Text-mining of children's answers to national web-survey questions on parental status during COVID-19

<p>Objectives　Behavior at school has been restricted due to infection control during the COVID-19 pandemic. This study aimed to clarify children's opinions by examining the structure of their text responses to open-ended web survey questions, as well as how they differed to the psychological and socio-economic status of their parents.</p><p>Method　From September to October 2020, approximately six months after the temporary closure of schools in Japan due to COVID-19, 2,111 respondents (elementary, junior high, and high school students) answered a questionnaire on “The third wave of CORONA-CODOMO project” website. The text data of 1,140 students who provided free-text answers to open-ended questions such as, “what are you concerned about and what do you want to say?” (comments) and “how should I get them to understand?” (proposals) were analyzed. Results were stratified by parental demographics such as age, employment status, the Kessler Psychological Distress Scale (K6), and financial status. The text mining assessed the frequency of word appearances, analyzed characteristic words, and showed a term network (word-relationship diagram).</p><p>Results　The total number of lines (number of respondents) of comments and proposal texts were 531 and 1,017, respectively, and the average line length (number of characters) was 21.5 and 31.5, respectively. In the dependency relations frequency analysis, the actions of proposals and comments such as “speaking-listening”, “event-disappearing”, and “mask-removing” appeared. In the word network, words such as “moving” and “speaking” formed a strong co-occurrence network to “I”, which had the largest node in proposals, while “COVID-19” in the comments strongly co-occurred with “end+?” (predicate attribute: question) and “I+want to disappear” (predicate attribute: aspirations). According to the parent attributes, in the proposals, the characteristic words (complementary similarity measure) were “talk” (35.9) in the employed group, “talk” (26.6) in the K6 lower score group, and “understand+not” (23.5/17.3, predicate attribute: negation) in the K6 higher score group/bad economic status. In the comments, it was “COVID-19” (28.1, 27.5) in the employed group and the high mental health group.</p><p>Conclusions　While children displayed discomfort and fear of COVID-19 when asked about proposals, the children were willing to talk directly, also, they wanted someone to listen to them. The characteristics of the proposals in their answers were “I don't know” in both the high mental burden group and the bad economic status group of the parents.</p>