System and method for accessing heterogeneous databases

• US 6,295,533 B2
• Filed: 02/24/1998
• Issued: 09/25/2001
• Est. Priority Date: 02/25/1997
• Status: Expired due to Fees

• Current Assignee: Rutgers University
• Original Assignee: AT&T Corporation (AT&T, Inc.)
• Inventors: Cohen, William W.
• Primary Examiner(s): Alam, Hosain T.
• Assistant Examiner(s): FLEURANTIN, JEAN B
• Application Number:
  US09/028,471
• Publication Number:
  US 20010013035A1
• Time in Patent Office: 1,309 Days
• Field of Search: 707/3-7, 704/9-10, 704/240
• US Class Current: 1/1
• CPC Class Codes: G06F 16/256 in federated or virtual databases
• CPC Class Codes: Y10S 707/99935 Query augmenting and refining, e.g. inexact access

Abstract

A system and method are provided for answering queries concerning information stored in a set of collections. Each collection includes a structured entity, and each structured entity includes a field. A query is received that specifies a subset of the set of collections and a logical constraint between fields that includes a requirement that a first field match a second field. The probability that the first field matches the second field is determined automatically based upon the contents of the fields. A collection of lists is generated in response to the query, where each list includes members of the subset of collections specified in the query, and where each list has an estimate of the probability that the members of the list satisfies the logical constraint specified in the query.

Claims

1. A method for answering a query containing a join operation, the method comprising:

representing each entry in a column of a first relation by a vector;

representing each entry in a column of a second relation by a vector;

selecting a subset of rows of the first relation;

for each of the entries in the column of the first relation that is part of a row from the subset of rows;

determining the value of a similarity metric function that is based upon the vector representing the entry in the column of the first relation and a vector representing an entry from the column of the second relation, for each entry in the column of the second relation;

joining the first relation with the second relation based upon the set of similanty metric function values determined; and

outputting the result of the joining.

2. A method for answering a query containing a join operation, the method comprising:

representing each entry in a field of a first relation by a vector;

representing each entry in a field of a second relation by a vector;

selecting a subset of tuples of the first relation;

for each of the entries in the field of the first relation that is part of a tuple from the subset of tuples;

determining the value of a similarity metric function that is based upon the vector representing the entry in the field of the first relation and a vector representing an entry from the field of the second relation, for each entry in the field of the second relation;

joining the first relation with the second relation based upon the set of similarity metric function values determined; and

outputting the result of the joining.

3. The method of claim 2 wherein the joining is based on the N highest values in a subset of the set of similarity metric function values determined, where N is an integer.

4. The method of claim 2 wherein the joining is based on the N lowest values in a subset of the set of similarity metric function values determined, where N is an integer.

5. The method of claim 2 wherein an estimate of probability for correct matching of each tuple of the result is additionally output.

6. The method of claim 5 wherein the estimate of probability for a tuple is based on the value of the similarity metric function associated with that tuple.

7. The method of claim 2 wherein the first relation and the second relation are heterogeneous relations that are maintained separately and in locations remote to one another.

8. The method of claim 2 wherein each component of each vector corresponds to at least one of the following terms in the corresponding entry:

a word, a word prefix, a word suffix, and a phrase.

9. The method of claim 8 wherein the value of each component of each vector is based on the frequency of the term corresponding to the component in the entry corresponding to the vector.

10. The method of claim 9 wherein the value of a component of each vector is additionally based on the frequency of occurrence of the term corresponding to the component, in all of the entries in a field including the vector.

11. The method of claim 2 wherein each vector is normalized to unit length.

12. The method of claim 2 wherein the similarity metric function is given by $\sum _{}$

i

vi·

wi

wherein vⁱ is the i^th component of a vector v corresponding to an entry in the field from the first relation, and wⁱ is the i^th component of vector w corresponding to an entry in the field from the second relation.

