Developing a standardized protocol for computational sentiment analysis research using health-related social media...

Area of Science:

• Computational linguistics
• Public health informatics
• Social media analytics

Background:

• Sentiment analysis is widely used for health-related social media content.
• Existing studies suffer from methodological issues and inconsistent reporting.
• These inconsistencies can lead to biased data and unreliable conclusions.

Purpose of the Study:

• To systematically analyze methodological and reporting inconsistencies in health sentiment analysis studies.
• To develop a standardized protocol for improving research validity and comparability.
• To address the need for a consistent approach in analyzing health-related social media data.

Main Methods:

• Conducted a systematic review of eligible studies published between 2010-2019.
• Analyzed common research design choices and their reporting.
• Identified 16 distinctive research design choices, noting inconsistencies in 9 key areas.

Main Results:

• Identified significant methodological and reporting inconsistencies in health sentiment analysis research.
• Found variations in data relevance determination and sentiment analysis tool validation.
• Developed the Protocol of Analysis of senTiment in Health (PATH) based on these findings.

Conclusions:

• Substantial inconsistencies exist in the current literature on health sentiment analysis.
• The developed PATH protocol offers a standardized approach to mitigate these issues.
• Implementing PATH can enhance the validity and comparability of future health-related social media studies.