Request flow log retrieval

1. A method, comprising:

extracting, with a computer, at least one keyword from a natural language description of an action which a user requests that a resource-provisioning system perform during a user session;

segmenting a user request requesting the action into a plurality of segments that include:

a user segment that provides a user indicator;

a timing segment that identifies a timeframe during which performing the action the user requested was attempted; and

an action segment from which the at least one keyword is extracted;

training a classifier model to classify actions performed by the resource provisioning system using training data comprising a labeled search query as an input, wherein the classifier model is a multilayer perceptron or a neural network;

determining a classification of the action based on a correlation value generated by the classifier model based on the at least one keyword;

identifying a request flow associated with the action based on the classification of the action;

rank ordering a plurality of system log entries identified as corresponding to the request flow by applying a ranking model trained using machine learning to each of the plurality of system log entries, the user indicator, a timing indicator provided by the timing segment, and the at least one keyword; and

returning the rank ordered plurality of system log entries.

2. The method of claim 1 , wherein the classifier model comprises a multilayer perceptron.

3. The method of claim 1 , further comprising generating a corpus of training data for training the classifier model.

4. The method of claim 3 , wherein the generating a corpus of training data comprises:

creating an input document comprising a user request and system log entries corresponding to a request flow associated with the user request;

adding to the input document at least one keyword extracted from a natural language description of the user request;

indexing the input document; and

electronically storing the input document in at least one of a search engine, file system, and database comprising the corpus of training data.

5. The method of claim 3 , further comprising adding data to the corpus of training data to refine the classifier model, wherein the adding comprises:

electronically recording a natural language description of a user request;

collecting and normalizing at least one log comprising system log entries corresponding to the user request;

creating an input document from the at least one log by adding to the at least one log at least one keyword extracted from the user request; and

adding the input document to a database comprising the corpus of training data.

6. A system, comprising:

a computer having at least one processor programmed to initiate executable operations, the executable operations comprising:

extracting at least one keyword from a natural language description of an action which a user requests that a resource-provisioning system perform during a user session;

segmenting a user request requesting the action into a plurality of segments that include:

a user segment that provides a user indicator;

a timing segment that identifies a timeframe during which performing the action the user requested was attempted; and

an action segment from which the at least one keyword is extracted;

training a classifier model to classify actions performed by the resource provisioning system using training data comprising a labeled search query as an input, wherein the classifier model is a multilayer perceptron or a neural network;

determining a classification of the action based on a correlation value generated by the classifier model based on the at least one keyword;

identifying a request flow associated with the action based on the classification of the action;

rank ordering a plurality of system log entries identified as corresponding to the request flow by applying a ranking model trained using machine learning to each of the plurality of system log entries, the user indicator, a timing indicator provided by the timing segment, and the at least one keyword; and

returning the rank ordered plurality of system log entries.

7. The system of claim 6 , wherein the classifier model comprises a multilayer perceptron.

8. The system of claim 6 , wherein the executable operations further include generating training data that is added to a corpus of training data for training the classifier model.

9. The system of claim 8 , wherein the generating training data comprises:

electronically recording a natural language description of a new user request;

collecting and normalizing at least one log corresponding to the new user request;

adding to the at least one log at least one keyword extracted from the new user request;

creating a new input document comprising the at least one log and the at least one keyword extracted from the new user request; and

adding the new input document to the corpus of training data.

10. A computer program product, comprising:

a computer-readable storage medium having program code stored thereon, the program code executable by computer hardware to initiate operations including:

extracting at least one keyword from a natural language description of an action which a user requests that a resource-provisioning system perform during a user session;

segmenting a user request requesting the action into a plurality of segments that include:

a user segment that provides a user indicator;

a timing segment that identifies a timeframe during which performing the action the user requested was attempted; and

an action segment from which the at least one keyword is extracted;

training a classifier model to classify actions performed by the resource provisioning system using training data comprising a labeled search query as an input, wherein the classifier model is a multilayer perceptron or a neural network;

determining a classification of the action based on a correlation value generated by the classifier model based on the at least one keyword;

identifying a request flow associated with the action based on the classification of the action;

rank ordering a plurality of system log entries identified as corresponding to the request flow by applying a ranking model trained using machine learning to each of the plurality of system log entries, the user indicator, a timing indicator provided by the timing segment, and the at least one keyword; and

returning the rank ordered plurality of system log entries.

11. The computer program product of claim 10 , wherein the classifier model comprises a multilayer perceptron.

12. The computer program product of claim 10 , wherein the operations further include generating a corpus of training data for training the classifier model.

13. The computer program product of claim 12 , wherein the generating a corpus of training data comprises:

creating an input document comprising a user request and system log entries corresponding to a request flow associated with the user request;

adding to the input document at least one keyword extracted from a natural language description of the user request;

indexing the input document; and

electronically storing the input document in at least one of a search engine, file system, and database comprising the corpus of training data.

14. The computer program product of claim 12 , wherein the operations further include adding data to the corpus of training data to refine the classifier model, wherein the adding comprises:

electronically recording a natural language description of a user request;

collecting and normalizing at least one log comprising system log entries corresponding to the user request;

creating an input document from the at least one log by adding to the at least one log at least one keyword extracted from the user request; and

adding the input document to a database comprising the corpus of training data.