INFORMATION RETRIEVAL

(BAI515B)

MODULE WISE ANSWERS

1. Define Information Retrieval and Explain Its Components (5 Marks)

Definition:
Information Retrieval (IR) is the process of finding relevant unstructured materials, such as
text documents, that satisfy a user's information need from large collections, typically stored
on computers.
Components of IR:
1. Text Operations: Transforming raw text into indexable tokens using processes like
stopword removal and stemming.
2. Indexing: Creating data structures (e.g., inverted index) to map terms to documents
for efficient retrieval.
3. Searching: Retrieving documents containing the query tokens from the index.
4. Ranking: Scoring retrieved documents based on relevance to the query using metrics
like TF-IDF or probabilistic methods.
5. Query Operations: Enhancing queries via expansion or transformation for better
retrieval.
6. User Interface: Facilitating user interaction with the system, including query input
and displaying ranked results.

2. Describe the IR Problem with Suitable Examples (10 Marks)

Definition:
The Information Retrieval (IR) problem involves retrieving all relevant documents for a
query while minimizing the retrieval of irrelevant documents. It requires understanding user
needs and providing precise and useful results.

Key Problems in IR:

1. Document and Query Indexing:
o Challenge: Representing document and query contents effectively for matching.
o Example: Using term frequency or inverted index to link documents with
keywords.
2. Query Evaluation:
 o Challenge: Determining the degree to which a document corresponds to a query.
o Example: Matching a user query like “Federal Government financing
AMTRAK” to documents discussing funding policies.
3. System Evaluation:
o Challenge: Assessing system effectiveness in terms of relevance and
completeness.
o Example: Using Precision (fraction of relevant documents retrieved) and Recall
(fraction of all relevant documents retrieved) to evaluate search results.

Three Big Issues in IR:

1. Relevance:
o Central to IR; a relevant document meets the user's information need.
o Factors influencing relevance include query-document similarity and ranking
models.
o Example: A query for “cats” could retrieve pet care articles or information about
the musical "Cats".
2. Evaluation:
o Metrics used:
 Precision = Relevant documents retrieved ÷ Total retrieved documents.
 Recall = Relevant documents retrieved ÷ Total relevant documents.
o Example: Testing IR systems with a known set of documents and queries.
3. User Needs:
o Users’ poor query formulations can impact retrieval quality.
o Techniques like query suggestions, query expansion, and relevance feedback
are used to refine queries.
o Example: Refining a query like "cheap smartphones" with additional keywords
like "under $500".

Example:
For the query “role of the Federal Government in financing AMTRAK”, the IR system
must:
 Translate the complex query into keywords like "Federal Government," "financing,"
and "AMTRAK."
  Retrieve documents discussing funding policies, not irrelevant content like unrelated
news.


3. Explain the key elements of designing search interfaces. (5 Marks)

Key Elements of Designing Search Interfaces
Designing search interfaces is important for helping users find what they need easily. Here
are the key elements to consider:
 User Focus:
 The design should be made for the users. Understand what they want and how
they search. This helps create a simple layout that is easy to use.
 Simplicity:
 Keep the interface clear and simple. Use easy words and clear labels. The search
box should be easy to find, and users should be able to submit their searches
without confusion.
 Result Display:
 Show search results in a way that is easy to read. Rank results by how relevant
they are. Include short descriptions so users can quickly see if the result is what
they want.
 Query Changes:
 Users often need to change their search terms. The interface should make it easy
to modify searches and suggest related terms. This helps users find better
results.
 Visual Feedback:
 Use visuals like icons and colors to guide users. Provide feedback when a
search is happening, like showing a loading symbol. This helps users know their
search is being processed.
4. How do people search for information? Discuss the factors influencing their
behavior. (10 Marks)
How People Search
 Task Type: User interaction varies based on the task, domain expertise, and available
time.
 Information Lookup:
o Involves fact retrieval or question answering.
o Satisfied by discrete pieces of information (e.g., numbers, dates).
o Works well with standard web search interactions.
 Exploratory Search:
o Learning Search: Requires multiple query-response pairs, time spent scanning
and synthesizing information.
o Investigative Search: A long-term process involving multiple iterations,
critically assessing results, and integrating findings into personal or professional
knowledge bases.
 Classic vs. Dynamic Model:
o Classic model: Problem identification → information need articulation → query
formulation → results evaluation.
o Dynamic model: Emphasizes learning during the search; users adjust their
information needs based on retrieval results, known as the berry picking model.
 Navigation vs. Search:
o Navigation: Browsing through information structures, effective when well-
matched to the user’s needs.
o Browsing is less mentally taxing than recalling but can be frustrating without
appropriate links.
 Search Process Insights:
o Users often reformulate queries slightly and search for previously accessed
information.
o Interfaces support query history and revisitation.
o Searchers tend to focus on top-ranked results and struggle to assess the
relevance of documents.
o Experts use different strategies compared to novices.
 Key Observations:
o Users often have difficulty determining the relevance of search results.
o They tend to believe the top results are more relevant, neglecting the rest.
o Studies suggest that search interfaces should account for query reformulation,
revisitation, and user knowledge.