Use of Machine Learning to Identify Topical Arguments

Technology is described for identifying topical arguments to be provided in order to enable problem solving. The method can include a first operation of storing a topical problem statement at a root of a graph. A plurality of topical arguments may be stored in a plurality of argument nodes and user answers to the topical arguments in the graph. Another operation may be grouping the argument nodes into section groups that define a sub-topic linked by the graph to the topical problem statement. Additionally, use requests for the topical arguments in the graph may be tracked by recording access to the topical arguments. A use pattern of topical arguments may be identified by tracking use requests to the topical arguments. The use pattern may be processed using a machine learning model. Another operation may be sending an additional section group with topical arguments and user answers in the graph based in part on processing of the use pattern using the machine learning model to determine which additional section group to send to a requestor.