Add documentation for MS Marco Passage Ranking using ElasticSearch

* update_elastic_msmarco_passage_file

* integrate msmarco-passage instruction into elastirini.md

* Delete elastirini-msmarco-passage.md

* Minor updates

docs/elastirini.md

@@ -56,7 +56,7 @@ If at some point one of the ELK components is failing for some reason, or if you
 
 Once we have a local instance of Elasticsearch up and running, we can index using Elasticsearch through Elastirini.
 
-First, let us create the index in Elasticsearch.
+First, let us create the index in Elasticsearch. We need to update <index_name> and BM25 parameters for our own purpose.
 
 ```
 curl --user elastic:changeme -XPUT -H 'Content-Type: application/json' 'localhost:9200/<index_name>' \
@@ -139,3 +139,39 @@ Evaluation can be performed using `trec_eval`:
 ```
 eval/trec_eval.9.0.4/trec_eval -m map -m P.30 src/main/resources/topics-and-qrels/qrels.robust2004.txt run.es.robust04.bm25.topics.robust04.301-450.601-700.txt
 ```
+# Elasticsearch on MSMARCO(Passage)
+For Msmarco-passage data preparation, check Anserini: [BM25 Baselines on MS MARCO (Passage)](https://github.com/castorini/anserini/blob/master/docs/experiments-msmarco-passage.md). Similarly, there are three steps:
+
+1.Create the index in Elasticsearch by calling `curl` command as above. Remember to update BM25 parameters with k1 = 0.82, b = 0.68.
+
+2.Index documents as `JsonCollection` through Elastirini:
+```
+sh target/appassembler/bin/IndexCollection -collection JsonCollection -generator JsoupGenerator 
+-es -es.index msmarco-passage -threads 9 -input msmarco-passage/collection_jsonl  -storePositions -storeDocvectors -storeRawDocs
+```
+3.Retrieving and Evaluating the dev set
+Since there are many queries (> 100k), it would take a long time to retrieve all of them. To speed this up, we use only the queries that are in the qrels file:
+```
+python ./src/main/python/msmarco/filter_queries.py --qrels msmarco-passage/qrels.dev.small.tsv \
+ --queries msmarco-passage/queries.dev.tsv --output_queries msmarco-passage/queries.dev.small.tsv
+```
+The output queries file should contain 6980 lines.
+
+We can now retrieve this smaller set of queries with Elastirini, it takes about half hour on a modern desktop with an SSD:
+```
+sh target/appassembler/bin/SearchElastic -topicreader TsvString -es.index msmarco-passage \
+   -topics msmarco-passage/queries.dev.small.tsv \
+   -output msmarco-passage/run.dev.small.tsv
+```
+There are also other -es parameters that you can specify as you see fit.
+To perform the evaulation with trec_eval, run:
+```
+./eval/trec_eval.9.0.4/trec_eval -c -mrecall.1000 -mmap \
+    msmarco-passage/qrels.dev.small.tsv msmarco-passage/run.dev.small.tsv 
+```
+The output should be:
+```
+map                   all       0.1956
+recall_1000           all       0.8573
+```
+Average precision and recall@1000 are the two metrics we care about the most. You can check the table in [BM25 Baselines on MS MARCO (Passage)](https://github.com/castorini/anserini/blob/master/docs/experiments-msmarco-passage.md) `BM25  Tuning` section for more information.