Contributions published at Requirements Engineering (Martin Glinz)

null

[Context and motivation] Software companies have recognized the value of end-user feedback for requirements engineering. They gather feedback data from multiple feedback channels used by a large number of end-users who share needs, experiences, and ideas about software systems. [Question/problem] Exploiting end-user feedback as valuable source for requirements engineering implies an incorporation of feedback data into software development. In practice, however, limited resources, unsuitable tools, and missing systematic approaches make feedback analysis and processing cumbersome for software companies. If the feedback is intractable and its impact only latent, at worst, relevant feedback data is ignored. [Principal ideas/results] We propose a tool-supported, semi-automatic feedback process that allows to establish a traceability link between incoming feedback and documented requirements. We iteratively developed the web application Feedback Tracer Tool as an integrator between feedback gathering and issue tracking systems. For the traceability method, parameters of syntactic text similarity, topic models, and information retrieval algorithms were optimized with Grid Search for a top-ten candidate list. The findings from a test-session with four practitioners show that the Feedback Tracer Tool is usable, utile, and self-explaining. Used for a test data set with natural language feedback and JIRA issues, the text analysis algorithms reached adequate recall rates and F2 scores. [Contribution] The usage of candidate lists allows for efficient feedback processing and replaces a manual search through all artefacts. Our tool and method evaluation provides promising initial insights, but is limited to the connected external systems, the used data sets, and the chosen test procedure. However, the Feedback Tracer Tool and traceability method can be easily integrated into industrial and research projects, stimulating further investigations on tool-supported feedback traceability.

Prioritizing requirements is a crucial ingredient of successful Requirements Engineering (RE). The popular prioritization techniques assume that stakeholders are known and can be mandated to contribute to the prioritization process. This prerequisite no longer holds for many of today's systems where significant stakeholders (end-users, in particular) are outside organizational reach: they are neither known nor can they be identified among the members of the involved organizations. Classic techniques for involving these stakeholders such as polls or questionnaires are neither interactive nor collaborative, which is detrimental for prioritization. Social media enable collaborative prioritization, but fall short in motivating stakeholders outside organizational reach to participate voluntarily. In this light, we are developing the Garuso platform, which combines social media with gamification for motivating stakeholders. While first approaches to employing gamification in RE are promising, research is still in its infancy. Especially, little is known about the influence of the gamification algorithms controlling single game elements on the stakeholders' activities. In this paper we report on a field experiment in which we investigated this influence with Garuso. We found statistically significant differences between different algorithms controlling single game elements on the contributions of stakeholders to the prioritization of requirements.

Keeping requirements and acceptance test documents aligned and up-to-date plays an important role in the success of software projects. In practice, these documents are not always aligned with each other, nor with the actual system behaviour. A previous study showed that even when requirements are updated, acceptance tests might stay outdated, which often leads to quality problems and unintended costs. In order to keep the requirements and test documents in a consistent state, we are developing an approach that automatically generates guidance on how to change impacted acceptance tests when changes in requirements occur. In this paper, we briefly present our approach and a prototype tool that implements it. A preliminary evaluation of our approach yielded encouraging results.

Feedback communication channels allow end-users to express their needs, which can be considered in software development and evolution. Although feedback gathering and analysis have been identified as an important topic and several researchers have started their investigation, information is scarce on how software companies currently elicit end-user feedback. In this study, we explore the experiences of software companies with respect to feedback gathering. The results of a case study and online survey indicate two sides of the same coin: on the one hand, most software companies are aware of the relevance of end-user feedback for software evolution and provide feedback channels, which allow end-users to communicate their needs and problems. On the other hand, the quantity and quality of the feedback received varies. We conclude that software companies still do not fully exploit the potential of end-user feedback for software development and evolution.

When designing a new presentation front-end called FlexiView for requirements modeling tools, we encountered a general problem: designing such an interface requires a lot of experimentation which is costly when the code of the tool needs to be adapted for every experiment. On the other hand, when using simplified user interface (UI) tools, the results are difficult to generalize. To improve this situation, we are developing an UI experimentation tool which is based on so called ImitGraphs. ImitGraphs can act as a simple, but accurate substitute for a modeling tool. In this paper, we define requirements for such an UI experimentation tool based on an analysis of the features of existing requirements modeling tools

null

Classifying requirements into functional requirements (FR) and non-functional ones (NFR) is an important task in requirements engineering. However, automated classification of requirements written in natural language is not straightforward, due to the variability of natural language and the absence of a controlled vocabulary. This paper investigates how automated classification of requirements into FR and NFR can be improved and how well several machine learning approaches work in this context. We contribute an approach for preprocessing requirements that standardizes and normalizes requirements before applying classification algorithms. Further, we report on how well several existing machine learning methods perform for automated classification of NFRs into sub-categories such as usability, availability, or performance. Our study is performed on 625 requirements provided by the OpenScience tera-PROMISE repository. We found that our preprocessing improved the performance of an existing classification method. We further found significant differences in the performance of approaches such as Latent Dirichlet Allocation, Biterm Topic Modeling, or Na¨ıve Bayes for the sub-classification of NFRs.

Involving stakeholders in requirements elicitation is a cornerstone of successful requirements engineering (RE). With the recent technological advances, the number of stakeholders of a system has significantly increased. Major stakeholders, end-users in particular, are increasingly difficult to reach, because they may be globally distributed and outside organizational reach, i.e., they are no members of the organizations that are involved in the development of a system. Online elicitation platforms allow to elicit requirements collaboratively from a large number of distributed stakeholders. However, such platforms are not sufficient for motivating stakeholders outside organizational reach to contribute voluntarily. Gamification is a potential means for creating and sustaining such motivation. However, there is little research on stakeholder engagement with gamification so far. Current approaches particularly do not consider that stakeholders learn during elicitation and that their motivational factors also change. In this paper, we address this gap with a motivation concept that is inspired by the theories of experiential learning and need satisfaction. Our contribution is threefold. First, we suggest how to characterize these stakeholders despite not knowing who they are. Second, we show the role of experiential learning and need satisfaction with respect to gamification in the context of requirements elicitation. Third, we present a three-dimensional concept of how to motivate these stakeholders towards requirements elicitation over the whole period of requirements elicitation.

