Visualizing UML’s Sequence and Class Diagrams Using Graph-Based Clusters
Download as PPTX, PDF0 likes45 views
This paper proposes a new methodology to generate UML class and sequence diagrams from source code and class files using a graph-based clustering approach. It performs keyphrase extraction from source code comments, constructs a source code dependency graph (SDG) based on contextual similarity, and clusters related documents together. Finally, it creates UML diagrams from the filtered top clusters. An analysis of existing UML diagram generation techniques is also provided. The proposed approach is useful for software maintenance engineers and developers to understand project structure for large open source projects.
![Algorithm1: Data Filtering
Input : Source code files SC, Class files CF.
Step 1: Read input source codes files SC.
Step 2: Read input class files CF.
Step 3:for each source code SCi in SC[]
Do
Parse source code SCi with methods M and Fields F.
Mi=ExtractMethods(SCi)
Fi=ExtractFields(SCi)
Mapping (Mi , Fi) to SCi
SC1 (M1,F1)
SC2 (M2,F2)
… …..
SCn (Mn,Fn)
done
Step 4: for each class file CFi in CF[]
Do
Parse class files CFi with methods M and Fields F.
Mi=ExtractMethods(CFi)
Fi=ExtractFields(CFi)
Mapping (Mi , Fi) to CFi
CF1 (M1,F1)
CF2 (M2,F2)
… …..
CFn (Mn,Fn)
done](https://image.slidesharecdn.com/chs202065-210304132047/85/Visualizing-UML-s-Sequence-and-Class-Diagrams-Using-Graph-Based-Clusters-9-320.jpg)
![Step 5: // Remove the duplicate methods and fields in each source code and class files
For each code Ci in i j
SC CF
Do
i i j
i i j
M Prob(M M / C);i j
F Prob(F F / C);i j
If( Mi!=0 AND Fi!=0)
Then
Remove Mi in Ci or Cj
Remove Fi in Ci or Cj
End if
Done
Step 6: //Pre-processing source code comments using Stanford NLP parser.
For each document di in D
Do
T[]=Tokenize(di)
For each token t in T[]
Do
Apply stemming, stopword removal using Stanford NLP library.
Done
Done](https://image.slidesharecdn.com/chs202065-210304132047/85/Visualizing-UML-s-Sequence-and-Class-Diagrams-Using-Graph-Based-Clusters-10-320.jpg)
![Contextual source code graph based clustering algorithm
Step 1: Read number of clusters c.
Step 2: Read number of iterations I.
Step 3: Initialize k random clusters as centroids.
Step 4: for each document at vertex V in graph
Do
TF-ID[]= Compute term frequency tf-id
Done
Step 5: Repeat until c clusters
Find nearest cluster distance metrics using the following equation
Let Document vector one V1, document vector two V2
2 2
3
Cosine(V1[i],V2[i])
Dist(V1,V2)
Correlation(V1,V2). V1[i] V2[i]
Done
Step 6: Merge the graph nodes using the nearest distance measure.
Step 7: Update cluster centroid using mean distance.
Step 8: Construct the class diagram using the plant UML library to the
filtered top k-clusters C[k].
Step 9: For each source code file SC[i] do
Check the source code file has distance metric >0
If(dist(SC[i],C[k])>0)
Then
Display class diagram in source code file SC[i].
End if
Step 10. done](https://image.slidesharecdn.com/chs202065-210304132047/85/Visualizing-UML-s-Sequence-and-Class-Diagrams-Using-Graph-Based-Clusters-15-320.jpg)
![Step 6: Class file dependency graph CDG is used to find the contextual similarity
between the vertex nodes to the neighbor metrics using the following proposed
measure.
Let U(CMi) (m1,m2,….mn) denotes the source codes metrics vector at vertex i.
V(CMj) (m1,m2,….mr) denotes the source code metrics vector at vertex j.
2 2 2
i 1 2 p
2 2 2
1 2 q
i j 1 1 2 2 p q
j
| U(CM ) | U(m ) U(m ) ....U(m )
| | V(m ) V(m ) ....V(m )
| U(CM ).V(CM ) | U(m ).V(m ) U(m ).V(m )... U(m ).V(m )
Pr oposed Contextual class code depenedency graph dissimilarity index
is computed as
CM
C
V
3
i j i j
i j
U(CM ).V(CM )*cos(| U(CM ) | | V(CM ) |)
CDGDI= ;where i j
2*(| U(CM ) |*| V(CM ) |)
Contextual class code depenedency graph similarity index
CCDGSI 1 CCDGDI;
For each class file in CF[i]
Do
Add to Sequence diagram designer S.
