Finding Reasons for Vaccination Hesitancy: Evaluating Semi-Automatic Coding of Internet Discussion Forums

Area of Science:

• Computational Linguistics
• Health Informatics
• Social Sciences

Background:

• Analyzing large text datasets is challenging.
• Manual text coding is time-consuming and labor-intensive.
• Computer-assisted methods can enhance efficiency.

Purpose of the Study:

• To evaluate a logistic regression classifier for ranking text codes.
• To assess the effectiveness of ranked code suggestions for analysts.
• To improve the analysis of vaccination stance reasons in discussion posts.

Main Methods:

• Utilized a dataset of manually coded discussion posts on vaccination stances.
• Trained a logistic regression classifier to predict the presence of specific reasons.
• Ranked reasons based on their probability of occurrence.
• Evaluated performance by checking if expected reasons were in the top 'n' ranked suggestions (n=1-6).

Main Results:

• Logistic regression-based ranking significantly outperformed a baseline frequency-based ranking.
• The proposed method effectively included expected reasons within the top-ranked suggestions.
• This approach demonstrated superior performance compared to traditional methods.

Conclusions:

• Ranked lists of text code suggestions generated by logistic regression are effective.
• This method can be a valuable feature for computer-assisted text coding tools.
• The approach enhances the efficiency and accuracy of analyzing large text collections.