Method and system for finding similar records in mixed free-text and structured data

• US 20060116995A1
• Filed: 01/13/2006
• Published: 06/01/2006
• Est. Priority Date: 03/07/2001
• Status: Active Grant

• Current Assignee: Eric Bloedorn
• Original Assignee: Eric Bloedorn
• Inventors: Bloedorn, Eric
• Granted Patent: US 7,440,946 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G06F 16/2465 Query processing support for facilitating data mining operations in structured databases
• CPC Class Codes: G06F 16/3346 using probabilistic model
• CPC Class Codes: Y10S 707/99931 Database or file accessing
• CPC Class Codes: Y10S 707/99936 Pattern matching access
• CPC Class Codes: Y10S 707/99945 Object-oriented database structure processing

Abstract

A technique for data mining where the available data contains both structured as well as unstructured (free-text) data. The present invention combines together the information available from different types of data to provide a single similarity score indicating the degree of similarity between records. Thus, a data evaluation application selects two records from a database and compares corresponding fields from the two records. The application determines whether to apply a nominal matching process, an ordinal matching process, or a vector-space matching process depending on the type of data in each pair of corresponding fields. The application sums the matching scores for all the fields in the records to compute the similarity score.

Claims

1. A method for determining whether records are similar in a database, the method comprising:

(a) selecting two records in the database;

(b) accessing two corresponding fields of the selected records;

(c) determining a type of data in the accessed fields; and

(d) applying a match function to the accessed corresponding fields to generate a match score based on the type of data in the accessed fields,

wherein,

if the type of data in the fields is nominal, the match function applied is a Boolean match function, if the type of data in the fields is ordinal, the match function applied is an ordinal match function, and if the type of data in the fields is unstructured data, the match function applied is a vector-based match function.

2. A method as defined in claim 1 further comprising:

(e) repeating steps b through d for one or more additional corresponding fields of the selected records to generate one or more additional match scores; and

(f) generating a similarity score that indicates a degree of similarity between the two records from the match scores.

3. A method as defined in claim 2 wherein the similarity score is generated as follows:

similarity_score=w₁*match(a_1i,a_1j)+w₂*match (a_2i,a_2j)+ . . . w_n*match(a_ni,a_nj)

wherein,

similarity score is the similarity score,

i identifies a first selected record in the database,

j identifies a second selected record in the database,

identifies a field position for a given field an in the first selected record and a corresponding field position for a given field a_nj in the second selected record,

match indicates the match function used to generate the match scores, and

w_n indicates a predefined weight for each match score.

4. A method as defined in claim 1 wherein the match score is generated by the Boolean match function as follows:

match_score=1 if a_ni equals a_nj else function_result=0

wherein,

match_score indicates the match score generated by the Boolean match function,

i identifies a first selected record in the database,

j identifies a second selected record in the database, and

n identifies a field position for a given field a_ni in the first selected record and a given field a_nj in the second selected record.

5. A method as defined in claim 1 wherein the match score is generated by the ordinal match function as follows:

match_score=1−

((a_ni−

a_nj)/|Domain a_n|)

wherein,

match_score indicates the match score generated by the ordinal match function,

i identifies a first selected record in the database,

j identifies a second selected record in the database,

n identifies a field position for a given field a_ni in the first selected record and a given field a_nj in the second selected record, and

|Domain a_n| is the size of the data domain of the corresponding fields.

6. A method as defined in claim 1 wherein the match score is generated by the vector-based match function as follows:

$\mathrm{match\_score}=\underset{x=1<br><br>\text{\hspace{1em}<br><br>}<br><br>\mathrm{to}<br><br>\text{\hspace{1em}<br><br>}<br><br>V}{\sum <br><br>}<br><br>\frac{{\mathrm{weight}}_{\mathrm{nix}}*{\mathrm{weight}}_{\mathrm{njx}}}{\sqrt{\left({\mathrm{weight}}_{\mathrm{nix}}\right)^2*\left({\mathrm{weight}}_{\mathrm{njx}}\right)^2}}$

wherein,

match_score indicates the match score generated by, the vector-based match function,

V is a size of a vocabulary associated with the data in the corresponding fields,

i identifies a first selected record in the database,

j identifies a second selected record in the database,

n identifies an accessed field in the selected records,

x identifies a word within an accessed field,

weight_nix is a weight of word x in field n of the first record, and

weight_njx is a weight of word x in field n of the second record.

