Systems, methods, and computer readable media for visualization of semantic information and inference of temporal signals indicating salient associations between life science entities

Patent metadata

Field	Value
Publication number	US-10360507-B2
Application number	US-201715713426-A
Country	US
Kind code	B2
Filing date	Sep 22, 2017
Priority date	Sep 22, 2016
Publication date	Jul 23, 2019
Grant date	Jul 23, 2019

Abstract

Disclosed systems, methods, and computer readable media can detect an association between semantic entities and generate semantic information between entities. For example, semantic entities and associated semantic collections present in knowledge bases can be identified. A time period can be determined and divided into time slices. For each time slice, word embeddings for the identified semantic entities can be generated; a first semantic association strength between a first semantic entity input and a second semantic entity input can be determined; and a second semantic association strength between the first semantic entity input and semantic entities associated with a semantic collection that is associated with the second semantic entity can be determined. An output can be provided based on the first and second semantic association strengths.

First claim

What is claimed is:

1. A computer-implemented method of detecting an association between semantic entities, comprising:
identifying semantic entities present in one or more knowledge bases, wherein the semantic entities include one or more of single words or multi-word phrases;
identifying semantic collections associated with the semantic entities, wherein the semantic entities of a semantic collection share the same entity type, and wherein the entity type is one or more of bio-molecules, bio-entities, diseases, phenotypes, hospitals, drugs, medical instruments, medical procedures, indications, biomolecular signals, or genes;
determining a time period for analysis;
dividing the time period into one or more time slices;
generating, for each time slice, a set of word embeddings for the identified semantic entities based on one or more corpora, wherein the set of word embeddings is generated for each time slice independently of a set of word embeddings generated for other time slices;
receiving a first semantic entity input and a second semantic entity input that correspond to two of the identified semantic entities;
determining, for each time slice, a first semantic association strength between the first semantic entity input and the second semantic entity input;
determining, for each time slice, a second semantic association strength between the first semantic entity input and a plurality of semantic entities in a semantic collection that is associated with the second semantic entity input;
generating one or more of summary statistics or temporal analysis for at least one of the one or more time slices, wherein said one or more of the summary statistics or the temporal analysis indicates an emerging association between the first semantic entity input and the second semantic entity input, based on one or more of the first semantic association strength or the second semantic association strength; and
applying said one or more of summary statistics or temporal analysis to a display template to produce an output graphical representation that visually indicates the emerging association between the first semantic entity input and the second semantic entity input to a user on a user device.

2. The computer-implemented method of claim 1 , wherein the one or more corpora comprise structured data and unstructured data.

3. The computer-implemented method of claim 1 , wherein identifying semantic entities includes one or more of: (1) automatic methods of identifying one or more single words or multi-word phrases as semantic entities belonging to semantic collections or (2) selecting one or more single words or multi-word phrases forcibly from the one or more knowledge bases.

4. The computer-implemented method of claim 3 , wherein the one or more single words or multi-word phrases are selected forcibly from information compiled from a structured database.

5. The computer-implemented method of claim 1 , wherein identifying semantic entities is performed on all text in the one or more knowledge bases for the time period.

6. The computer-implemented method of claim 1 , wherein the word embeddings are generated using one or more of Word2vec, AdaGram, fastText, or Doc2vec.

7. The computer-implemented method of claim 1 , wherein the set of word embeddings for a time slice is generated by leveraging a set of word embeddings from a previous time slice.

8. The computer-implemented method of claim 1 , wherein the plurality of semantic entities associated with the semantic collection that is associated with the second semantic entity input does not include the second semantic entity input.

9. The computer-implemented method of claim 1 , wherein the second semantic association strength is a mean, a median, or a percentile of a set of semantic association strengths between the first semantic entity input and the plurality of semantic entities associated with a semantic collection that is associated with the second semantic entity input.

10. The computer-implemented method of claim 1 , further comprising:
detecting an increase in the first semantic association strength of a first time slice relative to the first semantic association strength of a second, subsequent time slice; and
determining whether the increase in the first semantic association strength is statistically significant relative to the corresponding second semantic association.

11. The computer-implemented method of claim 10 , wherein the statistical significance of the increase is determined based on a p-value as a measure of statistical significance of the first semantic association strength relative to the corresponding second semantic association.

12. The computer-implemented method of claim 1 , further comprising:
selecting the first semantic entity input and the second semantic entity input based on a level of co-occurrence between the first semantic entity input and the second semantic entity input in the one or more knowledge bases.

13. The computer-implemented method of claim 12 , wherein the level of co-occurrence between the first semantic entity input and the second semantic entity input is zero.

14. The computer-implemented method of claim 1 , further comprising:
receiving the first semantic entity input and the second semantic entity input from a user.

15. The computer-implemented method of claim 1 , further comprising:
determining, for each time slice, a count of documents present in the one or more corpora containing the first semantic entity input and the second semantic entity input; and
determining a time difference between (1) a first date associated with an increase in the first semantic association strength for a first time slice relative to the first semantic association strength for a second, subsequent time slice and (2) a second date associated with an increase in a count of documents containing the first semantic entity input and the second semantic entity input for a third time slice relative to a count of documents containing the first semantic entity input and the second semantic entity input for a fourth time slice.

16. The computer-implemented method of claim 15 , further comprising:
detecting the increase in the count of documents containing the first semantic entity input and the second semantic entity input based on a slope of a curve in a fixed axis, wherein the curve is based on the time period on an x-axis of the curve and the count of documents on a y-axis of the curve.

17. The computer-implemented method of claim 15 , further comprising:
detecting the second increase in the count of documents containing the first semantic entity input and the second semantic entity input based on a document count threshold.

18. The computer-implemented method of claim 1 , wherein the graphical representation includes a graph line that is created by plotting each of the first semantic association strengths for each of the time slices over the time period based on the display template.

19. The computer-implemented method of claim 1 , wherein the graphical representation includes a graph line that is created by plotting each of mean second semantic association strengths for each of the time slices over the time period based on the display template.

20. The computer-implemented method of claim 1 , wherein the graphical representation includes a graph line that is created by plotting a count of documents present in the one or more corpora containing the first semantic entity input and the second semantic entity input for each of the time slices o

Assignees

• Nference Inc

Inventors

• Aravamudan Murali
• Soundararajan Venkataramanan
• Rajasekharan Ajit
• Gibson William

Classifications

• G06N3/044

  Recurrent networks, e.g. Hopfield networks · CPC title

• G06F40/30 Primary

  Semantic analysis · CPC title

• G06N5/022

  Knowledge engineering; Knowledge acquisition · CPC title

• G06N5/04 Primary

  Inference or reasoning models · CPC title

• G16B50/00

  ICT programming tools or database systems specially adapted for bioinformatics · CPC title