Can Development Work Describe Itself?
- 1. MSR’2010, Cape Town, South Africa, Mai 2010 Can Development Work Describe Itself? Walid Maalej, Technische Universität München Hans‐Jörg Happel, FZI Research Center Karlsruhe
- 2. ExecuIve Summary Informal notes that To a large extent, describe developers’ work descripKons work contain well‐ can be generated by defined semanKcs, observing the work granularity levels, and context of developers informaKon paMers and their interacKons 1 2 Grounded Theory on Work DescripIons © W. Maalej, Mai 2010 Analyzing Work DescripIon – MSR 2010 2
- 3. Outline 1 MoIvaIon 2 Research SeRng 3 Research Results 4 Work DescripIon AutomaIon © W. Maalej, Mai 2010 Analyzing Work DescripIon – MSR 2010 3
- 4. What Are Work DescripIons? Personal note Commit message ArIfacts Comments including work descripIons Social media Time sheet A work descripIon is an informal text wriTen by a knowledge worker to summarize achievements and other notable issues of a parIcular work session © W. Maalej, Mai 2010 Analyzing Work DescripIon – MSR 2010 4
- 5. Previous Studies Showed InteresIng ProperIes of Work DescripIons Regularities in Content and Effort and Quality Issues Metadata 5% of developers‘ Ime is The overall vocabulary spent for describing work usage seems to be (30 min. per day ) predictable 10% of the sessions have The vocabulary size is pseudo descripIons rather small (either no Ime or not Different projects have moIvaIon) similar ranking of terms To which extent can developers‘ work descriptions be automated? © W. Maalej, Mai 2010 Analyzing Work DescripIon – MSR 2010 5
- 6. Outline 1 MoIvaIon 2 Research SeRng 3 Research Results 4 Work DescripIon AutomaIon © W. Maalej, Mai 2010 Analyzing Work DescripIon – MSR 2010 6
- 7. Research QuesIons Content of Work DescripIons The semanIcs of informaIon included in work descripIons InformaIon EnIIes InformaIon Granularity Text fragments with similar The levels of detail included semanIcs (abstracIon levels) Occurrences CombinaIons Preferences Levels Causes Which infor‐ Do certain How are these What are Which properIes maIon enIIes developers prefer enIIes granularity effect the are included and certain combined? levels? granularity? how o`en? informaIon? © W. Maalej, Mai 2010 Analyzing Work DescripIon – MSR 2010 7
- 8. Data Sets Collected in Different Contexts Data set Summary Period Developers Entries Developers‘ personal MyComp notes at a German 2001 – 2009 25 38,005 soTware company Commit messages Apache and code comments 2001 – 2009 1,145 598,418 of all Apache projects Commit messages Unicase and code comments 2008 – 2009 18 5097 of the unicase project Personal notes in a Eureka observaKonal study at 2008 21 91 5 companies © W. Maalej, Mai 2010 Analyzing Work DescripIon – MSR 2010 8
- 9. The Data Analysis Process © W. Maalej, Mai 2010 Analyzing Work DescripIon – MSR 2010 9
- 10. Outline 1 MoIvaIon 2 Research SeRng 3 Research Results 4 Work DescripIon AutomaIon © W. Maalej, Mai 2010 Analyzing Work DescripIon – MSR 2010 10
- 11. InformaIon EnIIes and Their Usage Frequencies Occurrences % Entity Average Apache Mycomp Unicase Eureka Activity 71 69 76 71 67 Artifact 55 60 53 49 58 Problem 47 47 47 49 45 Rationale 28 30 29 25 31 New Work 24 24 20 28 22 Status 19 24 20 17 15 Reference 15 15 19 17 10 Solution 15 19 15 16 11 Experience 10 11 6 9 13 © W. Maalej, Mai 2010 Analyzing Work DescripIon – MSR 2010 11
