Utilizing unsumbitted user input data for improved task performance

1. A computer implemented method comprising:

identifying terms of unsubmitted input data entered by a user during composition of submitted input data for the performance of a task;

determining, for an identified unsubmitted term, a relevance score indicating the relevance of the unsubmitted term to user intent, wherein the relevance score of the unsubmitted term is determined using a graphical semantic model based on the submitted input data as a representation of user intent, and

selecting, based on the determined relevance score, the identified unsubmitted term, for use in supplementing the submitted input data for the performance of the task.

2. The method of claim 1 , wherein determining the relevance score comprises:

determining a value of one or more of a set of relevance features for the unsubmitted term, wherein each relevance feature is a parameter associated with a representation of the unsubmitted term in the graphical semantic model, and

determining the relevance score based on the determined one or more relevance feature values.

3. The method of claim 2 , wherein the relevance score is determined as a function of the determined one or more relevance feature values, wherein the function defines one or more of thresholds and weights for each relevance feature value.

4. The method of claim 2 , wherein the relevance score is determined using a predictive model of relevance to user intent based on the set of relevance features.

5. The method of claim 1 , further comprising:

obtaining a predefined semantic network graph;

performing spreading activation of the semantic network graph, based on terms of the submitted user data, to generate an activated sub-graph, and

utilizing the activated sub-graph as the graphical semantic model.

6. The method of claim 5 , wherein the relevance score is determined based on one or more values of a set of relevance features comprising parameters associated with at least one node representing the unsubmitted term in the activated sub-graph.

7. The method of claim 6 , wherein the relevance features include absence of a node representing the unsubmitted term in the activated sub-graph, wherein the relevance score is determined to be least relevant if a node representing the unsubmitted term is absent from the activated sub-graph.

8. The method of claim 6 , wherein the relevance features comprise one or more parameters selected from the group consisting of:

absence of a node representing the term in the activated sub-graph;

distance of a node representing the term from focus node in the activated sub-graph;

raw or activated weight of a node representing the term in the activated sub-graph;

number of in, out or total edges connected to a node representing the term in the activated sub-graph;

variations in the types of edges and nodes that are in proximity to a node representing the term in the activated sub-graph;

number of nodes representing the term in the activated sub-graph, and

other features or metrics derived from any of the above.

9. The method of claim 1 , wherein identifying terms of unsubmitted input data entered by a user during composition of submitted input data comprises:

recording input data entered by the user during composition of the submitted input data;

in response to submission of the input data by the user, identifying, from the recorded input data, the unsubmitted input data, and

deriving a set of unsubmitted terms from the identified unsubmitted input data, wherein the set of unsubmitted terms are potentially relevant to user intent.

10. The method of claim 1 , wherein selecting the identified unsubmitted term, for use in supplementing the submitted input data for the performance of the task, comprises:

selecting the unsubmitted term if the determined relevance score exceeds a predefined threshold.

11. A device comprising:

a processor, data storage and a user interface, wherein the user interface is configured for a user to enter input data for the performance of a task, wherein the processor is configured to:

identify terms of unsubmitted input data entered by a user during composition of submitted input data for the performance of a task;

determine, for an identified unsubmitted term, a relevance score indicating the relevance of the unsubmitted term to user intent, wherein the relevance score of the unsubmitted term is determined using a graphical semantic model based on the submitted input data as a representation of user intent, and

select, based on the determined relevance score, the identified unsubmitted term, for use in supplementing the submitted input data for the performance of the task.

12. The device of claim 11 , wherein the processor is configured to determine the relevance score by:

determining a value of one or more of a set of relevance features for the unsubmitted term, wherein each relevance feature is a parameter associated with a representation of the unsubmitted term in the graphical semantic model, and

determining the relevance score based on the determined one or more relevance feature values.

13. The device of claim 12 , wherein the processor is configured to determine the relevance score using a method selected from the group consisting of:

calculating the relevance score as a function of the determined one or more relevance feature values, wherein the function defines one or more of thresholds and weights for each relevance feature value, and

determining the relevance score using a predictive model of relevance to user intent based on the set of relevance features.

14. The device of claim 11 , wherein the processor is further configured to:

obtaining a predefined semantic network graph;

perform spreading activation of the semantic network graph, based on terms of the submitted user data, to generate an activated sub-graph, and

utilize the activated sub-graph as the graphical semantic model.

15. The device of claim 14 , wherein the processor is configured to determine the relevance score based on one or more values of a set of relevance features comprising parameters associated with at least one node representing the unsubmitted term in the activated sub-graph.

16. The device of claim 15 , wherein the relevance features include absence of a node representing the unsubmitted term in the activated sub-graph, wherein the processor is configured to determine the relevance score to be least relevant if a node representing the unsubmitted term is absent from the activated sub-graph.

17. The device of claim 15 , wherein the relevance features comprise one or more parameters selected from the group consisting of:

absence of a node representing the term in the activated sub-graph;

distance of a node representing the term from focus node in the activated sub-graph;

raw or activated weight of a node representing the term in the activated sub-graph;

number of in, out or total edges connected to a node representing the term in the activated sub-graph;

variations in the types of edges and nodes that are in proximity to a node representing the term in the activated sub-graph;

number of nodes representing the term in the activated sub-graph, and

other features or metrics derived from any of the above.

18. The device of claim 11 , wherein the processor is configured to identify terms of unsubmitted input data entered by a user during composition of submitted input data by:

recording input data entered by the user during composition of the submitted input data;

in response to submission of the input data by the user, identifying, from the recorded input data, the unsubmitted input data, and

deriving a set of unsubmitted terms from the identified unsubmitted input data, wherein the set of unsubmitted terms are potentially relevant to user intent.

19. The device of claim 11 , wherein the processor is configured to select the identified unsubmitted term, for use in supplementing the submitted input data for the performance of the task, by:

selecting the unsubmitted term if the determined relevance score exceeds a predefined threshold.

20. A computer program product comprising a computer readable storage medium having program instructions embodied therewith, wherein the program instructions are executable by a processor to cause the processor to:

identify terms of unsubmitted input data entered by a user during composition of submitted input data for the performance of a task;

determine, for an identified unsubmitted term, a relevance score indicating the relevance of the unsubmitted term to user intent, wherein the relevance score of the unsubmitted term is determined using a graphical semantic model based on the submitted input data as a representation of user intent, and

select, based on the determined relevance score, the identified unsubmitted term, for use in supplementing the submitted input data for the performance of the task.