System and method for semantic processing of natural language commands
A system, method and computer-readable storage devices are for processing natural language commands, such as commands to a robotic arm, using a Tag & Parse approach to semantic parsing. The system first assigns semantic tags to each word in a sentence and then parses the tag sequence into a semantic tree. The system can use statistical approach for tagging, parsing, and reference resolution. Each stage can produce multiple hypotheses, which are re-ranked using spatial validation. Then the system selects a most likely hypothesis after spatial validation, and generates or outputs a command. In the case of a robotic arm, the command is output in Robot Control Language (RCL).
1. A method comprising:
assigning, via a sequence tagger, a part of speech, a semantic tag and a label to each work in a natural language command addressed to a robot to yield a tagged natural language command; and
semantically parsing, via a processor and a semantic parsor, the tagged natural language command to yield a parsed natural language command, wherein a data set used to train the semantic parsor does not include any tag from which the sequence tagger selects for assigning the semantic tag,
receiving the parsed natural language command for execution by the robot; and
moving the robot according to a spatial validation of a physical context of the parsed natural language command when applied to the robot.
2. The method of claim 1 , wherein the robot comprises a robotic arm.
3. The method of claim 1 , further comprising:
identifying a command type for the parsed natural language command and wherein moving the robot is according to the command type.
4. The method of claim 3 , further comprising, after identifying the command type:
performing the spatial validation, based on the command type, for the physical context of the natural language command when applied to the robot; and
if the spatial validation indicates that a certainty of the command type meets a certainty threshold, executing the natural language command; and
if the spatial validation indicates that the certainty of the command type does not meet the certainty threshold, prompting for clarification of the natural language command.
5. The method of claim 4 , wherein the spatial validation is based on a condition of a target area associated with the parsed natural language command.
6. The method of claim 5 , wherein the target area comprises a working area of the robot, and wherein the condition of the target area comprises presence and positions of objects in the working area.
7. The method of claim 6 , further comprising:
performing reference resolution on the parsed natural language command to uniquely identify one of the objects in the working area.
8. The method of claim 3 , wherein the parsed natural language command is directed to one of a physical robotic appendage or a virtual robotic appendage.
9. The method of claim 8 , further comprising outputting the parsed natural language command according to the command type and according to a robot control language.
10. The method of claim 1 , wherein the semantic tag identifies entity types and event types in the natural language command.
11. The method of claim 1 , wherein the semantic tag is assigned using a maximum entropy sequence tagger.
12. The method of claim 1 , wherein a semantic parser trained on a non-lexical semantic tree parses the tagged natural language command.
13. A system comprising:
a processor; and
a computer-readable storage medium having instructions stored which, when executed by the processor, cause the processor to perform operations comprising:
receiving a parsed natural language command for execution by a robot; and
moving the robot according to a spatial validation of a physical context of the parsed natural language command when applied to the robot, wherein the parsed natural language command was generated by operations comprising:
assigning, via a sequence tagger, a part of speech, a semantic tag and a label to each word in a natural language command addressed to the robot to yield a tagged natural language command; and
semantically parsing, via a processor and a semantic parsor, the tagged natural language command to yield the parsed natural language command, wherein a data set used to train the semantic parsor does not include any tag from which the sequence tagger selects for assigning the semantic tag.
14. The system of claim 13 , wherein the robot comprises a robotic arm.
15. The system of claim 13 , wherein the computer-readable storage medium stores additional instructions stored which, when executed by the processor, cause the processor to perform operations further comprising:
identifying a command type for the parsed natural language command and wherein moving the robot is according to the command type.
16. The system of claim 15 , wherein the computer-readable storage medium stores additional instructions stored which, when executed by the processor, cause the processor to perform operations further comprising:
after identifying the command type:
performing the spatial validation, based on the command type, for the physical context of the natural language command when applied to the robot; and
if the spatial validation indicates that a certainty of the command type meets a certainty threshold, executing the natural language command; and
if the spatial validation indicates that the certainty of the command type does not meet the certainty threshold, prompting for clarification of the natural language command.
17. The system of claim 13 , wherein the semantic tag identifies entity types and event types in the natural language command.
18. The system of claim 13 , wherein the spatial validation is based on a condition of a target area associated with the natural language command.
19. A non-transitory computer-readable storage device having instructions stored which, when executed by a computing device, cause the computing device to perform operations comprising:
receiving a parsed natural language command for execution by a robot; and
moving the robot according to a spatial validation of a physical context of the parsed natural language command when applied to the robot, wherein the parsed natural language command was generated by operations comprising:
assigning, via a sequence tagger, a part of speech, a semantic tag and a label to each word in a natural language command addressed to the robot to yield a tagged natural language command; and
semantically parsing, via a processor and a semantic parsor, the tagged natural language command to yield the parsed natural language command, wherein a data set used to train the semantic parsor does not include any tag from which the sequence tagger selects for assigning the semantic tag.