System and method for faster interfaces on text-based tasks using adaptive memory networks

1. A method of performing question answer (QA) tasks comprising:

entering an input into an encoder portion of an adaptive memory network, wherein the encoder portion parses the input into entities of text for arrangement of memory banks;

organizing the entities into progressively weighted banks within the arrangement of memory banks with a bank controller of the adaptive memory network, wherein an initial memory bank having lowest relevance for lowest relevance entities being closest by logical relationship to the encoder, and a final memory bank having a highest relevance for highest relevance entities being closest by logical relationship to a decoder; and

inferring an answer for the question answer (QA) task with the decoder analyzing only the highest relevance entities in the final memory bank, wherein by limiting the decoder analysis to only the highest relevance entities decreases inference time.

2. The method of claim 1 , wherein the encoder creates entities from words processed from a story and sentences processed from the story.

3. The method of claim 1 , wherein the encoder creates entities directed to scheduling data, the highest relevance entities directed to scheduling data being in the final memory bank, the decoder providing a schedule based upon the highest relevance entities directed to scheduling data.

4. The method of claim 1 , wherein said gradually increasing relevance banks extend from the initial memory bank to the final memory bank, the gradually increasing relevance banks having gradually increasing relevant entities.

5. The method of claim 1 , wherein the memory bank controller organizes the entities using at least one memory bank function selected from the group consisting of memory bank creation, moving entities across memory banks, adding entities to a memory bank, propagating changes across entities, and combinations thereof.

6. The method of claim 5 , wherein the memory bank controller organizes the entities with said memory bank creation, wherein new memory banks for the entities is formed by the bank controller when the entities entropy exceeds a maximum preset entropy.

7. The method of claim 6 , wherein by creating said new memory banks, the adaptive memory network spreads out a concentration of information, which in turn increases the discretizes nodes according to relevance.

8. The method of claim 5 , wherein the memory bank controller organizes the entities with said moving entities across said banks, wherein entities relevant to question are moved from memory banks having a low relevance to memory banks having a higher relevancy.

9. A system for performing question answer (QA) tasks comprising:

a bank controller using a processor device operatively coupled to a memory for organizing entities from a question answer (QA) task into progressively weighted banks within an arrangement of memory banks of an adaptive memory network; and

a decoder that infers an answer for the question answer (QA) task using only highest relevance entities in the progressively weighted banks within the arrangement of memory banks of the adaptive memory network, wherein by limiting the decoder analysis to only the highest relevance entities decreases inference time, wherein the bank controller organizes the entities into progressively weighted banks within the arrangement of memory banks with the bank controller of the adaptive memory network, wherein a final memory bank having a highest relevance for highest relevance entities being closest by logical relationship to the decoder.

10. The system of claim 9 , wherein the memory bank controller organizes the entities with said moving entities across said banks, wherein entities relevant to question are moved from memory banks having a low relevance to memory banks having a higher relevancy.

11. The system of claim 9 , wherein the entities directed to scheduling data, the highest relevance entities directed to scheduling data being in the final memory bank, the decoder that provides a schedule based upon the highest relevance entities directed to scheduling data.

12. The system of claim 9 , wherein the memory bank controller organizes the entities using at least one memory bank function selected from the group consisting of memory bank creation, moving entities across memory banks, adding entities to a memory bank, propagating changes across entities, and combinations thereof.

13. A non-transitory computer readable storage medium comprising a computer readable pro-gram for performing question answer (QA) tasks, wherein the computer readable program when executed on a computer causes the computer to perform the steps of:

organizing entities into progressively weighted banks within an arrangement of memory banks with a bank controller of an adaptive memory network, wherein a final memory bank having a highest relevance for highest relevance entities is positioned closest by logical relationship to a decoder; and

inferring an answer for the question answer (QA) task with the decoder analyzing only the highest relevance entities in the final memory bank, wherein by limiting the decoder analysis to only the highest relevance entities decreases inference time.

14. The non-transitory computer readable storage medium as recited in claim 13 , wherein an encoder creates the entities from words processed from a story, and sentences processed from the story.

15. The non-transitory computer readable storage medium as recited in claim 13 , wherein the entities are directed to scheduling data, the highest relevance entities directed to scheduling data being in the final memory bank, the decoder providing a schedule based upon the highest relevance entities directed to scheduling data.

16. The non-transitory computer readable storage medium as recited in claim 13 , wherein said gradually increasing relevance banks extend from the initial memory bank to the final memory bank, the gradually increasing relevance banks having gradually increasing relevant entities.

17. The non-transitory computer readable storage medium as recited in claim 13 , wherein the memory bank controller organizes the entities using at least one memory bank function selected from the group consisting of memory bank creation, moving entities across memory banks, adding entities to a memory bank, propagating changes across entities, and combinations thereof.

18. The non-transitory computer readable storage medium as recited in claim 17 , wherein the memory bank controller organizes the entities with said memory bank creation, wherein new memory banks for the entities is formed by the bank controller when the entities entropy exceeds a maximum preset entropy.

19. The non-transitory computer readable storage medium as recited in claim 17 , wherein by creating said new memory banks, the adaptive memory network spreads out a concentration of information, which in turn increases the discretizes nodes according to relevance.