Processing text with domain-specific spreading activation methods

• US 10,713,440 B2
• Filed: 10/02/2018
• Issued: 07/14/2020
• Est. Priority Date: 01/04/2007
• Status: Active Grant

• Current Assignee: Children's Hospital & Medical Center
• Original Assignee: Children's Hospital & Medical Center
• Inventors: Pestian, John P., Matykiewicz, Pawel, Duch, Wlodzislaw, Glauser, Tracy A., Kowatch, Robert A., Grupp-Phelan, Jacqueline M., Sorter, Michael
• Primary Examiner(s): Ortiz-Sanchez, Michael
• Application Number:
  US16/149,698
• Publication Number:
  US 20190042563A1
• Time in Patent Office: 651 Days
• CPC Class Codes: G06F 16/3344 using natural language analysis
• CPC Class Codes: G06F 40/117 Tagging; Marking up details of markup languages G06F40/143; Designating a block; Setting of attributes style sheets, e.g. eXtensible Stylesheet Language Transformation [XSLT], G06F40/154
• CPC Class Codes: G06F 40/205 Parsing
• CPC Class Codes: G06F 40/232 Orthographic correction, e.g. spell checking or vowelisation
• CPC Class Codes: G06F 40/237 Lexical tools
• CPC Class Codes: G06F 40/253 Grammatical analysis; Style critique
• CPC Class Codes: G06F 40/30 Semantic analysis
• CPC Class Codes: G06F 40/40 Processing or translation of natural language natural language analysis G06F40/20; semantic analysis G06F40/30
• CPC Class Codes: G06Q 10/10 Office automation; Time management
• CPC Class Codes: G16H 40/20 for the management or administration of healthcare resources or facilities, e.g. managing hospital staff or surgery rooms
• CPC Class Codes: G16H 50/70 for mining of medical data, e.g. analysing previous cases of other patients
• CPC Class Codes: G16H 70/60 relating to pathologies

Abstract

A method for performing natural language processing of free text using domain-specific spreading activation. Embodiments of the present invention ontologize free text using an algorithm based on neurocognitive theory by simulating human recognition, semantic, and episodic memory approaches. Embodiments of the invention may be used to process clinical text for assignment of billing codes, analyze suicide notes or legal discovery materials, and for processing other collections of text. Further, embodiments of the invention may be used to more effectively search large databases, such as a database containing a large number of medical publications.

Claims

1. One or more non-transitory electronic memory devices including computer instructions for performing a method comprising using a using a central processing unit (CPU) to create, from a text of one or more suicide notes containing natural language stored in a non-transitory electronic memory device, a semantic network having a plurality of nodes corresponding to sets of semantic and episodic concepts and weighted links between the sets of semantic and episodic concepts,

the sets of semantic concepts consisting of a plurality of semantic concepts associated with a plurality of known words;

the sets of episodic concepts consisting of a plurality of episodic concepts associated with the semantic concepts; and

utilizing, using the CPU, spreading activation algorithms to refine the weighted links in the semantic network.

2. The one or more non-transitory memory devices of claim 1, further comprising, prior to the step of creating a semantic network, the steps of

identifying, using the CPU, one or more of a plurality of groups of characters of the text as corresponding to at least one of the plurality of known words;

querying a first database comprising associations between known words and semantic concepts to obtain the sets of one or more semantic concepts associated with each of the known words;

querying a second database comprising associations between episodic concepts, known words, and semantic concepts to obtain the sets of one or more episodic concepts associated with the sets of semantic concepts, the episodic concepts being separate from the semantic concepts.

3. The one or more non-transitory memory devices of claim 1, further comprising, following the step of creating a semantic network, the step of selecting, using the CPU, at least one of the concepts from the sets of semantic and episodic concepts based upon an associated weight for at least one node of the plurality of nodes derived from the step of utilizing spreading activation.

4. The one or more non-transitory memory devices of claim 2, further comprising, using the CPU, configuring the text prior to the identifying step, by including at least one of tagging parts of speech, replacing abbreviations with words, and correcting misspelled words.