- 12. Findings on InformaIon EnIIes 1 The majority of informaKon on performed acKviKes (82%) is combined with concerned arKfacts 2 InformaKon on problems is used to describe work done, work need to be done, and the context of experiences 3 The combinaKon paMerns show that sharing knowledge and managing work are two goals of work descripKons 4 There two clusters of developers: those who prefer to use arKfacts and those who prefer to use problems to describe work © W. Maalej, Mai 2010 Analyzing Work DescripIon – MSR 2010 12
- 13. Granularity Levels and Usage Frequencies Granularity Occurrences % Level Average Apache Mycomp Unicase Eureka Implementation 54 58 37 62 60 Domain Project 31 29 34 29 30 Requirement 12 10 26 6 7 Method 33 33 49 28 20 Class 29 29 25 31 32 Object Line 17 17 8 17 27 Component 15 14 16 19 10 Edit 53 55 41 57 60 Activity SE Process 36 34 42 30 39 Knowledge 12 13 15 11 9 © W. Maalej, Mai 2010 Analyzing Work DescripIon – MSR 2010 13
- 14. Findings on InformaIon Granularity 1 The majority of work descripKons (62%) include informaKon from a single granularity level 2 Developers think consistently (in a single abstracKon level) when taking notes about arKfacts 3 The shorter the session is the more fine‐grained are the described arKfacts 4 Levels of acKvity granularity overlap (edit, process and knowledge) © W. Maalej, Mai 2010 Analyzing Work DescripIon – MSR 2010 14
- 15. Outline 1 MoIvaIon 2 Research SeRng 3 Research Results 4 Work DescripIon AutomaIon © W. Maalej, Mai 2010 Analyzing Work DescripIon – MSR 2010 15
- 16. Two Main Enablers For AutomaIng Work DescripIons AutomaKng Work DescripKon Shared semanKcs of developers’ HeurisKcs derived from working context, i.e. acKviKes, empirical findings on arKfacts, and problems developers’ behavior © W. Maalej, Mai 2010 Analyzing Work DescripIon – MSR 2010 16
- 17. Shared SemanIcs to Annotate Context: Developers’ InteracIons © W. Maalej, Mai 2010 Analyzing Work DescripIon – MSR 2010 17
- 18. Shared SemanIcs to Annotate Context Developers’ ArIfacts © W. Maalej, Mai 2010 Analyzing Work DescripIon – MSR 2010 18
- 19. HeurisIcs to Generate Work DescripIons Appropriate Granularity • Guess the appropriate Relevant vs. Irrelevant Context level of detail 2 • Only a subset of artifacts concerned by the interactions 1 Problem-Solution States is included in the description • Detect if a developer is Four factors to • Useful metrics are accumulated encountering a problem, generate work usage duration, usage age, and searching for a solution, or description usage frequency applying a solution 3 • Indictors are are error messages, breakpoint usage, searches, or 4 usage of particular keywords Developers Preferences • Learn from previous behavior of developers and which information they describe in which situation © W. Maalej, Mai 2010 Analyzing Work DescripIon – MSR 2010 19
- 20. Summary of the Talk • InformaKon on acKviKes, arKfacts, problems, new InformaIon work, and status is included for work management • Most informaIon EnIIes • InformaKon on soluKons, raKonale, and experience enIIes can be is included to capture and share knowledge created automaIcally by observing • There are different levels of domain, object, and developer’s InformaIon acKvity granularity context Granularity • These are used consistently and with common paMerns • For that we propose a set of ontologies and • Developers either think problem‐centered or heurisIcs to be Developers‘ arKfact‐centered when describing their work used Preference • They use well defined informaKon paMerns such as <acKvity concerns arKfacts> © W. Maalej, Mai 2010 Analyzing Work DescripIon – MSR 2010 20
- 21. Feedback, QuesIons, SuggesIons and CollaboraIon are Welcomed! Walid Maalej Hans‐Jörg Happel TUM FZI maalejw@cs.tum.edu happel@fzi.de © W. Maalej, Mai 2010 Analyzing Work DescripIon – MSR 2010 21