Process for constructing a semantic knowledge base using a document corpus

• US 8,155,951 B2
• Filed: 04/03/2009
• Issued: 04/10/2012
• Est. Priority Date: 06/12/2003
• Status: Expired due to Fees

• Current Assignee: Patrick William Jamieson
• Original Assignee: Patrick William Jamieson
• Inventors: Jamieson, Patrick William
• Primary Examiner(s): Han, Qi
• Application Number:
  US12/417,737
• Publication Number:
  US 20100063799A1
• Time in Patent Office: 1,103 Days
• Field of Search: 704/10, 704/4, 704/1, 704/9
• US Class Current: 704/10
• CPC Class Codes: G06F 16/36 Creation of semantic tools, e.g. ontology or thesauri

Abstract

Related free-text documents, a corpus, are used to empirically derive a semantic knowledge base through a method in which documents are segmented into unique sentences, and then used to define sentential propositions which are arranged in a knowledge hierarchy. The method takes compound natural language sentences and transforms them to simple sentences by a process that is a part of the invention. A knowledge editor enables a domain expert using the methods of the invention to map the sentences in the corpus to sentential proposition(s). The resulting knowledge base can be used to semantically analyze documents in data mining and decision support applications, and can assist word processors or speech recognition devices. The invention is illustrated in connection with radiology reports, but it has wide applicability.

Claims

1. A computer assisted method for creating a semantic knowledge base for a document corpus describing physical or tangible objects, consisting of unique sentential propositions, said method comprising the steps of:

(a) segmenting at least a portion of a document corpus into individual sentences using a computer processor; and

(b) reducing each said sentence to one or more simple or complex sentences; and

(c) displaying candidate propositions to a domain expert for each said simple or complex sentence using a computer processor; and

(d) said domain expert creating a unique sentential proposition to represent the entire meaning for each said simple or complex sentence whose meaning is not represented in said knowledge base using a knowledge editor; and

(e) adding said sentential proposition(s) using a computer processor to said knowledge base, wherein a single sentential proposition represents the entire meaning of a simple or complex sentence.

2. The method of claim 1 wherein a new sentential proposition is placed in a knowledge hierarchy characterized by relationships that may include subsumption, “

is-a”

, “

part-of”

, and causal relationships.

3. The method of claim 1 wherein sentences from the document corpus are displayed in the knowledge editor in order of frequency.

4. The method of claim 1 wherein a selected sentence from the corpus is displayed in a knowledge editor window that includes the documents from which the sentence is derived.

5. The method of claim 1 wherein a newly created sentential proposition is further characterized by its semantic type.

6. The method of claim 1 wherein a sentential proposition is further displayed in a knowledge editor with a distinctive mark, or unique color, depending on the semantic type.

7. The method of claim 1 wherein candidate sentential propositions that may represent the meaning of a selected sentence are retrieved based on their string similarity to sentential propositions in the knowledge base.

8. The method of claim 1 wherein candidate sentential propositions that may represent the meaning of a selected sentence from the corpus are retrieved based on their string similarity to previously mapped sentences in the corpus.

9. The method of claim 1 wherein an unmappable sentence from the corpus is marked with the attribute “

delete”

, “

invalid”

, or “

skip”

or similar attribute.

10. A computer assisted method for creating a semantic mapping table consisting of associations between a sentence from a corpus and sentential propositions in a knowledge base describing physical or tangible objects, comprising the steps of:

a. reducing each said sentence to one or more said simple or complex sentences; and

b. for each said simple or complex sentence using a computer processor to display candidate propositions which may represent the entire meaning of said simple or complex sentence to a domain expert; and

c. said domain expert using a knowledge editor to associate one said candidate sentential proposition for each said simple or complex sentence; and

,

d. storing said associations using a computer processor in said mapping table, wherein a single sentential proposition represents the entire meaning of a simple or complex sentence.

11. The method of claim 10 wherein the associations are further qualified by one or more context markers.

12. The method of claim 10 wherein candidate sentential propositions that may represent the meaning of a selected sentence are retrieved based on their string similarity to sentential propositions in the knowledge base.

13. The method of claim 10 wherein candidate sentential propositions that may represent the meaning of a selected sentence from the corpus are retrieved based on their string similarity to previously mapped sentences in the corpus.

14. The method of claim 10 wherein similar sentences are found in the corpus and mapped to the same propositions.