Methods and systems for facilitating stratified elusion and recall
A computer-implemented method stratified elusion includes selecting hypothetical cutoff ranks when a stopping point is reached, calculating for each respective cutoff rank a recall value, elusion value, and remaining count; and displaying each respective cutoff rank, recall value, elusion value, and remaining count. A stratified elusion system includes a processor and a memory storing instructions that, when executed, cause the system to select cutoff ranks when a stopping point is reached, calculate for each respective cutoff rank a recall value, elusion value, and remaining count; and display each respective cutoff rank, recall value, elusion value, and remaining count. A non-transitory computer readable medium storing program instructions that when executed, cause a computer system to select cutoff ranks when a stopping point is reached, calculate for each respective cutoff rank a recall value, elusion value, and remaining count; and display each respective cutoff rank, recall value, elusion value, and remaining count.
1. A computer-implemented method of facilitating validation of an analysis of a set of documents in an active learning process, comprising:
selecting, in response to detecting a stopping point of the active learning process, an initial set of hypothetical cutoff ranks;
calculating, for each hypothetical cutoff rank, a respective recall value, a respective elusion value, and a respective number of documents left to review; and
displaying, in a display device, for each hypothetical cutoff rank, the hypothetical cutoff rank, the respective recall value, the respective elusion value, and the respective number of documents left to review.
2. The computer-implemented method of claim 1 , wherein selecting, in response to detecting the stopping point of the active learning process, the initial set of hypothetical cutoff ranks includes generating a list of ranks of unreviewed documents in the set of documents using a systematic random sampling algorithm.
3. The computer-implemented method of claim 1 , wherein calculating, for each hypothetical cutoff rank, the respective recall value includes
analyzing the number of responsive documents in the set of documents, the number of unreviewed documents in the set of documents, and a sample size; and
calculating the number of responsive documents in a sample data set, an interpolated number of sample responsive documents above a cutoff, and an estimated proportion of responsive documents above the cutoff.
4. The computer-implemented method of claim 1 , wherein calculating, for each hypothetical cutoff rank, the respective elusion value includes
analyzing an interpolated number of sample responsive documents below a cutoff, an interpolated sample size below the cutoff, a number of documents in a ranking below the cutoff, and a number of unresponsive documents in the set of documents; and
calculating an estimated number of responsive documents below the cutoff.
5. The computer-implemented method of claim 1 , wherein displaying, in the display device, for the each hypothetical cutoff rank, the hypothetical cutoff rank, the respective recall value, the respective elusion value, and the respective number of documents left to review includes generating one or both of (i) a table depicting the hypothetical cutoff rank, the respective recall value, the respective elusion value, and the respective number of documents left to review, and (ii) a chart depicting the hypothetical cutoff rank, the respective recall value, the respective elusion value, and the respective number of documents left to review.
6. The computer-implemented method of claim 1 , further comprising
calculating one or both of (i) a recall confidence, and (ii) an elusion confidence.
7. The computer-implemented method of claim 6 , wherein calculating the one or both of the recall confidence and the elusion confidence includes simulating Monte Carlo iterations to determine a richness of segment above-cutoff and a richness of segment below-cutoff.
8. A family-based review computing system of validating an active learning process, comprising
one or more processors; and
a memory storing instructions that, when executed, cause the computing system to:
select, in response to detecting a stopping point of the active learning process, an initial set of hypothetical cutoff ranks;
calculate, for each hypothetical cutoff rank, a respective recall value, a respective elusion value, and a respective number of documents left to review; and
display, in a display device, for each hypothetical cutoff rank, the hypothetical cutoff rank, the respective recall value, the respective elusion value, and the respective number of documents left to review.
9. The family-based review computing system of claim 8 , wherein the memory stores further instructions that when executed, cause the computing system to:
generate a list of ranks of unreviewed documents in the set of documents using a systematic random sampling algorithm.
10. The family-based review computing system of claim 8 , wherein the memory stores further instructions that when executed, cause the computing system to:
analyze the number of responsive documents in the set of documents, the number of unreviewed documents in the set of documents, and a sample size; and
calculate the number of responsive documents in a sample data set, an interpolated number of sample responsive documents above a cutoff, and an estimated proportion of responsive documents above the cutoff.
11. The family-based review computing system of claim 8 , wherein the memory stores further instructions that when executed, cause the computing system to:
analyze an interpolated number of sample responsive documents below a cutoff, an interpolated sample size below the cutoff, a number of documents in a ranking below the cutoff, and a number of unresponsive documents in the set of documents; and
calculate an estimated number of responsive documents below the cutoff.
12. The family-based review computing system of claim 8 , wherein the memory stores further instructions that when executed, cause the computing system to:
generate one or both of (i) a table depicting the hypothetical cutoff rank, the respective recall value, the respective elusion value, and the respective number of documents left to review, and (ii) a chart depicting the hypothetical cutoff rank, the respective recall value, the respective elusion value, and the respective number of documents left to review.
13. The family-based review computing system of claim 8 , wherein the memory stores further instructions that when executed, cause the computing system to:
calculate one or both of (i) a recall confidence, and (ii) an elusion confidence.
14. The family-based review computing system of claim 13 , wherein the memory stores further instructions that when executed, cause the computing system to:
simulate Monte Carlo iterations to determine a richness of segment above-cutoff and a richness of segment below-cutoff.
15. A non-transitory computer readable medium storing program instructions that when executed, cause a computer system to:
select, in response to detecting a stopping point of the active learning process, an initial set of hypothetical cutoff ranks;
calculate, for each hypothetical cutoff rank, a respective recall value, a respective elusion value, and a respective number of documents left to review; and
display, in a display device, for each hypothetical cutoff rank, the hypothetical cutoff rank, the respective recall value, the respective elusion value, and the respective number of documents left to review.
16. The non-transitory computer readable medium of claim 15 , including further program instructions that when executed, cause a computer system to:
generate a list of ranks of unreviewed documents in the set of documents using a systematic random sampling algorithm.
17. The non-transitory computer readable medium of claim 15 , including further program instructions that when executed, cause a computer system to:
analyze the number of responsive documents in the set of documents, the number of unreviewed documents in the set of documents, and a sample size; and
calculate the number of responsive documents in a sample data set, an interpolated number of sample responsive documents above a cutoff, and an estimated proportion of responsive documents above the cutoff.
18. The non-transitory computer readable medium of claim 15 , including further program instructions that when executed, cause a computer system to:
analyze an interpolated number of sample responsive documents below a cutoff, an interpolated sample size below the cutoff, a number of documents in a ranking below the cutoff, and a number of unresponsive documents in the set of documents; and
calculate an estimated number of responsive documents below the cutoff.
19. The non-transitory computer readable medium of claim 15 , including further program instructions that when executed, cause a computer system to:
generate one or both of (i) a table depicting the hypothetical cutoff rank, the respective recall value, the respective elusion value, and the respective number of documents left to review, and (ii) a chart depicting the hypothetical cutoff rank, the respective recall value, the respective elusion value, and the respective number of documents left to review.
20. The non-transitory computer readable medium of claim 15 , including further program instructions that when executed, cause a computer system to:
calculate one or both of (i) a recall confidence, and (ii) an elusion confidence.