The major goal of our study is to test methodsin NLP in the domain of health care educationrelated to Covid-19 of vulnerable groups suchas indigenous people from Latin America. Inorder to achieve this goal, we asked participantsin a survey questionnaire to provide answersabout health related topics. We used these answersto measure the health education status ofour participants. In this paper, we summarizethe results from our NLP-application on theparticipants’ answers. In the first experiment,we use embeddings-based tools to measure thesemantic similarity between participants’ answersand “expert” or “reference” answers. Inthe second experiment, we use synonym-basedmethods to classify answers under topics. Wecompare the results from both experiments withhuman annotations. Our results show that thetested NLP-methods reach a significantly loweraccuracy score than human annotations in bothexperiments. We explain this difference by theassumption that human annotators are muchbetter in pragmatic inferencing necessary toclassify the semantic similarity and topic classificationof answers.