Analogy based workflow identification
View Patent ↗A method of creating an analogous workflow is provided. A first workflow is received at a first device, the first workflow including a plurality of first modules that are connected. A second workflow is received at the first device, the second workflow including a plurality of second modules that are connected. A third workflow is received at the first device, the third workflow including a plurality of third modules that are connected. An analogy workflow is determined based on a difference between the received first workflow and the received second workflow. The determined analogy workflow is applied to the received third workflow to define a fourth workflow. The defined fourth workflow is presented to a user at the first device. A method of identifying a workflow of a plurality of workflows is provided. A query workflow is received at a first device, which includes a plurality of modules that are connected. A workflow is identified of a plurality of workflows, the identified workflow at least partially matching the received query workflow. The identified workflow is presented to a user at the first device.
1. A method of identifying a workflow of a plurality of workflows which contain a plurality of connected modules, the method comprising:
receiving a query workflow by a first computing device, wherein the query workflow comprises a plurality of modules, and further wherein the plurality of modules are connected;
identifying a workflow of a plurality of workflows by the first computing device, the identified workflow at least partially matching the received query workflow, wherein identifying the workflow comprises determining a similarity score between the query workflow and a first workflow of the plurality of workflows, wherein determining the similarity score comprises:
(a) converting the query workflow to a first graph representation, wherein the first graph representation comprises a first module represented as a vertex and a first connection represented as an edge;
(b) converting the first workflow to a second graph representation, wherein the second graph representation comprises a second module represented as a vertex and a second connection represented as an edge; and
(c) creating a graph representation that combines the first graph representation and the second graph representation, wherein the created graph representation is a graph categorical product of the first graph representation and the second graph representation, wherein creating the graph representation comprises
(d) defining a vertex for each pair of vertices of the first graph representation and the second graph representation; and
(e) defining an edge v i,j ˜v k,l when v i ˜v k in the first graph representation and v j ˜v l in the second graph representation exists;
wherein determining the similarity score further comprises;
(f) determining a normalized vector, wherein the normalized vector includes a score for paired vertices of the created graph representation, wherein the score for the paired vertices is determined as
ports
(
v
a
)
⋂
ports
(
v
b
)
ports
(
v
a
)
+
ports
(
v
b
)
where ports(v a ) denotes ports of the first module and ports(v b ) denotes ports of the second module; and presenting the identified workflow to a user by the first computing device.
2. The method of claim 1 , wherein the plurality of workflows are stored as a single data structure in a first memory.
3. The method of claim 2 , wherein the single data structure includes a first plurality of actions taken by a first workflow creator in creating the identified workflow.
4. The method of claim 3 , wherein the received query workflow includes a second plurality of actions taken by a second workflow creator in creating the query workflow.
5. The method of claim 1 , wherein determining the similarity score further comprises:
(g) determining an adjacency matrix, wherein the adjacency matrix is normalized so that a sum of each row is one.
6. The method of claim 5 , wherein determining the similarity score further comprises:
(h) calculating a similarity vector as π k+1 =αA(G)π k +(1−α)c(G), wherein A(G) is the adjacency matrix, c(G) is the normalized vector, α is a parameter that determines a tradeoff between vertex similarity and connectivity similarity, and π k is a first similarity vector.
7. The method of claim 6 , wherein determining the similarity score further comprises:
(i) iteratively calculating the similarity vector until a convergence value is satisfied.
8. The method of claim 7 , wherein at least partially matching the received query workflow is determined based on a module of the second workflow associated with a maximum value of the calculated similarity vector.
9. The method of claim 8 , further comprising repeating (a)-(i) with each workflow of the plurality of workflows as the first workflow.