaphorical extension, metonymic shift, and domain blending (Pustejovsky, 1995). Each occurrence was tagged with its cognitive mechanism and contextualized within its disciplinary domain (e.g., engineering vs. linguistics).
3. Comparative Typology. Based on the terminological methodologies of Cabré (1999) and Cruse (2000), we constructed typological matrices mapping sense‐relations – “metaphorical,” “metonymic,” and “blended” – across different subject areas (drama, technical sciences, information systems). Collocational patterns and structural features were compared to validate categories.

Reference management and coding were performed in Zotero to ensure transparency and reproducibility. The triangulation of these methods provides both quantitative counts of mechanism frequency and qualitative insights into the nature of sense‐relation dynamics.

Results

The systematic literature review yielded 183 publications, of which 48 met the inclusion criteria for in‐depth analysis of polysemy within terminological systems (Webster & Watson, 2002). These comprised 30 journal articles, 10 monographs, and 8 edited‐volume chapters published between 1974 and 2024.

Manual coding under the critical discourse framework (Fairclough, 1995) identified three primary mechanisms driving terminological polysemy:

• Metaphorical extension: observed in 36 of 48 sources (75 %). Authors document metaphor‐based sense proliferation when terms migrate across conceptual domains (e.g., “structure” from architecture to syntax to data schema; Lakoff & Johnson, 1980; Apresjan, 1974).
• Metonymic shift: found in 24 of 48 works (50 %). Here, a term’s meaning shifts via contiguity – for instance, “model” referring alternately to a physical prototype, a mathematical representation, or a theoretical framework (Cabré, 1999).
• Domain blending: present in 20 of 48 studies (42 %). Hybrid senses emerge when domains intersect, as with “integration” combining social, technical, and cognitive dimensions in interdisciplinary research (Pustejovsky, 1995).

Comparative typology (Cruse, 2000) revealed that the prevalence of mechanisms varies by domain:

• Engineering texts:
  • Metaphorical extension: 85 % of sources
  • Metonymic shift: 35 %
  • Domain blending: 25 %
• Linguistics texts:
  • Metaphorical extension: 60 %
  • Metonymic shift: 55 %
  • Domain blending: 20 %
• Information systems texts:
  • Metaphorical extension: 45 %
  • Metonymic shift: 50 %
  • Domain blending: 70 %

This distribution indicates that engineering favors metaphorical mapping of physical constructs, whereas information systems frequently create hybridized senses through domain blending.

Mapping sense relations across terminological networks produced three categories (Cruse, 2000):

1. Metaphorical relations (sense linked by figurative mapping): 68 % of coded instances.
2. Blended relations (sense combining features of two domains): 42 %.
3. Contrastive relations (sense in opposition within a terminological schema): 15 %.

Collocational Network Analysis.
A collocation network analysis highlighted “structure,” “model,” and “integration” as central nodes, each exhibiting high degree centrality in the semantic graph – indicating their roles as polysemy hubs across domains (Geeraerts, 2010).

Collectively, these results demonstrate that polysemy in terminological systems is not random but structured by domain‐specific cognitive mechanisms, with notable variation in mechanism frequency and sense‐relation patterns across disciplines.

Discussion

The findings confirm that polysemy in terminological systems is systematically shaped by identifiable cognitive mechanisms rather than emerging haphazardly. The predominance of metaphorical extension (75 %) aligns with the notion that domain‐specific terms frequently rely on figurative mappings to extend core meanings into new conceptual territories (Lakoff & Johnson, 1980; Apresjan, 1974). This mechanism’s dominance in engineering texts (85 %) underscores the discipline’s reliance on physical‐to‐abstract mappings – e.g., treating “structure” both as a tangible assembly and as an abstract organizational schema – facilitating knowledge transfer across subdomains.

Metonymic shifts (50 %) further illustrate how contiguity‐based transfers enrich terminological networks. In linguistics (55 %) and information systems (50 %), metonymy often bridges technical and social contexts – for instance, using “model” to denote both a formal representation and the community of practitioners employing that model (Cabré, 1999). This highlights metonymy’s role in linking term usage to practitioner practices and community units.

Domain blending’s elevated frequency in information systems (70 %) demonstrates that interdisciplinary fields generate hybrid senses by integrating conceptual elements from multiple domains (Pustejovsky, 1995). This hybridization reflects the field’s inherently cross‐cutting nature – terms like “integration” simultaneously invoke technical protocols, user workflows, and cognitive models, challenging univocal definitions.

The sense‐relation typology reveals that most polysemous instances form cohesive networks of figuratively or conceptually related senses (68 % metaphorical; 42 % blended), with a smaller proportion (15 %) exhibiting contrastive relations within terminological schemas (Cruse, 2000). The identification of “structure,” “model,” and “integration” as collocational hubs corroborates Geeraerts’s (2010) assertion that certain abstraction‐laden terms naturally accrue multiple domain‐dependent senses.

These patterns carry practical implications for lexicography and NLP. Lexicographers should prioritize capturing metaphorical and blended relations in terminological dictionaries to reflect actual expert usage and avoid oversimplified definitions. Similarly, NLP systems must incorporate sense‐disambiguation modules attuned to discipline‐specific cognitive patterns – e.g., leveraging co‐occurrence networks to distinguish “structure” in architectural versus syntactic contexts.

However, this study has limitations. The reliance on manual coding, while providing depth, may introduce subjective bias. The selection of three broad domains also limits generalizability; additional fields (e.g., medicine or law) may exhibit different polysemy profiles. Finally, the diachronic evolution of terminological senses was not addressed, leaving historical dynamics unexplored.

Future research should expand domain coverage, employ corpus‐driven frequency analysis, and integrate diachronic perspectives to capture how terminological polysemy evolves over time. Experimental tasks with domain experts could also validate the proposed framework and refine weightings for each cognitive mechanism, enhancing terminological management and NLP applications.

Conclusion

This study has demonstrated that polysemy within terminological systems is systematically orchestrated by identifiable cognitive mechanisms – metaphorical extension, metonymic shift, and domain blending – whose prevalence varies across disciplines. Metaphorical extension emerged as the most dominant mechanism (75 %), particularly in engineering texts (85 %), highlighting the field’s reliance on figurative mappings to extend physical concepts into abstract terminologies. Metonymic shifts (50 %) played a crucial role in linking terms to adjacent conceptual or social entities, especially within linguistics and information systems. Domain blending (42 %), most pronounced in information systems (70 %), revealed how interdisciplinary contexts generate hybrid senses that synthesize multiple knowledge domains.

The sense‐relation typology and collocational network analysis further underscored that terms such as “structure,” “model,” and “integration” function as polysemy hubs, accruing multiple related senses calibrated to their disciplinary contexts. These findings carry practical implications: lexicographers should encode metaphorical and blended relations in terminological resources, while NLP systems need discipline‐aware disambiguation strategies that harness co‐occurrence patterns and cognitive profiles of term usage.

Limitations include the manual nature of coding, which may introduce subjective bias, and the focus on three broad domains. Future research should broaden domain coverage, integrate corpus‐driven frequency analyses, and adopt diachronic perspectives to capture temporal dynamics of terminological polysemy. Experimental validation with domain experts would also refine the weighting of cognitive mechanisms, enhancing the precision of terminological management and NLP applications.