the method comprising:

parsing, by at least one computing device, a set of question-answer groups to identify a plurality of questions and a plurality of answers, wherein a respective question-answer group comprises a question and at least one answer;

classifying, by the at least one computing device, the set of question-answer groups according to a plurality of dialog acts, wherein a respective dialog act is stored in a data structure in association with a respective one of the question-answer groups;

transforming, by the at least one computing device, the set of question-answer groups into declarative sentences, wherein a respective one of the declarative sentences is stored in a data structure in association with the respective one of the question-answer groups;

performing, by the at least one computing device, sentence correction to modify at least one of the declarative sentences, wherein the sentence correction comprises inputting the at least one of the declarative sentences into a Neural Machine Translation (NMT) based process that translates the at least one of the declarative sentences into at least one language and back-translates the respective declarative sentence back into an original language to remove noise;

classifying, by the at least one computing device, the set of question-answer groups according to a predetermined set of aspects, wherein a respective aspect is stored in the data structure in association with the respective one of the question-answer groups;

identifying, by the at least one computing device, segment boundaries for the set of question-answer groups based at least in part on the declarative sentences;

identifying, by the at least one computing device, candidate summary sentences from the declarative sentences; and

generating, by the at least one computing device, a summary based on a file comprising at least one of the candidate summary sentences arranged according to an aspect layout.