Automated unit finding for numeric information retrieval

• US 8,275,747 B2
• Filed: 04/05/2010
• Issued: 09/25/2012
• Est. Priority Date: 04/05/2010
• Status: Active Grant

• Current Assignee: Quantifind Incorporated
• Original Assignee: Quantifind Incorporated
• Inventors: Tuchman, Ari K., Stockton, John K.
• Primary Examiner(s): Woo, Isaac M
• Application Number:
  US12/754,249
• Publication Number:
  US 20110246458A1
• Time in Patent Office: 904 Days
• Field of Search: 707800-899
• US Class Current: 707/640
• CPC Class Codes: G06F 16/901 Indexing; Data structures therefor; Storage structures for retrieval from the web G06F16/951
• CPC Class Codes: G06F 16/951 Indexing; Web crawling techniques

Abstract

The present invention is related to the task of retrieving numeric information in response to a textual keyword-based query by automatically associating a unit to the type of data being retrieved. An information retrieval system is presented which suggests a unit for data exploration by leveraging the local environment of numeric data across the corpus. This local environment is parsed, including through natural language processing and proximity-based techniques, to determine units relevant to particular keyword phrases. The system also relies on knowledge of semantically and scientifically related units to optimize their binning for suggested unit scoring.

Claims

1. A method using a computer system for automated retrieval of numeric information from a set of electronic source documents, comprising:

receiving an electronic query comprising one or more keywords from a remote user computer system via a telecommunications network;

accessing a computer-searchable index comprising a plurality of searchable index entries, at least a plurality of the entries each including a representation of an associated unit of measure extracted from the electronic source documents;

automatically determining a query unit, at least partly by identifying one or more related occurrences within the electronic source documents of said one or more keywords and one or more of said units of measure and by scoring the one or more related occurrences to reflect how closely related each of said one or more keywords and one or more of said units of measure appear to be within the source document; and

assessing the relevance of one or more of the index entries to the electronic query, at least partly based on a comparison of the query unit and the unit associated with one or more of said index entries.

2. The method of claim 1, wherein scoring the related occurrences employs one or more natural language processing techniques.

3. The method of claim 1, wherein scoring the related occurrences comprises calculating a measure of proximity between the keyword and the unit within text of the source document.

4. The method of claim 3, wherein the measure of proximity within the source document reflects one or more of:

a number of words separating the unit and the keyword, and punctuation separating the unit and the keyword.

5. The method of claim 1, wherein automatically determining a query unit further comprises aggregating a plurality of scores assigned respectively to each of the related occurrences for said one or more units of measure.

6. The method of claim 5, wherein aggregating further comprises scaling the aggregation of said scores.

7. The method of claim 5, wherein aggregating further comprises combining a plurality of scores assigned to each of the related occurrences for two or more non-identical, but related, units of measure.

8. The method of claim 7, wherein determining a query unit further comprises selecting, as the query unit, the unit of measure with a highest individual score from among said two or more non-identical but related units of measure whose scores were combined.

9. The method of claim 1, wherein the electronic query further includes one or more document-level constraint keywords, and wherein said method does not assess proximity within the electronic source documents of said document-level constraint keywords and said units of measure in determining the query unit.

10. The method of claim 1, further including displaying graphical output based on the assessed relevance of the index entries to the electronic query.

11. The method of claim 10, wherein:

the graphical output includes an interactive facility listing one or more alternative suggested query units and allowing a user to select among the suggested query units, and

the method further includes assessing the relevance of the index entries based on comparison with the user-selected query unit, and displaying revised graphical output reflecting said assessment.

12. The method of claim 1, further including automatically determining one or more additional keywords related to said query, at least partly based on a frequency of occurrence of said additional keywords within the source documents corresponding to the index entries assessed as relevant to the query.

13. The method of claim 12, wherein one or more of said additional keywords is used to further assess the relevance of said index entries.

14. The method of claim 1, wherein each of the index entries represents an item of numerical data comprising a unit of measure and a numerical value, said items of numerical data extracted from the electronic source documents.

15. The method of claim 1, wherein the computer system comprises a plurality of distributed computers in communication with each other via a telecommunications network.

16. Apparatus for numeric information retrieval, comprising:

electronic memory storing a computer-searchable index, the index comprising a plurality of searchable index entries, at least a plurality of the entries each including a representation of an associated unit of measure extracted from the electronic source documents;

a computer system, being programmed to receive a query, said query comprising one or more keywords;

said computer system being further programmed to automatically generate a query unit, at least partly by determining a correlation within the electronic source documents between the keywords and one or more of the units included in the index, wherein determining the correlation includes scoring how closely related each of said one or more keywords and one or more of said units appear to be within the electronic source documents; and

said computer system being further programmed to assess the relevance of one or more of the index entries to the query, based at least partly on a comparison of the query unit and the unit associated with each of said index entries.

17. The apparatus of claim 16, wherein scoring comprises calculating a measure of proximity between the keyword and the unit within text of the source document.

18. The apparatus of claim 16, wherein the electronic query further includes one or more document-level constraint keywords, and wherein said computer system is not programmed to asses proximity within the electronic source documents of said document-level constraint keywords and said units of measure in generating the query unit.

19. The apparatus of claim 16, said computer system being further programmed to display interactive graphical output based on the assessed relevance of the index queries to the electronic query, and wherein said interactive graphical output includes one or more alternative suggested query units and allows a user to select among the suggested query units.

20. The apparatus of claim 16, said computer system being further programmed to automatically determine one or more additional keywords related to said query, at least partly based on a frequency of occurrence of said additional keywords within the source documents corresponding to the index entries assessed as relevant to the query.

21. The apparatus of claim 16, wherein each of the index entries represents an item of numerical data comprising a unit of measure and a numerical value, said items of numerical data extracted from the electronic source documents.

22. The apparatus of claim 16, wherein the computer system comprises a plurality of computers in communication with each other via a telecommunications network.