OBJECTIVES:

Screen media have become an integral part of life in the United States. Screen time guidelines have been issued to restrict children’s screen media use. This study explored the sentiment related to screen media use and screen media parenting on Twitter.

METHODS:

A random sample of tweets (n=1,148,016) representing 748,104 individuals in the United States was retrieved from 2019 using Twitter’s application programming interface. Natural language processing and sentiment analysis were conducted to process the unstructured text and classify the tweets into positive, negative, and neutral sentiments. SentiStrength was used for this analysis and most frequent sentimental words were presented using wordcloud. Manual annotation of 300 parenting-related sample tweets was conducted to validate the performance of SentiStrength.

RESULTS:

Retweets made up 54.3% of the total tweets. After excluding marketing tweets and irrelevant topics, 1,074,626 tweets were included in the study for sentiment analysis. Overall, 41.6% of tweets were classified as neutral 30.4% as positive, and 28.0% as negative. The distribution of positive words was highly skewed whereas negative words were more evenly distributed. The positive sentiments were mostly focused on connecting with friends, family, and relaxing. However, the negative sentiments involved a lot more negative words concerning violence, illness, drinking, and tiredness. Similar findings were observed for parenting-related tweets. When compared to manual annotation, SentiStrength’s performance varied among different sentiments, with high sensitivity for positive tweets (82.9%) but low sensitivity for negative tweets (57.3%).

CONCLUSIONS:

The number of positive tweets was slightly higher than negative tweets related to screen media use. However, positive sentiments related to screen media use were limited in scope while the negative sentiments related to screen media use were more wide-ranging. SentiStrength lacks the domain-specific feature and assigned neutral words positive scores. Human intelligence is needed to evaluate the performance and interpret the results.