Method for performing information-preserving DTD schema embeddings

1. A method for performing information-preserving DTD (Document Type Definition) schema embeddings between a source schema and a target schema, comprising:

finding each DTD-legal string of the target, schema, wherein each of the source schema and the target schema comprise a Directed Acyelie Graph (DAG) DTD, wherein a DTD-legal string is a string, where after deleting all marked string nodes in the string at levels>_i, a node at level i-1 satisfies at least conditions of:

(a) determining that a production of the node does not contain both conjunctive and disjunctive edge,

(b) determining that where the node was originally a conjunctive node before the deletions, the deletions do not cause two or more disjunctive nodes to be merged into the node, and

(c) determining that where the node was originally a disjunctive node before the deletions, the deletions do not cause two or more conjunctive nodes to be merged into the node;

determining, for each DTD-legal string, a bottom-up estimate of DAG substructure root nodes in reduced DTD graphs to provide thereby initial estimates of the best cumulative edit distances of the relevant DAG substructures;

determining a top down adjustment of the initial estimates to provide thereby a one to one matching value and corresponding match costs; and

estimating a minimal match cost in memory, wherein the minimal match cost is the smallest among the match costs.

2. The method of claim 1 wherein the step of finding further comprises:

determining from a sequence of nodes comprising the target schema, a maximal sequence of tree nodes starting from a first node whose parent is either a root or a node having more than one children and ending at a node having more than one children; wherein each intermediate node therebetween has one child, thereby the determined maximal sequence of nodes forms a string.

3. The method of claim 1 where the top down adjustment comprises preparing a best cumulative edit distance of the first string of the target schema and a first string of the source schema routed at respective same level nodes.

4. The method of claim 3 wherein the preparing of the best cumulative edit distance further comprises:

(a) checking that input nodes from a reduced source schema graph and a reduced target schema graph are at the same level;

(b) producing an estimate of the cost of matching sets of strings corresponding to the input nodes; and

(c) comparing out-degree values for the input nodes.

5. The method of claim 4 wherein the cost estimate is computed by finding a first optimal complete bipartite matching between the sets of strings.

6. The method of claim 5 wherein if the out-degree values for the input nodes equal zero, then the edit distance equals the first optimal complete bipartite matching between the sets of strings.

7. The method of claim 5 wherein if the out-degree values for the input nodes do not equal zero, then the edit distance is computed by performing a second complete bipartite matching is solved.

8. The method of claim 7 wherein the second complete bipartite matching determines the best way of matching the input nodes' children using recursively computed cost estimates.

9. The method of claim 1 wherein the determining a top down adjustment step further comprises taking the minimal match cost of the determining step and evaluating parent nodes of the same level input nodes.

10. The method of claim 9 wherein if the parent node evaluation results in a finite cost, the same level input nodes enter a matching and the estimated minimal match cost is updated.