This paper describes a natural language interface to database systems which is based on the query formation capabilities of a High-level Query Formulator. The formulator relies on the Semantic Graph of the database, which is a model of the data stored in the database. The natural language interface accepts a user input in natural language and extracts the necessary information needed by the formulator. This extraction process is performed using keywords obtained from the Semantic Graph and the database. Because keywords may have several meanings within a given domain, keyword meaning disambiguation is done using a statistical approach which involves comparing vectors of n-grams. N-grams are n contiguous words within a given text of natural language and they are capable of capturing lexical context. Traditionally, natural language interfaces have been heavy with grammars and other knowledge, but have been wide-ranging in functionality. The interface presented in this paper is more porta...

This paper describes a case study on application of natural language processing in very early stages of software development. At this early stage it is very important for the domain expert (who is, most probably, the future user) and the software expert to define a common language, understood by both of them. To define such a common language, we extract terms from the text written by domain expert, classify these terms and build a domain ontology using them. In our previous

Software engineers must systematically account for the broad scope of environmental behavior, including nonfunctional requirements, intended to coordinate the actions of stakeholders and software systems. The Inquiry Cycle Model (ICM) provides engineers with a strategy to acquire and refine these requirements by having domain experts answer six questions: who, what, where, when, how, and why. Goal-based requirements engineering has led to the formalization of requirements to answer the ICM questions about when, how, and why goals are achieved, maintained, or avoided. In this article, we present a systematic process called Semantic Parameterization for expressing natural language domain descriptions of goals as specifications in description logic. The formalization of goals in description logic allows engineers to automate inquiries using who, what, and where questions, completing the formalization of the ICM questions. The contributions of this approach include new theory to conceptually compare and disambiguate goal specifications that enables querying goals and organizing goals into specialization hierarchies. The artifacts in the process include a dictionary that aligns the domain lexicon with unique concepts, distinguishing between synonyms and polysemes, and several natural language patterns that aid engineers in mapping common domain descriptions to formal specifications. Semantic Parameterization has been empirically validated in three case studies on policy and regulatory descriptions that govern information systems in the finance and health-care domains.

A new query formulation system based on a semantic graph model is presented. The graph provides a semantic model for the data in the database with userdefined relationships. The query formulator allows users to specify their requests and constraints in highlevel concepts. The query candidates are formulated based on the user input by a graph search algorithm and ranked according to a probabilistic information measure. English-like query descriptions can also be provided for users to resolve ambiguity when multiple queries are formulated from a user input. For complex queries, we introduce an incremental approach, which assists users to achieve a complex query goal by formulating a series of simple queries. A prototype system with a multimodal interface using the high-level query formulation techniques has been implemented on top of a cooperative database system (CoBase) at UCLA. 1 Introduction Many database applications require users to formulate ad-hoc queries instead of invocation ...

This paper describes an approach for analysis of natural language requirements documents and two case studies conducted to prove the feasibility of the approach. The goal of the analysis is to define a common language understood both by the domain expert and the software engineer. To define such a common language, it is necessary to extract terms from the text written by domain expert. The extracted terms must be classified to build a taxonomy. This taxonomy is augmented by associations between terms to build a domain ontology. The paper introduces two case studies that illustrate feasibility of the approach. The first case study was conducted to test the approach itself and showed that the approach works, but certain interaction with human analyst is necessary. The second case study showed that this interaction does not consume too much time so that the approach scales to larger documents. 1

Abstract. Surprisingly little research has investigated the factors contributing to the successful practice of process modeling, in particular those contributing to the development of process models that facilitate human understanding. This research discusses the use of text and icons for labeling the graphical constructs in a process model. We discuss the use of two verb classification schemes to propose a systematic approach for describing, and graphically representing, domain semantics in labels used in process model constructs such that human understanding can be improved. We argue that the systematic use of domain labels and corresponding icons will result in process models that are easier and more readily understandable by end users. Our findings contribute to an ongoing stream of research investigating the practice of process modeling and thereby contribute to the body of knowledge about conceptual modeling quality overall. 1

An input to the conceptual database design phase contains a description of the target database. This description is usually given in some natural language, for instance in English. Conceptual design aims at transforming English sentences into a conceptual database schema. A conceptual data model should therefore possess modeling features that can represent English sentence structures. Since Chen himself has argued that the basic ER model has such constructs many extensions of this basic ER model have been proposed. Based on these new features we revise the several correspondences between English sentence structures and concepts of ER modeling. It turns out that EER modeling can provide both a well-defined semantics and improved modeling elements that naturally reflect English language sentence structures. 1

A requirements analysis step plays a significant role on the development of information systems, and in this step we produce various kinds of abstract models of the systems (called requirements models) according to the adopted development processes, e.g. class diagrams in the case of adopting object-oriented development. However, constructing these models of sufficient quality requires highest intellectual tasks and skills of human requirements analysts. In this paper, we develop a computerized tool to extract from a set of Japanese text documents conceptual information, called conceptual graph, which can be used as intermediate representation to generate software requirements models. More concretely, by applying the variation of text-mining techniques that we have developed, we extract significant words from text documents referring to the same problem domain and identify relevant relationships among them. The extracted words can be considered as concepts and they are constituents of a conceptual graph in the domain. This constructed graph can be used for generating requirements models, e.g. object oriented models, feature model, and even as a domain ontology that can be utilized during requirements analysis activities. We have made experimental analyses of our tool. This paper also includes the discussion on how the extracted conceptual graph can act as an object-oriented model, a feature model and a domain ontology, in order to show its wide applicability.

The current practice of graphical icon design is very much an art than a science. This paper proposes a scientific-oriented methodology for graphical icon design..It incorporates several ideas from data/system modeling and natural languages in the design and interpretations of icons. Specifically, it incorporates ideas of multi-leveling modeling and type-instance concepts. It also incorporates the concepts of grammar structures from alphabetic-based languages such as English and the principles of ideogram construction of graphical iconbased natural languages such as Chinese and Ancient Egyptian Hieroglyphs. The methodology also draws on other scientific disciplines such as cognitive science and user testing. 1.

Policies and government regulations impose restrictions on information practices in healthcare and finance. These restrictions govern the use and disclosure of information that spans organizations and their business practices. To comply with policies and the law, organizations must demonstrate that they have verifiable procedures in-place to implement these restrictions. To this end, we present techniques that software engineers can use to systematically acquire software artifacts from natural language policies and regulations based on our in-depth analysis of the U.S. Health Insurance Portability and Accountability Act 1 (HIPAA). The techniques apply semantic primitives to regulatory statements to express class structures using the Z notation. From these structures, software engineers distinguish between necessary and discretionary software requirements and acquire the following software artifacts: specifications for transactions including interfaces between software and business processes; data schemas and data maintenance requirements; and event-based test cases for ensuring that systems comply with policies and regulations. 1.