Adaptive semantic reasoning engine

• US 20070124263A1
• Filed: 11/30/2005
• Published: 05/31/2007
• Est. Priority Date: 11/30/2005
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Liu, Jun, Katariya, Sanjeev, Yao, Qi, Gao, Jianfeng, Ramsey, William
• Granted Patent: US 7,822,699 B2
• Time in Patent Office: Days
• US Class Current: 706/14
• CPC Class Codes: G06F 16/3334 Selection or weighting of terms from queries, including natural language queries

Abstract

Provided is an adaptive semantic reasoning engine that receives a natural language query, which may contain one or more contexts. The query can be broken down into tokens or a set of tokens. A task search can be performed on the token or token set(s) to classify a particular query and/or context and retrieve one or more tasks. The token or token set(s) can be mapped into slots to retrieve one or more task result. A slot filling goodness may be determined that can include scoring each task search result and/or ranking the results in a different order than the order in which the tasks were retrieved. The token or token set(s), retrieved tasks, slot filling goodness, natural language query, context, search result score and/or result ranking can be feedback to the reasoning engine for further processing and/or machine learning.

Claims

1. A system that facilitates intent to action mapping, comprising:

an interface component that receives a natural language input and a context;

a reasoning engine that processes the received input and maps the input to an action based in part on the natural language input and the context; and

an output component that outputs the mapped action, the mapped action is fed back to the reasoning engine.

2. The system of claim 1, the reasoning engine comprising:

a lexical processing component that breaks the input into at least one token set; and

a task search component that performs a task search on the least one token set to retrieve a plurality of task lists.

3. The system of claim 2, the task search component reduces the plurality of task lists into a manageable number of task lists.

4. The system of claim 2, the reasoning engine further comprising a slot filling component that fills in slots with the task list information.

5. The system of claim 4, the slot filling component utilizes ranking algorithms to enable slot matching and filling.

6. The system of claim 1, the reasoning engine further comprising a ranking component that scores a task score or ranks a task list information.

7. The system of claim 1, the reasoning engine utilizes one of Hidden Markov Model (HMM), Max Entropy Min Divergence (MEMD) and IR technology.

8. The system of claim 1, further comprising a feedback and training flow to train the reasoning engine for mapping an action.

9. A method for reasoning an action based on a natural language input, comprising:

receiving a query;

assigning tokens associated with the query into slots;

determining the appropriateness of the assigned slots; and

outputting the assigned slot result.

10. The method of claim 9, after receiving a query, further comprising dividing the query into at least two tokens;

retrieving a task list; and

reducing the task list.

11. The method of claim 9, determining the appropriateness of the assigned slots, further comprising:

retrieving a task search result from the assigned tokens; and

scoring the task search result.

12. The method of claim 9, determining the appropriateness of the assigned slots, further comprising ascertaining the appropriateness of an original ranking of the search result.

13. The method of claim 12, further comprising ranking the search result differently if the original ranking is not appropriate.

14. The method of claim 9, further comprising receiving a context associated with a user action after the result is output.

15. The method of claim 9, further comprising feeding back the user action during a next query receipt.

16. The method of claim 9, further comprising utilizing one of Hidden Markov Model, Naï

ve-Bayesian model;

Max Entropy Min Divergence and IR technology to construct an adaptive reasoning engine.

17. A system that converts a natural language input into an executed action, comprising:

means for receiving a natural language inquiry and a context associated with the inquiry;

means for filling slots;

means for ascertaining an overall ranking of the filled slots;

means for determining an asset mapping; and

means for executing the action.

18. The system of claim 17, further comprising means for partitioning the inquiry into at least one token.

19. The system of claim 17, further comprising:

means for acquiring a user feedback; and

means for learning from the user feedback.

20. The system of claim 17, further comprising:

means for utilizing feedback technologies to construct an accurate and adaptive reasoning engine.