Managing documents in question answering systems

1. A method for answering a time-sensitive question received by a computer system, the method comprising:

identifying a first candidate answer to the time-sensitive question with a first document;

calculating a first temporal confidence subscore of the first candidate answer with a first temporal confidence rule;

generating a first general confidence score of the first candidate answer, the first general confidence score being based on at least the first temporal confidence subscore;

identifying a second candidate answer to the time-sensitive question with a second document;

calculating a second temporal confidence subscore of the second candidate answer with a second temporal confidence rule;

generating a second general confidence score of the second candidate answer, the second general confidence score being based on at least the second temporal confidence subscore;

comparing the first general confidence score and the second general confidence score; and

selecting a final answer from the first candidate answer and the second candidate answer based on the comparison of the general confidence scores.

2. The method of claim 1 , wherein the time-sensitive question has a question time frame, wherein the first document has a first base time and a first information type, the first information type being associated with the first temporal confidence rule, and wherein the second document has a second base time and a second information type, the second information type being associated with the second temporal confidence rule, the method further comprising:

the calculating the first temporal confidence subscore of the first candidate answer with the first temporal confidence rule comprising:

applying the first temporal confidence rule to the first base time and the question time frame; and

the calculating the second temporal confidence subscore of the second candidate answer with the second temporal confidence rule comprising:

applying the second temporal confidence rule to the second base time and the question time frame.

3. The method of claim 2 , wherein the first information type is further associated with a first update rule, the method further comprising:

applying the first update rule to at least the first base time;

determining, based on the application of the first update rule to at least the first base time, that the first document is outdated; and

in response to determining that the first document is outdated, ingesting an updated version of the first document, the updated version of the first document having an updated base time and the first information type.

4. The method of claim 3 , wherein the second information type is further associated with a second update rule, and wherein the first base time is more recent than the second base time, the method further comprising:

after determining that the first document is outdated, determining that the second document is not outdated by applying the second update rule to at least the second base time.

5. The method of claim 3 , wherein the at least the first base time comprises the first base time and the question time frame.

6. The method of claim 2 , wherein the first base time is more recent than the second base time, and wherein the first temporal confidence subscore indicates a lesser degree of confidence than the second temporal confidence subscore.

7. The method of claim 2 , further comprising:

ingesting a plurality of source documents, the plurality of source documents including the first document and the second document; and

labeling each of the plurality of source documents with a base time and one of a plurality of information types, the plurality of information types including the first information type and the second information type, each of the plurality of information types being associated with one of a plurality of temporal confidence rules, the plurality of temporal confidence rules including the first temporal confidence rule and the second temporal confidence rule.

8. The method of claim 7 , wherein the plurality of source documents comprises at least one thousand source documents, and wherein the plurality of information types comprises at least ten information types.

9. The method of claim 2 , wherein the first information type and second information type are different.

10. The method of claim 2 , wherein the at least the first temporal confidence subscore comprises the first temporal confidence subscore and a third temporal confidence subscore, the method further comprising:

supporting the first candidate answer with a third document, the third document having a third base time and a third information type, the third information type being associated with a third temporal confidence rule; and

calculating the third temporal confidence subscore by applying the third temporal confidence rule to the third base time and the question time frame.

11. The method of claim 2 , wherein the at least the first temporal confidence subscore comprises the first temporal confidence subscore and a third temporal confidence subscore, the method further comprising:

identifying a third document as relevant to the first candidate answer, the third document having a third base time and a third information type, the third information type being associated with a third temporal confidence rule and a third update rule;

subsequent to identifying the third document as relevant to the first candidate answer, applying the third update rule to at least the third base time to determine that the third document is outdated;

in response to determining that the third document is outdated, obtaining an updated version of the third document, the updated version of the third document having a updated third base time and the third information type; and

calculating the third temporal confidence subscore by applying the third temporal confidence rule to the updated third base time and the question time frame.