Done
Visualize sequence diagram to all the class files in the given relational packages.](https://image.slidesharecdn.com/chs202065-210304132047/85/Visualizing-UML-s-Sequence-and-Class-Diagrams-Using-Graph-Based-Clusters-16-320.jpg)
![Result in Form of
Creation of Clusters
Thursday, March 4, 2021
Cluster-1{
MultiNomialBMAEstimator.java
SimpleEstimator.java
}
[D@c88a32 = [2]
Cluster-2{
DiscreteEstimatorBayes.java
}
[D@17c2f4f = [0]
Cluster-3{
BayesNetEstimator.java
}
[D@80cdf3 = [1]
Cluster-4{
BMAEstimator.java
}
[D@f9296d = [3]
Cluster-5{
DiscreteEstimatorFullBayes.java](https://image.slidesharecdn.com/chs202065-210304132047/85/Visualizing-UML-s-Sequence-and-Class-Diagrams-Using-Graph-Based-Clusters-18-320.jpg)
Visualizing UML’s Sequence and Class Diagrams Using Graph-Based Clusters
- 1. Visualizing UML’s Sequence and Class Diagrams Using Graph-Based Clusters Paper ID 65 Nakul Sharma, Dr. Prasanth Yalla Department of Computer Science and Engineering Koneru Lakshmiah Education Foundation Vaddeswaram,Guntur-522502, India
- 2. Agenda • Abstract • Introduction • Literature Review • Proposed Methodology • Results & Discussion • Conclusion & Future Scope
- 3. Abstract The paper discusses the creation of UML diagram based recommendation system using java and class files as the input. The existing systems do not make use of techniques available in text-mining for creating UML diagrams. The overall methodology makes use of keyphrase extraction, contextual similarity calculation, and graph-based clusters in creating UML diagrams. The existing systems survey of state-of-art UML diagram generation techniques and keyphrase extraction survey is also provided. A comparative analysis of the existing tools for generating UML-diagrams is also provided. The recommendation system generated is useful to maintenance engineers and software developers.
- 4. Introduction • In the current work, a literature review of UML diagram construction from text or source code is done. A comparative analysis of different methods used in UML diagram construction is also proposed. • In this paper, the authors propose a contextual similarity approach combined with cluster and graph creation. A multi-step approach involves keyphrase extraction, graph construction, clustering of related documents together and finally creation of UML class & sequence diagrams.
- 5. Literature Review • There are several methodologies being used in developing UML diagrams. The most common diagrams which are developed are use-case and class diagrams. • However little work has been done wrt using text, source code, API documentation for generating UML diagrams. In addition text mining techniques are not used extensively in generating UML diagrams.
- 6. Existing System Developed (UML Diagram Generation) Sr. No. Name of Tool Generated Title of Publication Name of Author Publication Venue 1 Extended ForUML (2019) Extended ForUML for Automatic Generation of UML Sequence Diagrams from Object-Oriented Fortran Aziz Nanthaamornphong, Anawat Leatongkam Scientific Programming, Hindwai Publications, 2019 2 Automatic Builder of Class Diagram (ABCD) (2016) Automatic Builder of Class Diagram (ABCD): an application of UML generation from functional requirements Wahiba Ben Abdessalem, Karaa Zeineb Ben Azzouz, Aarti Singh, Nilanjan Dey, Amira S. Ashour, Henda Ben Ghazala Software Practice and Experience: Wiley Publication, 2016 3 RECAA (2015) From requirements to UML models and back: How automatic processing of text can support requirements engineering Mathias Landhaußer , Sven J. Korner, Walter F. Tichy Software Qual J,Springer, 2014 4 ForUML (2015) Extracting UML Class Diagrams from Object-Oriented Fortran: ForUML Aziz Nanthaamornphong,Jeffrey