Roshni R. Ramnani

A framework for identifying and analyzing non-functional requirements from text

Early identification of Non-Functional Requirements (NFRs) is important as this has direct bearing on the design and architecture of the system. NFRs form the basis for architects to create the technical architecture of the system which acts as the scaffolding in which the functionality of the same is delivered. Failure to identify and analyze NFRs early-on can result in unclassified, incomplete or conflicting NFRs, and this typically results in costly rework in later stages of the software development. In practice, this activity is primarily done manually. In this paper, we present a framework to automatically detect and classify non-functional requirements from textual natural language requirements. Our approach to identify NFRs is based on extracting multiple features by parsing the natural language requirement whereby the presence of a certain combination of and relationship among the features uniquely identifies the requirement as an NFR of a particular category. These features are specified as pattern based rules which can be specified in a human readable language through the use of a domain specific language that we have defined. This enables great ease and flexibility in creating and extending rules. Our approach has been implemented as a prototype tool and here we also present the results of applying our approach on a publicly available requirement corpus.

Verb-based Semantic Modelling and Analysis of Textual Requirements

Automated machine analysis of natural language requirements poses several challenges. Complex requirements such as functional requirements and use cases are hard to parse and analyze, the language itself is un-constrained, the flow of requirements may be haphazard, and one requirement may contradict another - to name a few challenges. In this paper, we present a lightweight semantic modeling technique through natural language processing to filter requirements and create a semi-formal semantic network of requirement sentences. We employ novel techniques of classification of verbs used in requirements, semantic role labeling, discourse identification, and a few verb entailment and dependency relationships to generate a lightweight semantic network and critique the requirements. We discuss the design of the model and some early results obtained from analyzing real-life industrial requirements.

Smart Entertainment - A Critiquing Based Dialog System for Eliciting User Preferences and Making Recommendations

We present a Critiquing based dialog system that can make media content recommendations to users by eliciting information through active exploration of user preferences for item attributes. The system and user communicate through a natural language mixed-initiative conversational interface in which the system guides the user to a specific choice. During the conversation, the system presents the user with several options and analyzes the responses or “critiques”. The system starts with general recommendations or relevant candidates and refines this as it learns more about user’s preferences in subsequent iterations. These choices made by the user and the textual feedback/reviews that can be optionally provided, is used to infer a user preference model for the item.

Semi-Automated Information Extraction from Unstructured Threat Advisories

One of the fundamental challenges for information officers of most organizations today is the growing number of cyber security threats. This has led to an emerging field of Cyber Threat Intelligence, which is a mechanism to acquire, categorize and prioritize information regarding impending security threats from disparate online sources, enabling organizations to take the necessary steps to avoid compromising client data and protecting their hardware and software resources. Such information is published as formal security advisories which are largely in the form of unstructured or semi structured data. In this work we describe an approach to read large volume of such unstructured data and automatically extract useful nuggets of information like the exploit targets, techniques for the exploitation and recommended prevention guidelines. We use natural language processing techniques and a pattern identification framework to extract these information nuggets. We present some early results and observations.