System and methods for units-based numeric information retrieval

• US 9,830,378 B2
• Filed: 05/12/2014
• Issued: 11/28/2017
• Est. Priority Date: 06/26/2009
• Status: Active Grant

• Current Assignee: Quantifind Incorporated
• Original Assignee: Quantifind Incorporated
• Inventors: Stockton, John K., Tuchman, Ari K.
• Primary Examiner(s): Fan, Shiow-Jy
• Application Number:
  US14/275,840
• Publication Number:
  US 20140250130A1
• Time in Patent Office: 1,296 Days
• CPC Class Codes: G06F 16/313 Selection or weighting of terms for indexing
• CPC Class Codes: G06F 16/316 Indexing structures
• CPC Class Codes: G06F 16/334 Query execution G06F16/335 takes precedence
• CPC Class Codes: G06F 16/3347 using vector based model
• CPC Class Codes: G06F 16/338 Presentation of query results
• CPC Class Codes: G06F 16/93 Document management systems
• CPC Class Codes: G06F 2216/11 Patent retrieval

Abstract

An information retrieval and analysis system for numeric data which provides high precision and recall for numeric search and uses a methodology for determining contextualization of the extracted data. The capabilities include extracting, parsing, and contextualizing numeric data including both a numeric value and an accompanying unit. This system facilitates the organization of largely unstructured numeric data into an inverted index and other database formats. An information retrieval system which enables the exploration and refinement of an extracted numeric data set defined by a search input that may be precise or initially vague. This system also facilitates analyzing and portraying numeric data graphically, creating knowledge by combining data from multiple sources, extracting correlations between seemingly disparate variables, and recognizing numeric data trends. This system uses local natural language processing, mathematical analysis, and expert-based scientific heuristics to score the numeric and contextual relevancy of the data to the query parameters.

Claims

1. A computer apparatus, comprising:

a processor;

a memory communicatively coupled to the processor and storing program instructions causing the processor to;

receive a user-defined query comprising one or more numerical data constraints;

determine the relevancy of the query to each of one or more entries in a computer-searchable electronic index, each of the index entries representing an item of data extracted from at least one of a plurality of electronic source documents, wherein the determining further comprises;

determine a spectral vector for at least a part of the query across the plurality of electronic source documents, the spectral vector comprising a first column of binned numerical values associated with the query and a second column of frequencies of occurrence of the binned numerical values within the plurality of electronic source documents,

identify at least one spectral feature comprising a local maximum or local minimum in the frequencies of occurrence within the spectral vector, and

limit the computer-searchable electronic index to those documents of the plurality of electronic source documents that are associated with the at least one spectral feature, the relevancy being based on association of documents with the at least one spectral feature;

generate a graph, based on the relevancy determination, comprising a separate data point for each of the index entries determined to be most relevant to the query, wherein a plurality of the most relevant index entries can be extracted from a single one of the source documents; and

provide an interactive display of the graph.

2. The apparatus of claim 1, wherein one or more visual characteristics of each point on the graph are correlated with information descriptive of a corresponding index entry.

3. The apparatus of claim 2, wherein the one or more visual characteristics include one or more of color, size, shape, and transparency, and the descriptive information includes one or more of a relevancy score and source document for the corresponding index entry.

4. The apparatus of claim 1, wherein the program instructions further comprise:

identify one or more regions of the graph by one or more clustering algorithms.

5. The apparatus of claim 4, wherein the regions of the graph are each labeled by one or more terms automatically extracted from the source documents.

6. The apparatus of claim 1, wherein the interactive display is responsive to selection of a point on the graph by displaying content from the source document from which the item of numerical data corresponding to the selected point was extracted, at least some of said content located proximate to said item of numerical data within said source document.

7. The apparatus of claim 1, said program instructions further comprising:

in response to interactive selection of a graph data point corresponding to one of the index entries, retrieve at least one source document from which the item of numerical data represented by said index entry was extracted.

8. The apparatus of claim 1, wherein the interactive display is configured to receive user feedback regarding the relevancy determination, and wherein the program instructions further comprise:

in response to the user feedback, modify one or more of;

the index, the query, a library of rules used in the determination of relevancy.

9. The apparatus of claim 1, wherein the extracted item of data comprises a unit of measure and a numerical value.

10. The apparatus of claim 1, wherein determining the relevancy of the query further comprises:

determining a topic comparison vector associated with at least part of the query;

performing a comparison between the topic comparison vector and a term frequency vector from the at least one of a plurality of electronic source documents containing each item of data; and

determining a global keyword score based on the comparison.

11. The apparatus of claim 1, wherein identify at least one spectral feature further comprises:

determine at least one fitting curve for the spectral vector; and

extract at least one center value and at least one width associated with the spectral feature from the at least one fitting curve.

12. A processor-implemented method, comprising:

interactively specifying a query comprising one or more numerical data constraints;

transmitting said query to a remote computing system via a network;

receiving a response to the query via the network from the remote computing system, said response defining a graph comprising a separate data point for each of a plurality of responsive items of numerical data occurring in one or more of a plurality of electronic source documents, said responsive items determined by the remote system to be relevant to the query and wherein a plurality of said relevant items can occur in a single one of the source documents, wherein determining relevance further comprises;

determining a spectral vector for at least a part of the query across the plurality of electronic source documents, the spectral vector comprising a first column of binned numerical values associated with the query and a second column of frequencies of occurrence of the binned numerical values within the plurality of electronic source documents,

identifying at least one spectral feature comprising a local maximum or local minimum in the frequencies of occurrence within the spectral vector, and

limiting the computer-searchable electronic index to those documents of the plurality of electronic source documents that are associated with the at least one spectral feature, the relevance being based on association of documents with the at least one spectral feature; and

interactively displaying the graph.