Confidence links between name entities in disparate documents

• US 8,527,522 B2
• Filed: 12/29/2008
• Issued: 09/03/2013
• Est. Priority Date: 09/05/2008
• Status: Expired due to Fees

• Current Assignee: Cxense ASA
• Original Assignee: Ramp Holdings Incorporated (Clean Harbors Incorporated)
• Inventors: Baron, Alex, Freedman, Marjorie Ruth, Weischedel, Ralph M., Boschee, Elizabeth Megan
• Primary Examiner(s): VU, BAI DUC
• Application Number:
  US12/344,871
• Publication Number:
  US 20100076972A1
• Time in Patent Office: 1,709 Days
• Field of Search: None
• US Class Current: 707/749
• CPC Class Codes: G06F 16/355 Class or cluster creation or modification
• CPC Class Codes: G06F 40/295 Named entity recognition

Abstract

The invention relates to cross-document entity co-reference systems in which naturally occurring entity mentions in a document corpus are analyzed and transformed into name clusters that represent global entities. In a first aspect of the invention, a name variation module analyzes naturally occurring names of entities extracted from the document corpus and provides an initial set of equivalent names that could refer to the same real world entity. In a second aspect of the invention, a disambiguation module takes the initial set of equivalent names and uses an agglomerative clustering algorithm to disambiguate the potentially co-referent named entities.

Claims

1. A system that detects similarities between name strings in a document set, comprising:

a processor and a memory, the memory comprising a preprocessing module, a matching module and a generation module;

the preprocessing module configured to;

extract a plurality of name strings from the document set by generating additional name strings based on an alternative spelling of one or more name strings in the document set, each name string comprising a similar entity with names that are misspelled, mistranslated, incorrectly transcribed, have multiple aliases, and/or have multiple equally valid spellings, the alternate spelling comprising determining typical misspellings, creating language specific lists of spelling corrections, and generating the alternative spelling based on the spelling corrections;

the matching module configured to;

detect possible matching pairs from the plurality of name strings, and

detect a plurality of similarity scores to each of the possible matching pairs using a plurality of algorithms that execute in parallel; and

the generation module configured to;

generate a set of equivalent names by its relating name strings from the possible matching pairs based on a comparison between the similarity scores and a threshold.

2. The system of claim 1, wherein the matching module is configured to assign similarity scores by a signing an algorithm-specific to each possible matching pair for each of the plurality of algorithms.

3. The system of claim 2, wherein the algorithm-specific score for an algorithm is based at least in part on a confidence level associated with a source of equivalency used by the algorithm.

4. The system of claim 1, wherein each of the plurality of algorithms the text a different type of equivalence between the name strings in each of the possible matching pairs.

5. The system of claim 1, wherein the generation module is configured to generate a set of equivalent names by selecting from the plurality of similarity scores for each possible matching pair the highest similarity score for the pair and comparing the highest similarity score for each pair to the threshold.

6. The system of claim 1, wherein the memory further comprises a storage module configured to store a set of equivalent name strings in a database.