Carver, KarlaMorris, Salvatore Filippone Scientific Programming, Hindwai Publications, 2015 5 Class-Gen (2010) Parsed use case descriptions as a basis for object-oriented class model generation Mosa Elbendak, Paul Vickers∗, Nick Rossiter The Journal of Systems and Software, Springer, 2011 6 UMGAR (2008) An Automated Tool for Generating UML Models from Natural Language Requirements. Deeptimahanti, D. K. and Babar, M. A IEEE Conference, 2008 7 ER convertor (2008) Heuristics-based entity relationship modeling through natural language processing. Nazlia Omar , Paul Hanna, and Paul Mc Kevitt 15th Artificial Intelligence and Cognitive Science Conference, Ireland
- 7. Input files Text Pre- processing Calculation of Similarity Measures Source Code Clustering Constructing the UML Class and Sequence Diagram using Clusters SDG Representation Key-phrase extraction Overall Architecture Of Proposed Methodology
- 8. Module-1:A Conceptual Dependency Graph Based Keyword Extraction Model for Source Code to API Documentation Mapping
- 9. Algorithm1: Data Filtering Input : Source code files SC, Class files CF. Step 1: Read input source codes files SC. Step 2: Read input class files CF. Step 3:for each source code SCi in SC[] Do Parse source code SCi with methods M and Fields F. Mi=ExtractMethods(SCi) Fi=ExtractFields(SCi) Mapping (Mi , Fi) to SCi SC1 (M1,F1) SC2 (M2,F2) … ….. SCn (Mn,Fn) done Step 4: for each class file CFi in CF[] Do Parse class files CFi with methods M and Fields F. Mi=ExtractMethods(CFi) Fi=ExtractFields(CFi) Mapping (Mi , Fi) to CFi CF1 (M1,F1) CF2 (M2,F2) … ….. CFn (Mn,Fn) done
- 10. Step 5: // Remove the duplicate methods and fields in each source code and class files For each code Ci in i j SC CF Do i i j i i j M Prob(M M / C);i j F Prob(F F / C);i j If( Mi!=0 AND Fi!=0) Then Remove Mi in Ci or Cj Remove Fi in Ci or Cj End if Done Step 6: //Pre-processing source code comments using Stanford NLP parser. For each document di in D Do T[]=Tokenize(di) For each token t in T[] Do Apply stemming, stopword removal using Stanford NLP library. Done Done
- 11. Module-2 & 3 Source Code Dependency Graph Based Contextual Probabilistic Clustering Approach for class dependency Diagrams
- 12. Probabilistic Weighted based contextual similarity measure for Source code and class files dependency graph Input : Project source codes SC, Project class files CF, Project source metrics (SMi,SFi) and Project class metrics (CMi,CFi). Procedure: Step 1: Read source code metrics , sci(SMi,SFi) and Project class metrics cfi(CMi,CFi) Step 2: Constructing a source code dependency graph SDG(V,E) with vertex set V and Edge set E using source code metrics. Here vertex set V is represented with source code methods and fields and edge set E is represented as weighted rank between the vertices. Step 3: The probabilistic weights of the edges are computed using the vertex terms ti and tj where i i t V and j j t V .
- 13. i, j i,j i j i j Prob(t t ) Edgeweight : w(i, j) 2.max{Prob(t ),Prob(t )} Prob(t ,t ) i j Prob(t , t ) is the number of times both terms i j (t , t ) occurred together. i Prob(t ) is the number of occurrence of i t in vertex Vi j Prob(t ) is the number of occurrence of j t in vertex Vj Step 4: The vertices with positive edge weights are sorted in ascending order in the dependency graph to find the contextual similarity between the source code metrics. Step 5: Source code dependency graph SDG is used to find the contextual similarity between the vertex nodes to the neighbor metrics using the following proposed measure. Let U(SMi) (m1,m2,….mn) denotes the source codes metrics vector at vertex i. V(SMj) (m1,m2,….mr) denotes the source code metrics vector at vertex j.
