Use of neural networks for annotating search results

• US 8,533,130 B2
• Filed: 11/15/2009
• Issued: 09/10/2013
• Est. Priority Date: 09/26/2005
• Status: Active Grant

• Current Assignee: Callahan Cellular LLC (Intellectual Ventures LLC)
• Original Assignee: Dranias Development LLC (Intellectual Ventures LLC)
• Inventors: Ershov, Alexander V.
• Primary Examiner(s): CHANG, LI WU
• Application Number:
  US12/618,733
• Publication Number:
  US 20110047111A1
• Time in Patent Office: 1,395 Days
• Field of Search: None
• US Class Current: 706/12
• CPC Class Codes: G06F 16/9535 Search customisation based on user profiles and personalisation

Abstract

A system for generating annotations of a document, including a plurality of neurons connected as a neural network, the neurons being associated with words, sentences and documents. An activity regulator regulates a minimum and/or maximum number of neurons of the neural network that are excited at any given time. The neurons are displayed to a user and identify the neurons that correspond to sentences containing a predetermined percentage of document meaning. The annotations can be also based on a context of the user'"'"'s search query. The query can include keywords, documents considered relevant by the user, or both. Positions of the neurons relative to each other can be changed on a display device, based on input from the user, with the change in position of one neuron changing the resulting annotations. The input from the user can also include changing a relevance of neurons relative to each other, or indicating relevance or irrelevance of a document or sentence.

Claims

1. An apparatus comprising:

a memory;

a processor operatively coupled to the memory; and

a neural network comprising;

a plurality of word neurons;

a plurality of sentence neurons;

at least one document neuron;

a plurality of first connections between at least a portion of the plurality of word neurons and the plurality of sentence neurons; and

a plurality of second connections between at least a portion of the word neurons and the at least one document neuron, wherein the neural network is configured to excite a first sentence neuron of the plurality of sentence neurons in response to excitation of the at least one document neuron;

wherein the processor is configured to change a position of the plurality of word neurons on a display based on an input, and wherein the change in the position of one word neuron changes annotation corresponding to at least one of the plurality of sentence neurons.

2. The apparatus of claim 1, wherein the plurality of word neurons are organized into an input layer that receives an input query.

3. The apparatus of claim 2, wherein at least some of the plurality of word neurons of the input layer have connections between one other.

4. The apparatus of claim 1, wherein the first sentence neuron has a highest relevance to document meaning based on a predetermined percentage of document meaning.

5. The apparatus of claim 4, wherein the predetermined percentage of document meaning is a degree of relevance of the first sentence neuron to the at least one document neuron.

6. The apparatus of claim 1, wherein the first sentence neuron represents an annotation of a document.

7. The apparatus of claim 1, wherein the neural network is a bidirectional neural network.

8. The apparatus of claim 1, wherein the processor is configured to receive an input query, and wherein the input query comprises one or more documents considered relevant by a user.

9. The apparatus of claim 1, wherein the processor is configured to receive the input, and wherein the input comprises a change in relevance of neurons relative to one other.

10. The apparatus of claim 1, wherein the processor is configured to receive the input, wherein the input comprises an indication of irrelevance of a sentence neuron, and wherein the indication of the irrelevance of the sentence neuron inhibits neurons of the neural network.

11. The apparatus of claim 1, wherein the processor is configured to regulate a sum of all activity of all active neurons of the neural network that are excited at any given time.

12. A method comprising:

generating a plurality of word neurons;

generating a plurality of sentence neurons;

generating at least one document neuron, wherein the plurality of word neurons, the plurality of sentence neurons, and the at least one document neuron form at least a portion of a neural network;

forming, with a processor of a computing device, a plurality of first connections between at least a portion of the plurality of word neurons and the plurality of sentence neurons;

forming, with the processor of the computing device, a plurality of second connections between at least a portion of the plurality of word neurons and the at least one document neuron;

exciting a first sentence neuron of the plurality of sentence neurons in response to excitation of the at least one document neuron; and

changing a position of the plurality of word neurons on a display based on an input, and wherein the change in the position of one word neuron changes annotation corresponding to at least one of the plurality of sentence neurons.

13. The method of claim 12, wherein the first sentence neuron represents an annotation of a document.

14. The method of claim 12, wherein the neural network is a bidirectional neural network.

15. The method of claim 12, further comprising receiving an input query, wherein the input query comprises one or more documents considered relevant by a user.

16. The method of claim 12, further comprising organizing the plurality of word neurons into an input layer that receives an input query.

17. The method of claim 12, further comprising regulating a sum of all activity of all active neurons of the neural network that are excited at any given time.

18. A non-transitory computer-readable medium having computer-readable instructions stored thereon, wherein the computer-readable instructions comprise:

instructions to generate a plurality of word neurons;

instructions to generate a plurality of sentence neurons;

instructions to generate at least one document neuron, wherein the plurality of word neurons, the plurality of sentence neurons, and the at least one document neuron form at least a portion of a neural network;

instructions to form a plurality of first connections between at least a portion of the plurality of word neurons and the plurality of sentence neurons;

instructions to form a plurality of second connections between at least a portion of the plurality of word neurons and the at least one document neuron;

instructions to excite a first sentence neuron of the plurality of sentence neurons in response to excitation of the at least one document neuron; and

instructions to change a position of the plurality of word neurons on a display based on an input, and wherein the change in the position of one word neuron changes annotation corresponding to at least one of the plurality of sentence neurons.

19. The non-transitory computer-readable medium of claim 18, further comprising instructions to organize the plurality of word neurons into an input layer that receives an input query.

20. The non-transitory computer-readable medium of claim 18, further comprising instructions to regulate a sum of all activity of all active neurons of the neural network that are excited at any given time.

21. The non-transitory computer-readable medium of claim 18, further comprising instructions to change a relevance of word neurons relative to one other in response to a received input.