Comments on the net1.input file. The net1.input file is a sample of the input that you will be using to create the underlying network of the formula-augmented network (FAN) that you are constructing for the term project. net1.input may change during the next few days. The general format of net1.input is: ( ("nodename" SuperTypes <["node1" [number] ..... "noden" [number] "]> [Cancels <"nodek" ..... "nodek+m">] ) (where objects inside square brackets are optional) That is, the file consists of a set of node descriptions. It is delimited by a open parenthesis ( and a closed parenthesis ). Each node description is itself delimited by an open parenthesis and a closed parenthesis. The node description consists of a name, delimited by quote marks, a SuperTypes field, giving the positive parents of the node, if any, that is, the set of nodes to which there are isa links from the node, and a possibly empty Cancels field (giving the nodes to which there are cancels links from the node). The SubTypes field is never empty: if a node is a root, the value of its SubTypes field is the empty string . If there are multiple supertypes, numbers may appear after each node. In general, numbers will not appear if the multiple supertypes can themselves be organized in a hierarchy. In general, if a node has two supertypes a and b, and a is followed by a lower number than b, then a is considered to be preferred to b, in the FANs algorithm. If a and b are followed by equal numbers, neither is preferred. The numbers provided after multiple supertypes give the partial order of the supertypes. Remember that the partial order is *never* used in Stein's algorithm, and is used only in the FANs inheritance algorithm. The value of the SuperTypes and Cancels fields is delimited by the open angle bracket < and the close angle bracket >. The data given is a small portion of the network used in the benefits inquiry system described in class. The data had to be modified considerably in order to allow distribution. It can easily be seen that the structure is very simple, much simpler than the benchmark examples on which inheritance algorithms are typically tested. For that reason, I have included at the end of the file the Royal Elephant network (Touretzky: A Mathematical Theory of Inheritance, 1984) and the Seedless Grapevine network (Stein: AIJ, 1992).