5. The one or more non-transitory memory devices of claim 3, further comprising outputting, using the CPU, the selected at least one of the concepts.

6. The one or more non-transitory memory devices of claim 1, wherein the text is of a plurality of documents and the method further comprises identifying, using the CPU, a subset of the plurality of documents by identifying at least two documents having associations with the selected at least one of the concepts.

7. The one or more non-transitory memory devices of claim 6, further comprising outputting, using the CPU, an identification of one or more portions of each of the at least two documents having associations with the selected at least one of the concepts.

8. The one or more non-transitory memory devices of claim 1, wherein the method further comprises evaluating, using the CPU, the suicide note for concepts indicative of suicidal intent.

17. The one or more non-transitory memory devices of claim 1, wherein the method further comprises, following the step of creating a semantic network,

iteratively expanding the semantic network with additional concepts taken from the one or more databases and linked to respective nodes in the semantic network to iteratively add new nodes to the semantic network for such additional concepts, each new node including a weighted link with an existing node, the additional concepts being separate from the first set of semantic concepts and including at least one episodic concept.

18. The one or more non-transitory memory devices of claim 17, wherein the computer instructions are further configured to perform the step of repeating the iteratively expanding continuously until a termination requirement is met.

19. The one or more non-transitory memory devices of claim 18, wherein the termination requirement is a value of a weighted link falling below a predefined threshold.

9. One or more non-transitory electronic memory devices including computer instructions for performing a method comprising using a CPU for processing a text of one or more suicide notes containing natural language, the method comprising

tagging, using the CPU, parts of speech in the text;

recognizing, using the CPU, known words in the text; and

creating, using the CPU, a semantic network,

the semantic network including at least one of the recognized known words and at least one relationship with at least one semantic concept associated with at least one of the recognized known words; and

supplementing the semantic network by iteratively adding additional concepts and additional relationships to the semantic network until a termination requirement is met, each additional concept being associated with at least a prior one of the concepts and additional concepts in the semantic network by a respective additional relationship, at least one of the additional concepts being an episodic concept separate from the at least one semantic concept, and evaluating, using the CPU, the suicide note for concepts indicative of suicidal intent.

10. The one or more non-transitory memory devices of claim 9, further comprising

weighting each of the at least one relationships and each of the additional relationships with a weighting value reflecting the strength of each relationship and additional relationship;

determining a minimum threshold weighting value; and

terminating the iterative growth of any network node in which the weighting between the relationships and the additional relationships do not satisfy the minimum threshold weighting value.

11. The one or more non-transitory memory devices of claim 10, further comprising

comparing the at least one semantic concept and the additional concepts to a list of known relevant concepts to generate a list of identified relevant concepts;

and

providing an output based on at least one of a number and a significance of the identified relevant concepts.

12. The one or more non-transitory memory devices of claim 11, wherein the output pertains to a probability of a particular occurrence.

13. The one or more non-transitory memory devices of claim 12, wherein the particular occurrence is a suicide attempt.

14. The one or more non-transitory memory devices of claim 10, wherein the text includes a plurality of documents and the method further comprises

receiving a query including a search concept;

and

displaying a list of documents including one or more of the plurality of documents that is associated with the at least one semantic concept and the additional concepts that matches the search concept.

15. The one or more non-transitory memory devices of claim 14, wherein the list of documents is sorted by the weighting value pertaining to at least one relationship or additional relationship between the search concept and the corresponding recognized known word.

16. The one or more non-transitory memory devices of claim 15, wherein the one or more episodic concepts are uniquely associated with a patient'"'"'s prior clinical history.

20. A method for machine learning from corpora, the method comprising annotating a training corpora, creating a disambiguated and anonymous part-of-speech tagged clinical corpus by disambiguating 30-50 clinical documents from the following subspecialties, radiology, nephrology, pulmonary, behavioral medicine, psychiatry, rheumatology, pathology, cardiology, allergy and immunology, critical care, hematology/oncology, and human genetics;

and employing spreading activation algorithms for concept mapping, wherein the step of employing spreading activation algorithms for concept mapping comprises assignment of identifiers, full concept identification, concept weighting, and decay function computation.