GLIS 691 Lecture Notes - Lecture 6: Relevance Feedback, Information Retrieval, Aboutness
18 Jun 2018
School
Department
Course
Professor
Lecture 5: Techniques 2: February 16th 2018
** Lecture 5 notes in Week 6 because of pushed back scheduling due to previously
cancelled class.
Finding information:
oThree phases;
Question
Retrieve answer
Evaluate answer
oIt’s an ITERATIVE process
Information Retrieval Cycle:
o*query reformulation, vocabulary learning, and relevance feedback are all
important parts of this cycle.
oSteps:
Source Selection Query Formulation (Resource) Query
(search) Selected (Ranked list) Examination (Documents)
Delivery (Documents)
Goal of Information Retrieval:
oTo retrieve ALL and ONLY the relevant documents in a collection for a
particular user with a particular need for information.
oA “GOOD” IR system is able to
Extract meaningful (relevant) information while withholding
NON-RELEVANT information
Evaluation:
oWhat can be measured that reflects the users’ ability to use the system?
Coverage of information
Form of presentation
Effort required/ Ease of use
Efficiency
Recall
Precision
Recall & Precision:
oRecall= # of relevant documents retrieved / number of relevant documents
in the database
oPrecision= number of relevant documents retrieved / number of
documents retrieved total
Results;
oHow well do the results answer the question?
Complete? Partial?
Background information?
Does it give hints for further exploration?
oHow relevant are the results to the user??
find more resources at oneclass.com
find more resources at oneclass.com
Document Summary
** lecture 5 notes in week 6 because of pushed back scheduling due to previously cancelled class. Evaluate answer: it"s an iterative process. Information retrieval cycle: *query reformulation, vocabulary learning, and relevance feedback are all important parts of this cycle, steps: Source selection query formulation (resource) query (search) selected (ranked list) examination (documents) . Goal of information retrieval: to retrieve all and only the relevant documents in a collection for a particular user with a particular need for information, a good ir system is able to. Recall & precision: recall= # of relevant documents retrieved / number of relevant documents in the database, precision= number of relevant documents retrieved / number of documents retrieved total. Proximity operators: distance and order between terms, proquest: Term a n/3 term b: when you shorten near to n, you must provide a number (ex, internet n/3 media)