13. The method of claim 2 wherein the query is probabilistically answered using the A* search algorithm.

14. A medium storing instructions for answering a query containing a join operation, the instructions adapted to be executed by a processor, the instructions including:

representing each entry in a field of a first relation by a vector;

representing each entry in a field of a second relation by a vector;

selecting a subset of tuples of the first relation;

for each of the entries in the field of the first relation that is part of a tuple from the subset of tuples;

determining the value of a similarity metric function that is based upon the vector representing the entry in the field of the first relation and a vector representing an entry from the field of the second relation, for each entry in the field of the second relation;

joining the first relation with the second relation based upon the set of similarity metric function values determined; and

outputting the result of the joining.

16. The medium of claim 14 wherein the joining is based on the N lowest values in a subset of the set of similarity metric function values determined, where N is an integer.

17. The medium of claim 14 wherein an estimate of probability for correct matching of each tuple of the result is additionally output.

18. The medium of claim 17 wherein the estimate of probability for a tuple is based on the value of the similarity metric function associated with that tuple.

19. The medium of claim 14 wherein the first relation and the second relation are heterogeneous relations that are maintained separately and in locations remote to one another.

20. The medium of claim 14 wherein each component of each vector corresponds to at least one of the following terms in the corresponding entry:

a word, a word prefix, a word suffix, and a phrase.

21. The medium of claim 14 wherein the value of each component of each vector is based on the frequency of the term corresponding to the component in the entry corresponding to the vector.

22. The medium of claim 21 wherein the value of a component of each vector is additionally based on the frequency of occurrence of the term corresponding to the component, in all of the entries in a field including the vector.

23. The medium of claim 14 wherein each vector is normalized to unit length.

24. The medium of claim 14 wherein the similarity metric function is given by $\sum _{}$

i

vi·

wi

wherein vⁱ is the i^th component of a vector v corresponding to an entry in the field from the first relation, and wⁱ is the i^th component of vector w corresponding to an entry in the field from the second relation.

25. The medium of claim 14 wherein the query is probabilistically answered using the A* search algorithm.

15. The medium 14 wherein the joining is based on the N highest values in a subset of the set of similarity metric function values determined, where N is an integer.

26. An apparatus for answering a query containing a join operation, the apparatus comprising:

a processor; and

a memory that stores instructions adapted to be executed by a processor, the instructions including;

representing each entry in a field of a first relation by a vector;

representing each entry in a field of a second relation by a vector;

selecting a subset of tuples of the first relation;

for each of the entries in the field of the first relation that is part of a tuple from the subset of tuples;

determining the value of a similarity metric function that is based upon the vector representing the entry in the field of the first relation and a vector representing an entry from the field of the second relation, for each entry in the field of the second relation;

joining the first relation with the second relation based upon the set of similarity metric function values determined; and

outputting the result of the joining.

27. A method for performing a join operation, the method comprising the steps of:

representing each entry in a field of a first relation and a field of a second relation as a vector;

evaluating a similarity metric function on a plurality of pairs of vectors, wherein each pair from the plurality of pairs of vectors includes a vector from the field of the first relation and a vector from the field of the second relation;

determining whether to join a tuple from the first relation and a tuple from the second relation based on i) the value of the similarity function evaluated on the vector corresponding to the entry that is in the tuple from the first relation and the field of the first relation, and the vector corresponding to the entry that is in the tuple from the second relation and the field of the second relation; and

ii) the values for the similarity metric function obtained in the evaluation step; and

outputting the tuple obtained by joining the tuple from the first relation and the tuple from the second relation.

28. The method of claim 27 wherein a probability estimate for correct joining of the tuple from the first relation and the tuple from the second relation is also output, wherein the probability estimate is based on the value of the similarity function evaluated on the vector corresponding to the entry that is in the tuple from the first relation and the field of the first relation, and the vector corresponding to the entry that is in the tuple from the second relation and the field of the second relation.