Ratna Sanyal

Semi-automatic generation of UML models from natural language requirements

Going from requirements analysis to design phase is considered as one of the most complex and difficult activities in software development. Errors caused during this activity can be quite expensive to fix in later phases of software development. One main reason for such potential problems is due to the specification of software requirements in Natural Language format. To overcome some of these defects we have proposed a technique, which aims to provide semi- automated assistance for developers to generate UML models from normalized natural language requirements using Natural Language Processing techniques. This technique initially focuses on generating use-case diagram and analysis class model (conceptual model) followed by collaboration model generation for each use-case. Then it generates a consolidated design class model from which code model can also be generated. It also provides requirement traceability both at design and code levels by using Key-Word-In-Context and Concept Location techniques respectively to identify inconsistencies in requirements. Finally, this technique generates XML Metadata Interchange (XMI) files for visualizing generated models in any UML modeling tool having XMI import feature. This paper is an extension to our existing work by enhancing its complete usage with the help of Qualification Verification System as a case study.

Static UML Model Generator from Analysis of Requirements (SUGAR)

In this paper, we propose a tool, named Static UML Model Generator from Analysis of Requirements (SUGAR), which generates both use-case and class models by emphasizing on natural language requirements. SUGAR aims at integrating both requirement analysis and design phases by identifying use-cases, actors, classes along with its attributes and methods with proper association among classes. This tool extends the idea of previously existing tools and implemented with the help of efficient natural language processing tools of Stanford NLP Group, WordNet and JavaRAP using the modified approach of Rational Unified Process with better accuracy. SUGAR has added new features and also able to incorporate solution for those problems existed in previous tools by developing both analysis and design class models. SUGAR generates all static UML models in Java in conjunction with Rational Rose and provides all functionalities of the system even though the developer is having less domain knowledge.

An Innovative Approach for Generating Static UML Models from Natural Language Requirements

Moving from requirements analysis to design is considered as one of the most complex and difficult activities of software development life cycle. Errors caused in this activity can be quite expensive to fix. Tool support for integrating both requirement analysis and design phases by automating some of the tasks involved in this activity is highly desirable. To this end we proposed a tool, named Static UML Model Generator from Analysis of Requirements (SUGAR), which generates static UML models by emphasizing on natural language requirements. This tool extends previously existing approaches and implemented with the help of efficient natural language processing tools using the modified approach of Rational Unified Process with better accuracy. SUGAR generates all static UML models in Java in conjunction with Rational Rose and provides all functionalities of the system even though the developer is having less domain knowledge.

Anaphora Resolution in Hindi Documents

This paper presents anaphora resolution as a technique of semantic analysis of text documents written in Hindi language. The focus is on texts that mainly employ simple sentences, such as childrens stories, short essays, etc. The technique works by locating sentences in the text that are semantically related through anaphors, analyzing their semantics and exploiting the latter for resolving referents of the respective anaphors. The approach used here is based on matching constraints for the grammatical attributes of different words. The algorithm for anaphora resolution has been tested extensively. The accuracy of anaphora resolution is nearly 96% for simple sentences and for compound and complex sentences; the accuracy is of the order of 80%. The causes of the errors are analyzed and possible techniques for improvements are discussed.

Clause Boundary Identification using Classifier and Clause Markers in Urdu Language

This paper presents the identification of clause boundary for the Urdu language. We have used Conditional Random Field as the classification method and the clause markers. The clause markers play the role to detect the type of subordinate clause, which is with or within the main clause. If there is any misclassification after testing with different sentences then more rules are identified to get high recall and precision. Obtained results show that this approach efficiently determines the type of sub-ordinate clause and its boundary.

Statistics and linguistic rules in multiword extraction:a comparative analysis

A hybrid methodology is proposed for extracting multiword expressions based on linguistic and statistical information. In the proposed methodology, N-grams are extracted by linguistic patterns and then various statistical measures are applied for classifying these N-grams as multiword expressions. To solve the problem of deciding cut-off boundary threshold in statistical filtering phase, a novel method for calculating boundary threshold is designed. Comparative analysis between the baseline method and the proposed methodology is presented. In the baseline method, firstly, N-grams are filtered by statistical measures and then linguistic filtering is applied. Precision, recall and ƒ-Score are calculated on manually annotated corpus. Observed results show that the proposed methodology provides good results for certain types of multiword expressions like compound nouns, verb-particles and verb-verb.

Semantic document classification and keyword spotting in digital repositories

The volume of documents in the digital repositories numbers in thousands and is increasing constantly. In such a scenario it becomes a very important issue to organize and retrieve these documents in a way that relates to the human mind. In this paper, we present a novel approach to classify the documents in a digital repository and find the semantically significant keywords related to those documents to make the organization and the retrieval of the documents expeditious. We approach this problem using probabilistic model with incomplete training data to organize them and mark the relevant keywords. This approach makes the classification faster and instead of the unlabeled clustering gives classification with well defined topics.