Method for automatic wrapper repair
A method for repairing a wrapper associated with an information source, includes defining a classifier, based on content features of extracted and labeled information using the wrapper, using the classifier to extract content information from the file according to a set of classifier extraction rules; analyzing the extracted content information according to the content features and assigning a label to any extracted content information which satisfies the label's rules; and defining a repaired wrapper as the classifier and those labels in the set which have been assigned to extracted content information. Additional content information and labels can be extracted by iteratively creating a classifier based on both content features and structure features of extracted strings.
1. A method for repairing a wrapper associated with an information source, comprising:
wherein an initial wrapper comprises an initial set of rules for extracting information and for assigning labels from a wrapper set of labels to the extracted information;
defining a classifier comprising a set of alternative rules for extracting information and for assigning labels to the extracted information;
providing a file from the information source after the information source has changed its file format;
using the classifier to extract information from the file and to assign a label to any extracted information which satisfies the label's definition; and
constructing a repaired wrapper comprising those alternative rules for which corresponding labels have been assigned to information extracted by the classifier.
2. The method of claim 1 , further comprising:
using the initial wrapper to extract information from the file and to assign a label from the set of wrapper labels to any extracted information which satisfies the label's definition; and
constructing a repaired wrapper comprising those initial rules for which corresponding labels have been assigned to information extracted by the wrapper and those alternative rules for which corresponding labels which have been assigned to information extracted by the classifier.
3. The method of claim 2 , wherein the initial set of rules comprises grammar/transducer-based rules.
4. The method of claim 1 , wherein the set of alternative rules comprises content features of information extracted by the initial wrapper from files produced by the information source before it changed its file format.
5. The method of claim 1 , wherein the set of alternative rules comprises context features of information extracted by the initial wrapper from files produced by the information source before it changed its file format.
6. The method of claim 1 , wherein the set of alternative rules comprises structure features of information extracted by the initial wrapper from files produced by the information source before it changed its file format.
7. The method of claim 1 , wherein the set of alternative rules comprises the initial set of rules applied in a direction opposite to the direction in which the initial wrapper applied them.
8. The method of claim 4 , wherein content features comprise syntactic features and semantic features and wherein syntactic features comprise length of a string, number of separators between strings, and number of words in a string.
9. The method of claim 8 , wherein semantic features comprise number of nouns, data strings, and numeric strings.
10. The method of claim 1 , further comprising:
defining a second classifier comprising a second set of alternative rules for extracting information and for assigning labels to the extracted information;
using the second classifier to extract information from the file and to assign a label to any extracted information which satisfies the label's definition; and
constructing a repaired wrapper comprising those alternative rules from the set of alternative rules for which corresponding labels have been assigned to information extracted by the classifier and those alternative rules from the second set of alternative rules for which corresponding labels have been assigned to information extracted by the second classifier.
11. The method of claim 10 , wherein the set of alternative rules comprises content features of information extracted by the initial wrapper from files produced by the information source before it changed its file format; and
wherein the second set of alternative rules comprises structure features of information extracted by the initial wrapper from files produced by the information source before it changed its file format.