Requirements articulating the needs of stakeholders are critical to successful system development and key to influencing their long-term effects. As the concept of sustainability has entered the discourse of a number of software-related computing fields, so has the term ‘sustainability requirement’. However, it is unclear whether sustainability requirements are and should be different from how we already understand software requirements. This paper presents the results of a corpusassisted discourse analysis study that explored the concept of sustainability requirements in order to understand how the term is being used in software and requirements engineering and related fields. The results of this study reveal that the term ‘sustainability requirement’ is generally used ambiguously and reveals significant segmentation across different fields. Our detailed analysis of selected influential papers highlights the segmented use of the term and suggests key focus questions that need to be addressed to establish a shared operative understanding of the term.

When developing or evolving software systems of non-trivial size, having the requirements properly documented is a crucial success factor. The time and effort required for creating and maintaining non-code artifacts are significantly influenced by the tools with which practitioners view, navigate and edit these artifacts. This is not only true for requirements, but for any artifacts used when developing or evolving systems. However, there is not much evidence about how practitioners actually work with artifacts and how well software tools support them. Therefore, we conducted an exploratory study based on 29 interviews with software practitioners to understand the current practice of presenting and manipulating artifacts in tools, how practitioners deal with the challenges encountered, and how these challenges affect the usability of the tools used. We found that practitioners typically work with several interrelated artifacts concurrently, less than half of these artifacts can be displayed entirely on a large screen, the artifact interrelationship information is often missing, and practitioners work collaboratively on artifacts without sufficient support. We identify the existing challenges of working with artifacts and discuss existing solutions proposed addressing them. Our results contribute to the body of knowledge about how practitioners work with artifacts when developing or evolving software, the challenges they are faced with, and the attempts to address these challenges.

Due to the increasing use of general purpose graphical models (e.g., UML diagrams) and domain specific graphical models in different stages of software development, software engineers who work with these models spend more time interacting with modeling tools. Thus, the usability of interaction techniques of modeling tools affects the overall productivity of software development. Tool developers and user interface designers rely on the feedback from usability tests to optimize the user interface of tools that provide a graphical editor. Developing a working prototype to test new techniques is costly due to the complexity and variety of graphical models. This results in either tests at the late stages of development when changes are more expensive, or tests with prototypes that only support a subset of the intended graphical models, thus not comprehensive. In order to simplify conducting usability tests, instead of using the intended graphical models in the tests, we propose to use simpler models that require similar interactions when being manipulated. For this purpose, we introduced ImitGraphs, an extended graph with additional properties so that it can be specialized in interacting similarly to an intended graphical model. Then, we designed a method to instruct test participants to create ImitGraphs. Specialized graphs enable us to develop prototypes for usability tests faster and consequently cheaper resulting in more usability tests at early stages of tool development and on a wider range of intended models.

null

It is essential to consider user need and expectations in order for software to remain relevant and useful throughout its evolution. However, users can provide feedback explicitly through a dedicated channel, or implicitly by discussing software with other users. In addition, users use different languages to express their needs. The variety of channels and the different languages used pose a challenge in understanding and capturing the whole picture of expressed user needs. In this thesis, we present an approach to automatically identify similar user feedback across different languages and from multiple channels. At the core of the approach is a word aligner that aligns words based on their semantic similarity and the similarity between their local semantic contexts. Additionally, we make use of Sentiment Analysis and Text Classification and compute a similarity score by means of a weighted function. We evaluate our approach on user feedback from five different mobile applications obtained from application distribution platforms (i.e. app stores) and social networking sites. The obtained results are encouraging. Compared to human assessment, the overall performance for monolingual user feedback pairs yielded a correlation of 79.38%, whereas for cross-lingual pairs the correlation was 77.95%.

Gauging usersí sentiment towards a product or feature is an integral part of software evolution, since feedback can be used to find faults or make improvements. Due to the large volumes of user feedback, it is not feasible to manually evaluate each review. Various approaches have been developed to automatically analyze the sentiment in texts. We compare the lexical approach SentiStrength, the statistical machine learning approaches NaÔve Bayes and Support Vector Machines, and the deep learning approach Recursive Neural Tensor Networks in terms of quality. Furthermore, we investigate strengths and weaknesses of each approach, as well as training data requirements, performance and transferability between data of different domains. We use a dataset of app reviews, which was created and annotated during this thesis, and found that when applied to this dataset, no approach clearly outperforms the others.

Crowd-based requirements engineering (CrowdRE) could significantly change RE. Performing RE activities such as elicitation with the crowd of stakeholders turns RE into a participatory effort, leads to more accurate requirements, and ultimately boosts software quality. Although any stakeholder in the crowd can contribute, CrowdRE emphasizes one stakeholder group whose role is often trivialized: users. CrowdRE empowers the management of requirements, such as their prioritization and segmentation, in a dynamic, evolved style through collecting and harnessing a continuous flow of user feedback and monitoring data on the usage context. To analyze the large amount of data obtained from the crowd, automated approaches are key. This article presents current research topics in CrowdRE; discusses the benefits, challenges, and lessons learned from projects and experiments; and assesses how to apply the methods and tools in industrial contexts. This article is part of a special issue on Crowdsourcing for Software Engineering.