Maurizio Pighin

1
Venezia, 14-16 Marzo 2001Using IR Techniques
Catalogue, Reuse and Maintain Existing Software
Objects

• Maurizio Pighin
• Dipartimento di Matematica e Informatica
• Università di Udine, Italy
• e-mail pighin_at_dimi.uniud.it

2
Introduction

• Using IR tools to catalogue software life-cycle
• Work done up-to-date
• First Model Definition
• Experimental Analysis
• Result Assessment
• Future perspectives
• Model extension to specification
• Definition of mobility feature
• New Experimental Analysis

3
Introduction

• Existing SW Objects
• Specifications
• Data-Model
• Code
• Test Chain
• Software Object Catalogue
• Repository of heterogeneous information

4
Introduction

• Difficulties
• Catalogue organization definition
• Adopted methodology to fill it
• Supported techniques to analyze it
• with Existing Software Objects
• The economic break-even point is very high
• Novel approach
• Understand how heterogeneous information must be
  organized
• Application of Information Retrieval techniques
• Experimental evaluation on these techniques

5
The Proposed Methodology

• First Step
• Verification on code-objects the soundness of
  proposal
• ALICE (AppLication of Information Retrieval to
  Catalogues of Existing software) is the
  Experimental System produced

6
The Proposed Methodology

• Search Functionalities which must be at disposal
  of user
• Structure of objects which must be catalogued
• task-analysis of Sw-engineer work permits us to
  built a basic functionalities model

7
The Proposed Methodology

• Four basic search methodologies on a Software
  Catalogue
• semantics-driven search
• data-definition driven search
• structure-driven search
• syntax-driven search
• These search mechanisms can be combined in more
  complex strategies
• From mechanism we found the feature which defines
  the keys of catalogue search

8
The Proposed Methodology

• The structure can automatically be built by
  SW-Objects and can be indexed in the catalogue
• Depending from object format and from requested
  feature specialized parsers must be built which
  analyze the objects end extract interesting
  feature references

9
The IR tool

• ALICE is built on Z-PRISE (statistical engine)
• We built the specialized parser which analyzes
  and indexes an industrial SW package (about
  150.000 C code lines)
• We defined a specialized client which easily
  interacts with query, features and the ranked
  list extracted by Z-Prise engine

10
(No Transcript)
11
The experiment

• Ten students, one expert
• Three management problems
• 20 minutes for every problem
• Operation objective analysis
• SW-Monitor with operation-logAnalisi soggettiva
• Operation Subjective analysis
• Questionnaries
• Semantic differential
• Likert Scale
• (used the know-how of project IRNLI-Information
  Retrieval, Natural Language Interface)

12
Results
13
Le Valutazioni

• Score mechanism
• Interesting results about
• Effectiveness
• Easy of use
• Efficiency
• The system seem powerful and sound
• This part of work is presented at
• International Conference on Software Maintenance
  and Reuse, CSMR 2001, Lisboa
• Information Software Technology, Vol 51/2-3,
  Elsevier Science Publications, Clare (IR), 2000,
  pp. 131-138

14
The Actual/Future Steps

• Attention on other SW objects
• Specification mechanism
• UML
• E-R
• Informal specification
• Actual activities
• Feature definition
• Particular attention at mobility aspects
  (PhD-Thesis Gian Pietro Picco Milan Politecnico
  team)
• Mobility code and data
• Security
• Traslation
• Communication
• ..

15
Reference schemas
16
I Passi Attuali/Futuri

• Continue Activity
• Definition of Data Base on which work (Ericsson -
  UTMS specification Team)
• Application of more specification schemas
• Parser (feature) definition
• Search interface definition
• Search schemas (feature) definition
• Test on a significant Data Base
• Assessment of results

17
The potential results

• (Reasonable) Automatic Cataloguing of existing
  specifications
• Homogenization of specifications defined in
  heterogeneously (like UTMS specification team)
• Use of powerful cataloguing and search IR tools