- 14. 2 2 2 i 1 2 p 2 2 2 1 2 q i j 1 1 2 2 p q j | U(SM ) | U(m ) U(m ) ....U(m ) | | V(m ) V(m ) ....V(m ) | U(SM ).V(SM ) | U(m ).V(m ) U(m ).V(m )... U(m ).V(m ) Pr oposed Contextual source code dependency graph dissimilarity index is computed as SM C V 1 3 i j i j i j U(SM ).V(SM )*tan (| U(SM ) | | V(SM ) |) SDGDI= ;where i j 2*(| U(SM ) |*| V(SM ) |) Contextual source code dependency graph similarity index CSDGSI 1 CSDGDI;
- 15. Contextual source code graph based clustering algorithm Step 1: Read number of clusters c. Step 2: Read number of iterations I. Step 3: Initialize k random clusters as centroids. Step 4: for each document at vertex V in graph Do TF-ID[]= Compute term frequency tf-id Done Step 5: Repeat until c clusters Find nearest cluster distance metrics using the following equation Let Document vector one V1, document vector two V2 2 2 3 Cosine(V1[i],V2[i]) Dist(V1,V2) Correlation(V1,V2). V1[i] V2[i] Done Step 6: Merge the graph nodes using the nearest distance measure. Step 7: Update cluster centroid using mean distance. Step 8: Construct the class diagram using the plant UML library to the filtered top k-clusters C[k]. Step 9: For each source code file SC[i] do Check the source code file has distance metric >0 If(dist(SC[i],C[k])>0) Then Display class diagram in source code file SC[i]. End if Step 10. done
- 16. Step 6: Class file dependency graph CDG is used to find the contextual similarity between the vertex nodes to the neighbor metrics using the following proposed measure. Let U(CMi) (m1,m2,….mn) denotes the source codes metrics vector at vertex i. V(CMj) (m1,m2,….mr) denotes the source code metrics vector at vertex j. 2 2 2 i 1 2 p 2 2 2 1 2 q i j 1 1 2 2 p q j | U(CM ) | U(m ) U(m ) ....U(m ) | | V(m ) V(m ) ....V(m ) | U(CM ).V(CM ) | U(m ).V(m ) U(m ).V(m )... U(m ).V(m ) Pr oposed Contextual class code depenedency graph dissimilarity index is computed as CM C V 3 i j i j i j U(CM ).V(CM )*cos(| U(CM ) | | V(CM ) |) CDGDI= ;where i j 2*(| U(CM ) |*| V(CM ) |) Contextual class code depenedency graph similarity index CCDGSI 1 CCDGDI; For each class file in CF[i] Do Add to Sequence diagram designer S. Done Visualize sequence diagram to all the class files in the given relational packages.
- 17. Thursday, March 4, 2021 Key Phrases in SDG :{m_items.iterator()} {m_items.add(i)} Score :0.9073701027137411 Key Phrases in SDG :{m_items.iterator()} {Collections.sort(m_items)} Score :0.9073701027137411 Key Phrases in SDG :{m_items.iterator()} {m_items.get(index)} Score :0.9073701027137411 Key Phrases in SDG :{m_items.iterator()} {m_items.size()} Score :0.8626786872190586 Key Phrases in SDG :{m_items.iterator()} {m_items.iterator()} Score :0.826985987428094 Key Phrases in SDG :{m_items.iterator()} {i.hasNext()} Score :1.0 Key Phrases in SDG :{m_items.iterator()} {i.next()} Score :1.0 Key Phrases in SDG :{m_items.iterator()} {i.next().toString()} Score :1.0 Key Phrases in SDG :{m_items.iterator()} {buff.append(i.next().toString() + “ “)} Score :1.0 Key Phrases in SDG :{i.hasNext()} {Collections.sort(m_items)} Score :1.0 Key Phrases in SDG :{i.hasNext()} {m_items.add(i)} Score :0.9073701027137411 Key Phrases in SDG :{i.hasNext()} {Collections.sort(m_items)} Score :1.0 Key Phrases in SDG :{i.hasNext()} {m_items.get(index)} Score :1.0 Key Phrases in SDG :{i.hasNext()} {m_items.size()} Score :1.0 Key Phrases in SDG :{i.hasNext()} {m_items.iterator()} Score :1.0 Key Phrases in SDG :{i.hasNext()} {i.hasNext()} Score :0.826985987428094 Key Phrases in SDG :{i.hasNext()} {i.next()} Score :0.8626786872190586
- 18. Result in Form of Creation of Clusters Thursday, March 4, 2021 Cluster-1{ MultiNomialBMAEstimator.java SimpleEstimator.java } [D@c88a32 = [2] Cluster-2{ DiscreteEstimatorBayes.java } [D@17c2f4f = [0] Cluster-3{ BayesNetEstimator.java } [D@80cdf3 = [1] Cluster-4{ BMAEstimator.java } [D@f9296d = [3] Cluster-5{ DiscreteEstimatorFullBayes.java
- 19. Results in form of Diagrams Generated • Class Diagram Sequence Diagram
