Updated replication doc for msmarco-passage and msmarco-doc

[adamyy]

Environment

OS: macOS Catalina 10.15.4
Java: openjdk 11.0.7 2020-04-14
Python: 3.7.5

Msmarco Doc

• With default BM25 parameters (k1=0.9, b=0.4)

• With tuned BM25 parameters (k1=3.44, b=0.87)

Msmarco Passage

• With BM25 parameters k1=0.9, b=0.4

• With BM25 parameters k1=0.82, b=0.68

Issues

Same issue with pyserini as mentioned in #1220. Had to fallback to the Java implementation as well.

[lintool]

I think I fixed the issue: #1238

Can you merge, fix conflict, and re-check?

[codecov]

Codecov Report

Merging #1237 into master will decrease coverage by 0.03%.
The diff coverage is n/a.

@@             Coverage Diff              @@
##             master    #1237      +/-   ##
============================================
- Coverage     48.37%   48.33%   -0.04%     
+ Complexity      741      739       -2     
============================================
  Files           147      147              
  Lines          8559     8559              
  Branches       1217     1217              
============================================
- Hits           4140     4137       -3     
- Misses         4078     4082       +4     
+ Partials        341      340       -1     

Impacted Files	Coverage Δ	Complexity Δ
...java/io/anserini/ltr/feature/CountBigramPairs.java	89.61% <0.00%> (-3.90%)	33.00% <0.00%> (-2.00%)

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 94893f1...af7320d. Read the comment docs.

[adamyy]

Re-ran msmarco-passage with the updated Python msmarco retrieve command, seems to work just fine this time. msmarco-doc does not seem to be affected by #1238 so I did not re-run it.