Information extraction from daily drilling reports using machine learning

1. A method for extracting information regarding a drill site, the method comprising:

forming one or more documents having raw data regarding a well site, wherein the raw data includes one or more raw comments directed to operational details of the well site;

extracting the raw data from the one or more documents regarding the well site to produce extracted raw data;

pre-processing the extracted raw data by removing one or more of ambiguity, artifacts, and formatting errors from the one or more raw comments to produce pre-processed data;

extracting topics data from the pre-processed data using a natural language processing (NLP) algorithm to produce extracted topics data, the NLP algorithm including a first NLP model for extracting the topics data;

extracting measurement data from the pre-processed data using the NLP algorithm to produce extracted measurement data, the NLP algorithm including a second NLP model for extracting the measurement data;

aggregating the extracted topics data and the extracted measurement data to form a set of discrete data points per comment to produce aggregated data;

identifying one or more discrete data points from the aggregated data; and

applying the one or more discrete data points as calibration points for geomechanical post drill analysis used to generate drilling risks and maps to improve well design or mitigation strategies in a field or area.

2. The method of claim 1 , wherein the one or more documents are daily drilling reports and the set of discrete data points is a set of calibration points.

3. The method of claim 1 , wherein the pre-processing step comprises iterating through the one or more raw comments and converting the one or more raw comments into cleaned comment tokens.

4. The method of claim 3 , wherein the pre-processing step comprises iterating through the cleaned comment tokens, and further wherein the NLP algorithm normalizes the cleaned comment tokens based on pre-defined domain knowledge inputs.

5. The method of claim 4 , wherein the pre-processing step comprises reducing or transforming the normalized cleaned comment tokens based on the pre-defined domain knowledge inputs and/or using a semantic analysis.

6. The method of claim 1 , wherein the first NLP model comprises two or more NLP models for extracting the topics data.

7. The method of claim 6 , wherein the second NLP model comprises two or more NLP models for extracting the measurement data.

8. The method of claim 1 , further comprising retraining the first NLP model or the second NLP model with user feedback.

9. A system comprising:

a processor; and

one or more documents having raw data regarding a well site, the raw data having one or more raw comments directed to operational details of the well site, wherein the processor is configured to:

extract the raw data from the one or more documents regarding the well site to produce extracted raw data;

pre-process the extracted raw data by removing one or more of ambiguity, artifacts, and formatting errors from the one or more raw comments to produce pre-processed data;

extract topics data from the pre-processed data using a natural language processing (NLP) algorithm to produce extracted topics data, the NLP algorithm including a first NLP model for extracting the topics data;

extract measurement data from the pre-processed data using the NLP algorithm to produce extracted measurement data, the NLP algorithm including a second NLP model for extracting the measurement data;

aggregate the extracted topics data and the extracted measurement data to form a set of discrete data points per comment to produce aggregated data;

identify one or more discrete data points from the aggregated data; and

apply the results of the one or more discrete data points as calibration points for geomechanical post drill analysis used to generate drilling risks and maps to improve well design or mitigation strategies in a field or area.

10. The system of claim 9 , wherein the one or more documents are daily drilling reports and the set of discrete data points is a set of calibration points.

11. The system of claim 9 , wherein the processor is configured to iterate through the one or more raw comments and convert the one or more raw comments into cleaned comment tokens.

12. The system of claim 11 , wherein the processor is configured to iterate through the cleaned comment tokens and further wherein the NLP algorithm normalizes the cleaned comment tokens based on pre-defined domain knowledge inputs.

13. The system of claim 12 , wherein the processor is configured to reduce or transform the normalized cleaned comment tokens based on the pre-defined domain knowledge inputs and/or using a semantic analysis.

14. The system of claim 9 , wherein the first NLP model comprises two or more NLP models for extracting the topics data.

15. The system of claim 14 , wherein the second NLP model comprises two or more NLP models for extracting the measurement data.

16. The system of claim 9 , further comprising retraining the first NLP model or the second NLP model with user feedback.

17. A method for extracting information from a daily drilling report (DDR), the method comprising:

extracting raw data from the DDR regarding a well site, wherein the raw data includes one or more raw comments directed to operational details of the well site;

iterating through the one or more raw comments and converting the one or more raw comments into cleaned comment tokens;

iterating through the cleaned comment tokens and normalizing the cleaned comment tokens;

reducing or transforming the normalized cleaned comment tokens to form pre- processed data;

extracting topics data from the pre-processed data using a natural language processing (NLP) algorithm to produce extracted topics data, the NLP algorithm including a first NLP model for extracting the topics data;

extracting measurement data from the pre-processed data using the NLP algorithm to produce extracted measurement data, the NLP algorithm including a second NLP model for extracting the measurement data;

aggregating the extracted topics data and the extracted measurement data to form a set of discrete data points per comment to produce aggregated data, the set of discrete data points including a set of calibration points;

identifying one or more calibration points from the aggregated data; and

applying the results of the one or more calibration points as calibration points for geomechanical post drill analysis used to generate drilling risks and maps to improve well design or mitigation strategies in a field or area.

18. The method of claim 17 , wherein the first NLP model comprises two or more NLP models for extracting the topics data.

19. The method of claim 18 , wherein the second NLP model comprises two or more NLP models for extracting the measurement data.

20. The method of claim 17 , further comprising retraining the first NLP model or the second NLP model with user feedback.