Effects on Text Simplification: Evaluation of Splitting Up Noun Phrases

Area of Science:

• Natural Language Processing
• Health Communication
• Human-Computer Interaction

Background:

• Health literacy is crucial for patient understanding and adherence.
• Patient education materials often contain complex language, hindering comprehension.
• Text simplification tools aim to improve accessibility of health information.

Purpose of the Study:

• To investigate the impact of splitting noun phrases on text difficulty.
• To evaluate the effect of pseudowords and noun phrase length on perceived and actual text difficulty.
• To inform the development of a text simplification tool for patient education materials.

Main Methods:

• Conducted a user study on Amazon's Mechanical Turk with 12,600 evaluations across 30 sentences per condition.
• Analyzed sentences varying in noun phrase length (2-4 grams), noun phrase splitting (split/not split), and pseudoword presence (present/not).
• Utilized a 3-way analysis of variance to assess perceived and actual difficulty.

Main Results:

• Splitting noun phrases positively impacted perceived difficulty but negatively affected actual difficulty.
• Pseudowords significantly increased both perceived and actual difficulty.
• Longer noun phrases increased perceived and actual difficulty when pseudowords were absent.

Conclusions:

• Splitting noun phrases enhances perceived ease but can reduce actual readability by creating less natural phrasing.
• Pseudowords and longer noun phrases are key indicators of text complexity.
• Findings guide the development of text simplification tools for more effective patient education.