Could Linguistic Complexity Be Automatically Evaluated? A Multilingual Study on WHO's Emergency Learning Platform

Area of Science:

• Linguistics
• Computational Linguistics
• Knowledge Management

Background:

• Effective knowledge reproduction is crucial, especially for tacit knowledge, which is difficult and costly to transfer, particularly during global health crises.
• Assessing linguistic complexity can enhance the accessibility and dissemination of information across different languages and expert groups.
• Existing models for linguistic complexity may lack cross-linguistic applicability or ease of implementation for non-specialists.

Purpose of the Study:

• To develop a simple, cross-linguistic model for assessing linguistic complexity in various content types.
• To enable non-linguists, such as communication experts and policymakers, to easily implement complexity assessments.
• To evaluate the model's effectiveness using a multilingual corpus from the World Health Organization (WHO) and control datasets.

Main Methods:

• Development of a novel, cross-linguistic model for linguistic complexity analysis.
• Application of the model to a corpus of 6 languages from the WHO's emergency learning platform.
• Utilizing open-access encyclopaedic entries as control data for comparative analysis.
• Analysis of complexity trends and potential features for automatic text classification.

Main Results:

• The developed model successfully assessed linguistic complexity across multiple languages.
• A discernible trend towards the minimization of linguistic complexity was observed in the analyzed WHO corpus.
• The complexity measurements derived from the model show potential as features for automatic text classification tasks.
• The model demonstrated ease of implementation for non-linguist groups.

Conclusions:

• The proposed model offers a practical approach to quantifying linguistic complexity in a cross-linguistic context.
• Minimizing linguistic complexity can improve knowledge reproduction and accessibility, particularly in critical information dissemination scenarios.
• The model's features are valuable for developing automated systems for text analysis and classification, enhancing information management.