- 21. Analysis of Existing Systems Name of UML Tool Techniques / Input files used for Conversion NLP SOFTWARES NLP AND Rules (Heuristics) XMI/XML Representation Source Code API Documentation Automatic Builder of Class Diagram (2016) No Yes Yes No Yes RECAA (2015) Yes Yes No No Yes CM-Builder (2000) Yes Yes No No No UMGAR (2008) Yes Yes No No No SENSE (2007) Yes Yes No No No ER convertor (2008) No Yes No No No LIDA (2001) No Yes No No No ForUML (2015) Yes Yes Yes Yes No Extended ForUML (2019) Yes Yes Yes Yes No SDG Graph Based Yes Yes No Yes Yes Thursday, March 4, 2021
- 22. Conclusion • The paper discusses how UML diagram can be used as a tool for recommending most essential classes within a given set of project. A large-scale open source project cannot be assessed using the existing similarity measures. Hence, a new hybrid probabilistic model is proposed for large open-source projects
- 23. References • Radoslav Kirkov, Gennady Agre, “Source Code Analysis – An Overview”, Cybernetics And Information Technologies, Volume 10, No 2, Bulgarian Academy Of Sciences, 2010. • Mohammed J. Zaki, Wagner Meira Jr., “Data Mining and Analysis: Fundamental Concepts and Algorithms”, Chapter 13, page 370. • “About the Unified Modeling Language Specification Version 2.5” , https://www.omg.org/spec/UML/2.5/About-UML/ • Nakul Sharma, Prasanth Yalla, “A Hybrid Weighted Probabilistic based source code graph clustering algorithm for class diagram and sequence diagram visualization”, --Under Review. • Mariem Abdouli, Wahiba Ben Abdessalem Karaa, Henda Ben Ghezala, "Survey of Works that Transform Requirements into UML Diagrams", SERA 2016, June 8-10, 2016, Baltimore, USA, ISBN: 978-1-5090-0809-4 • B.A.K.Wahiba, B.A. Zeineb. S.Aarti. D.Nilanjan. A.Amira. B.G. Henda. Automatic builder of class diagram (ABCD): an application of • UML generation from functional requirements. Software: Practice and Experience (2015). Published online in Wiley Online Library. • Mathias Landha ̈ußer • Sven J. Ko ̈rner • Walter F. Tichy, "From Requirements to UML Models & Back : How automatic processing of text can support requierments engineering", Software Qual J, DOI 10.1007/s11219-013-9210-6, pp 1-29. • Harmain Mohamed Harmain and Robert J. Gaizauskas. CM-Builder: An automated NLbased CASE tool. In ASE, pages 45-54, 2000. • Herchi H, Ben Abdessalem W (2012). From user requirements to UML class diagram. International Conference on Computer Related • Knowledge. 4 Nov 2012. • Deeptimahanti, D. K. and Babar, M. A. An Automated Tool for Generating UML Models from Natural Language Requirements. IEEE/ ACM int.Conf. on ASE, 2009. • Fabbrini F., M. Fusani, Gnesi S., Lami G., "An automatic quality evaluation for natural language requirements", • 7th International Workshop on Requirements Engineering: Foundation for Software Quality, pp. 150-164, Interlaken, Switzerland, 4-5 Giugno 2001. • Omar N, Hanna P, Mc Kevitt P (2004) Heuristics-based entity relationship modeling through natural language processing. Proceedings • of the 15th Irish Conference on Artificial Intelligence and Cognitive Science (AICS-04) 302-313. • Zhenchang Xing and Eleni Stroulia. Umldiff: an algorithm for object oriented design differencing. In Proceedings of the 20th IEEE/ACM international Conference on Automated software engineering, ASE '05, pages 54{65, New York, NY, USA, 2005. ACM. ISBN 1-58113-993-4. • Overmyer, S., Benoit, L., Rambow, O. Conceptual Modeling through Linguistic Analysis Using LIDA. 23rd International Conference on • Software Engineering. 2001. • Aziz Nanthaamornphong,Jeffrey Carver,Karla Morris,Salvatore Filippone, "Extracting UML Class Diagrams from Object-Oriented Fortran: ForUML", Hindawi Publishing Corporation, Scientific Programming, Volume 2015, 15 pages, http://dx.doi.org/10.1155/2015/421816 • Aziz Nanthaamornphong, Anawat Leatongkam, "Extended ForUML for Automatic Generation of UML Sequence Diagrams from Object-Oriented Fortran" Hindawi, Scientific Programming, Volume 2019, https://doi.org/10.1155/2019/2542686
- 24. Thank You, Any Questions ?