-
We should be able to take the full text of a ResultSet and generate a List where each Passage is maybe 4 sentences long has at least one of the words from the question.
-
Something is still wrong with the subtitle regex.
The problem can be observed in the second example of our subtitle demo at http://demos.flowplayer.org/basics/subtitle.html
I believed that the problem…
-
Improve the passage retrieval task by coming up with new ways of estimating the relevant window of text.
-
-
Deadline: 5pm
-
Implement a passage retrieval using sentences as the passages and adding the previous and proceeding sentences to add relevance. Use the sentences as indexes for Lucene.
-
-
Current passage extraction is complex, but has arbitrary weights assigned to various features which score a passage
I intend to investigate passage retrieval starting from the basic keyword frequency…
-
Could you check your code and passage extractor result to find which phase need improvement?
-
Discovered when I tried to benchmark new passage retrieval code.
Wanted to pass gold standard document IDs. But, couldn't get it to run.