Research--probabilistic Models In Information Retrieval

Norbert Fuhr

PROBABILISTIC MODELS IN INFORMATION RETRIEVAL

Introduction  The

intrinsic uncertainty of IR.  Two approaches: Relevance models Proof-theoretic model

Relevance models  A user

assigns relevance judgments to document w.r.t. his/her query.

 The

IR systems yield the approximation of the set of relevant documents.

 Some

models: BIR model, BII model, DIA model, etc…

Relevance models  Binary

independence retrieval model

(BIR) A document d_m is composed of a set of

terms and represented as a vector.  Assumptions: “cluster hypothesis”: Terms are distributed

differently within relevant and non-relevant documents. A query q_k is also a set of terms.

Relevance models

Relevance models 

An example

Ranking is (1,1),(1,0),(0,1),(0,)

The probability ranking principle 

Let C be the costs for the retrieval of a relevant document. for non-relevant documents.

Retrieve that document for which the expected costs of retrieval are a minimum.

Proof-theoretic model  IR

is interpreted as uncertain inference.

 A generation

of deductive databases:

queries and contents are treated as logical

formulas. The query has to be proved from the formulas.  A document

is an answer for a query iffthe logic formula is true.

Jane Cleland-Huang, Reffaella Settimi, Oussama BenKhadra, Eugenia Berezhanskaya, Selvia Christina

GOAL-CENTRIC TRACEABILITY FOR MANAGING NONFUNCTIONAL

 Non-Functional

Requirements (NFR) are

difficult to trace: Global impact upon a software system Extensive network of interdependencies and

trace-offs

 Goal

centric traceability (GCT) approach:

NFRs are modeled as goals and

operationalizations within SIG.

Dynamically establish traces from impacted

functional design element to elements in SIG.

Softgoal Interdependency Graph

GCT Model

 Impact

detection in GCT

Documents Queries Index terms

 The

relevance of a document query q is pr( ,q)

to a

Jane Cleland-Huang, Reffaella Settimi, Chuan Duan, Xuchang Zou

UTILIZING SUPPORTING EVIDENCE TO IMPROVE DYNAMIC

Introduction  Current

work

Recall level close to 90% Precision from 10% to 45%.

 Target: Maintain recall level at least 90% Precision at least 20%

Introduction  Three

strategies to improve the performance of dynamic requirements traceability: Hierarchical modeling Logical clustering of artifacts Semi-automated pruning of the probabilistic

network.

Enhancement strategies

Motivation Example

Hierarchical Enhancement

 R3

label is “De-icing”  Using hierarchical information in R3 -> R5 describe de-icing service.  Similarly, C4 describe about truck maintenance service. The link between C4 and R5 is not correct !!!

Hierarchical Enhancement

 Solution:

Build a DAG graph to display the direct

relationship between artifacts.

 Results

Clustering Enhancements

 Links

tend to occur

in clusters: q <-> d_j => higher prob that q <-> d q <-> q_i => higher prob that d <-> q  Care

about relationship of sibling artifacts.

Clustering Enhancements

 Solution

Clustering Enhancements

 Evaluation

Graph Pruning Enhancement  Observation: Word “schedule” used for both de-icing

schedule and truck maintenance schedules Query with “schedule” will returns artifacts from both domains make precision lower.

Graph Pruning Enhancement  Solution: Utilize initial decision made by the analyst to

place constraints and improve precision in “problematic” area. Rules to place constrains:

1. One or more links between two groups are all rejected by an analyst. 2. Basic retrieval algorithm generated candidate links between two groups.

Graph Pruning Enhancement  Evaluation