Query Reformulation

1
Query Reformulation

• Query 4 Operation Barbarossa Battle of
  Stalingrad
• This a good narrowed down query as this has given
  us a list of documents which have specific
  information on both these events. The search
  results have a document
• http//www.thehistorychannel.co.uk/classroom/gcse/
  staling.htm
• which has specific information found in the
  abstract. In fact there are a couple of lines
  which match details specified in the abstract. It
  looks as if this document might have been used
  while making this abstract. This particular web
  page has a lot of pertinent information and was
  the second document in the list of retrieved
  documents. So this query has given us a lot of
  specific details of the events in the abstract.

2
Query Reformulation
3
Query Reformulation

• Query 5 Operation Barbarossa Battle of
  Stalingrad Russian Victory
• This query was used to combine three different
  facts together in the same query so that we could
  have a bunch of documents highlighting these
  important events. Since all these three put
  together form a major segment of the abstract the
  objective was to get documents which would have
  more information related to the event. This query
  has given interesting results. In fact the same
  web page http//www.thehistorychannel.co.uk/classr
  oom/gcse/staling.htm also appears in the search
  results but this time at a lower rank which does
  not seem to help in narrowing down the results to
  match most of the information in the abstract.

4
Query Reformulation
5
Query Reformulation

• But what is interesting is that on rank 14 is a
  document http//www.alief.isd.tenet.edu/instructio
  n/SocialStudies/stalingr.htm which seems to have
  a lot of relevant information found in the
  abstract. So this query has given us results
  which have a lot of pertinent information but the
  results seem to be scattered.

6
Query Reformulation

• Query 6 Battle of Stalingrad Southern Steppes
• Interesting query We combine the general
  information which is the main topic of the
  abstract with specific information like Southern
  Steppes which is a region involved with the war.
  This query has the following document as the
  first result
• http//campus.northpark.edu/history//WebChron/East
  Europe/Stalingrad.html
• This page has a lot of relevant information found
  in the abstract, so this combination of detailed
  information Battle of Stalingrad a specific
  detail Southern Steppes has given useful results.

7
Query Reformulation

• Query 7 Battle of Stalingrad Southern Steppes
  Caucasus region. Going one step further and
  adding the second relevant geographical region
  Caucasus region does not change the results much.
  The first document is still http//campus.northpar
  k.edu/history//WebChron/EastEurope/Stalingrad.html
  
• Query8 Battle of Stalingrad Franklin Roosevelt
• This query combines the main event with an
  individual associated with this event. The result
  of this query is that most of the retrieved
  documents are concerned with Franklin Roosevelt
  as the individual has more significance rather
  than the event in the relevant context. Hence
  this combination has not resulted in any fruitful
  results.

8
Query Reformulation

• Query 9 Battle of Stalingrad most important
  battles of WWII Again this particular query is
  similar to the exact description of the event
  found in the abstract. This has resulted in the
  web page http//www.jumboshrimp.simplenet.com/WW2/
  stalingrad.htmla as the top result of the query.

9
Query Reformulation
10
Query Reformulation

• Observations
• After looking at the abstract and studying it
  carefully, we observed that the parts of the
  abstract that help us in distinguishing the
  search results are the unique key words. In this
  particular example the key words that are
  relevant are
• Operation Barbaross
• Battle of Stalingrad
• Southern Steppes and/or Caucasus region
• World War II Soviet victory

11
Query Reformulation

• Combining these correctly is critical in
  achieving the correct outcomes for the queries.
  Hence in query formulation it is very important
  to look for such key words and combine them
  appropriately.
• We observed that combining a more general
  attribute like Battle of Stalingrad and a
  specific quality like Operation Barbarossa
  results in narrowing down the results to the
  desired set of documents and even in doing this
  it is difficult to get more than a couple of
  documents which have the exact information
  available in the extract. Hence combining these
  attributes is very important for correct
  retrieval of documents.

12
Query Reformulation

• Also the documents retrieved tend to be those
  which have the key words mentioned above. It is
  very difficult to formulate queries which would
  retrieve the other documents as they dont have a
  distinguishing attribute. An abstract example
  like the one chosen which is a general
  description of a particular event thus tends to
  result in a wide variety of search outcomes which
  gives general information but getting detailed
  information is difficult.
• In general all kinds of words are useful for
  improving the retrieval performance but nouns
  contribute the most, adjectives and adverbs less
  and the verbs least.

13
Query Reformulation

• Results Of A Study
• A study shows that most users enter only a single
  query.
• A third of users go beyond the single query, with
  a smaller group using either query modification
  or relevance feedback, or viewing more than the
  first page of results.
• The study shows that the distribution of query
  type shifts as the length of session increases.
  For the user sessions of two and three queries,
  the relevance feedback query is dominant. As the
  length of the queries increase, the occurrences
  of relevance feedback as a percentage of all
  query types decreases.

14
Query Reformulation

• Given the low occurrences of relevance feedback
  queries, the study attempted to determine if the
  user sessions containing relevance feedback were
  successful or not. Given relevance feedback the
  benefit of the doubt, 63 of the relevance
  feedback sessions could be construed as being
  successful. If the partially successful user
  sessions are included, then almost 80 of the
  relevance feedback session provide some measure
  of success.
• The study suggests that relevance feedback is
  successful for Web users, although only a small
  percentage of Web users take advantage of this
  feature.

15
Query Reformulation

• How many phrases or terms should be added depends
  on the queries and associated phrases. For some
  queries many phrases can be added whereas others
  may degrade with only a few additional phrases.