Generating searchable keywords

• US 7,849,078 B2
• Filed: 06/07/2006
• Issued: 12/07/2010
• Est. Priority Date: 06/07/2006
• Status: Active Grant

• Current Assignee: SAP SE
• Original Assignee: SAP AG (SAP SE)
• Inventors: Meyer, Marek, Hildebrandt, Tomas, Faatz, Andreas
• Primary Examiner(s): Mofiz; Apu M
• Assistant Examiner(s): Nguyen; Cindy
• Application Number:
  US11/449,030
• Publication Number:
  US 20070288451A1
• Time in Patent Office: 1,644 Days
• Field of Search: 707/3, 707/5, 707/6, 707/723, 706/12, 706/47
• US Class Current: 707/723
• CPC Class Codes: G06F 16/313 Selection or weighting of terms for indexing
• CPC Class Codes: G06F 16/907 Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually

Abstract

Generating searchable keywords includes obtaining sets of parameters that include weights, where the weights correspond to actions for changing first words into predicted keywords, using the sets of parameters to obtain predicted keywords from the first words, determining a prediction value for each set of parameters, where the prediction values are based on an extent to which predicted keywords for a set of parameters match predefined keywords, ranking the sets of parameters based on prediction values, and applying a selected set of parameters to a second word to obtain a new keyword, wherein the selected set of parameters is applied based on rank.

Claims

1. A method that is computer-implemented, the method comprising:

obtaining sets of parameters comprising weights associated with actions to change one or more characters in first words to generate predicted keywords, the weights also being associated with locations of the first words within a set of metadata;

using the sets of parameters to obtain predicted keywords from the first words;

determining a prediction value for each set of parameters, the prediction values being based on at least one of a precision value and a recall value, the precision value including a number of predicted keywords that are also in a predefined keyword vocabulary relative to a total number of predicted keywords, and the recall value including a number of keywords in the predefined keyword vocabulary that are also in the predicted keywords relative to a total number of keywords in the predefined keyword vocabulary;

ranking the sets of parameters based on prediction values, wherein the ranking results in ranks that are stored in memory;

applying a selected set of parameters to a second word to obtain a new keyword, wherein the selected set of parameters that is applied is applied in order of rank;

receiving feedback indicating whether the new keyword is acceptable; and

if the new keyword is not acceptable, applying another set of parameters, in order of rank, to the second word to obtain an additional keyword.

2. The method of claim 1, wherein the selected set of parameters has a prediction value that indicates a best match between predicted keywords and predefined keywords.

3. The method of claim 2, wherein the other set of parameters has a prediction value that indicates a next best match between predicted keywords and predefined keywords.

4. The method of claim 1, wherein the weights comprise costs assigned to actions required to change the first words into the predicted keywords, the actions comprising character transitions that comprise the actions to change one or more characters at different positions.

5. The method of claim 4, wherein the character transitions comprise one or more of:

deleting an end position character, deleting a middle position character, deleting a beginning position character, editing the end position character, editing a middle position character, editing the beginning position character, inserting the end position character, inserting a middle position character, and inserting the beginning position character; and

wherein the end position comprises a last character position in the first word, the beginning position comprises a first character position in the first word, and a middle position comprises a position in the first word between the end position and the beginning position.

6. The method of claim 1, wherein a weight for a set of parameters is based on whether a first word is part of a title of data or a body of the data.

7. The method of claim 1, wherein the new keyword is part of metadata for a learning object, the learning object being stored in a repository that is part of an electronic learning system.

8. The method of claim 1, wherein the prediction value is determined for plural predicted keywords, the plural predicted keywords corresponding to keywords that are related to the first word hierarchically.

9. The method of claim 1, further comprising:

storing the new keyword as metadata for a data object;

receiving a search query; and

returning information relating to the data object if at least part of the search query matches the new keyword.

10. The method of claim 1, further comprising obtaining the first words by removing stop words from text that includes the first words.

11. The method of claim 2, wherein the feedback is received from a user;

and wherein the method further comprises;

receiving feedback indicating whether the additional keyword is acceptable;

if the additional keyword is not acceptable, applying another set of parameters to the second word to obtain a next keyword; and

repeating receiving and applying until an acceptable keyword is obtained.

12. One or more machine-readable media for storing instructions that are executable by one or more processing devices to perform the following actions:

obtaining sets of parameters comprising weights associated with actions to change one or more characters in first words to generate predicted keywords, the weights also being based on locations of the first words within a set of metadata;

using the sets of parameters to obtain predicted keywords from the first words;

determining a prediction value for each set of parameters, the prediction values being based on at least one of a precision value and a recall value, the precision value including a number of predicted keywords that are also in a predefined keyword vocabulary relative to a total number of predicted keywords, and the recall value including a number of keywords in the predefined keyword vocabulary that are also in the predicted keywords relative to a total number of keywords in the predefined keyword vocabulary;

ranking the sets of parameters based on prediction values;

applying a selected set of parameters to a second word to obtain a new keyword, wherein the selected set of parameters that is applied is applied in order of rank;

receiving feedback indicating whether the new keyword is acceptable; and

if the new keyword is not acceptable, applying another set of parameters, in order of rank, to the second word to obtain an additional keyword.

13. The one or more machine-readable media of claim 12, wherein the selected set of parameters has a prediction value that indicates a best match between predicted keywords and predefined keywords.

14. The one or more machine-readable media of claim 13, wherein the other set of parameters has a prediction value that indicates a next best match between predicted keywords and predefined keywords.

15. The one or more machine-readable media of claim 12, wherein the weights comprise costs assigned to actions required to change the first words into the predicted keywords, the actions comprising character transitions that comprise the actions to change one or more characters at different positions.

16. The one or more machine-readable media of claim 15, wherein the character transitions comprise one or more of:

deleting an end position character, deleting a middle position character, deleting a beginning position character, editing the end position character, editing a middle position character, editing the beginning position character, inserting the end position character, inserting a middle position character, and inserting the beginning position character; and

wherein the end position comprises a last character position in the first word, the beginning position comprises a first character position in the first word, and a middle position comprises a position in the first word between the end position and the beginning position.

17. The one or more machine-readable media of claim 12, wherein a weight for a set of parameters is based on whether a first word is part of a title of data or a body of the data.

18. The one or more machine-readable media of claim 12, wherein the new keyword is part of metadata for a learning object, the learning object being stored in a repository that is part of an electronic learning system.

19. The one or more machine-readable media of claim 12, wherein the prediction value is determined for plural predicted keywords, the plural predicted keywords corresponding to keywords that are related to the first word hierarchically.

20. The one or more machine-readable media of claim 12, further comprising instructions that are executable by one or more processing devices to perform the following actions:

storing the new keyword as metadata for a data object;

receiving a search query; and

returning information relating to the data object if at least part of the search query matches the new keyword.