7. A data processing system comprising:

a database having a plurality of records;

a processor capable of accessing records in the database, the processor configured to;

(a) select two records in the database;

(b) access two corresponding fields of the selected records;

(c) determine a type of data in the accessed fields; and

(d) apply a match function to the accessed corresponding fields to generate a match score based on the type of data in the accessed fields, wherein,

if the type of data in the fields is nominal, the match function applied is a Boolean match function, if the type of data in the fields is ordinal, the match function applied is an ordinal match function, and if the type of data in the fields is unstructured data, the match function applied is a vector-based match function.

8. A data processing system as defined in claim 7 further wherein the processor is further configured to:

(e) repeat steps b through d for additional corresponding fields of the selected records to generate one or more additional match scores; and

(f) generate a similarity score that indicates a degree of similarity between the two records from the match scores.

9. A data processing system as defined in claim 8 wherein the similarity score is generated as follows:

similarity_score=w₁*match(a_1i,a_1j)+w₂*match(a_2i,a_2j)+ . . . w_n*match(a_ni,a_nj)

wherein,

similarity_score is the similarity score,

i identifies a first selected record in the database,

j identifies a second selected record in the database,

n identifies a field position for a given field a_ni in the first selected record and a corresponding field position for a given field a_nj in the second selected record,

match indicates the match function used to generate the match scores, and

w_n indicates a predefined weight for each match score.

10. A data processing system as defined in claim 7 wherein the match score is generated by the Boolean match function as follows:

match_score=1 if a_ni equals a_nj else function_result=0

wherein,

match_score indicates the match score generated by the Boolean function,

i identifies a first selected record in the database,

j identifies a second selected record in the database, and

n identifies a field position for a given field a_ni in the first selected record and a given field a_nj in the second selected record.

11. A data processing system as defined in claim 7 wherein the match score is generated by the ordinal match function as follows:

match_score=1−

((a_ni−

a_nj)/|Domain a_n|)

wherein,

match_score indicates the match score generated by the ordinal match function,

i identifies a first selected record in the database,

j identifies a second selected record in the database,

n identifies a field position for a given field a_ni in the first selected record and a given field a_nj in the second selected record, and

|Domain a_n| is the size of the data domain of the corresponding fields.

12. A data processing system as defined in claim 7 wherein the match score is generated by the vector-based match function as follows:

$\mathrm{match\_score}=\underset{x=1<br><br>\text{\hspace{1em}<br><br>}<br><br>\mathrm{to}<br><br>\text{\hspace{1em}<br><br>}<br><br>V}{\sum <br><br>}<br><br>\frac{{\mathrm{weight}}_{\mathrm{nix}}*{\mathrm{weight}}_{\mathrm{njx}}}{\sqrt{\left({\mathrm{weight}}_{\mathrm{nix}}\right)^2*\left({\mathrm{weight}}_{\mathrm{njx}}\right)^2}}$

wherein,

match_score indicates the match score generated by the vector-based match function,

V is a size of a vocabulary associated with the data in the corresponding fields,

i identifies a first selected record in the database,

j identifies a second selected record in the database,

n identifies an accessed field in the selected records,

x identifies a word within an accessed field,

weight_nix is a weight of word x in field n of the first record, and

weight_njx is a weight of word x in field n of the second record.

13. A data processing system as defined in claim 7 further comprising:

a communications interface configured to manage communications between the processor and the database.

14. A data processing system as defined in claim 13 wherein the communications interface is a computer bus.

15. A data processing system as defined in claim 13 wherein the communications interface is a network interface that provides access to the database over a data network.