João Pedro Santos

VML* – a family of languages for variability management in software product lines

Managing variability is a challenging issue in software-product-line engineering. A key part of variability management is the ability to express explicitly the relationship between variability models (expressing the variability in the problem space, for example using feature models) and other artefacts of the product line, for example, requirements models and architecture models. Once these relations have been made explicit, they can be used for a number of purposes, most importantly for product derivation, but also for the generation of trace links or for checking the consistency of a product-line architecture. This paper bootstraps techniques from product-line engineering to produce a family of languages for variability management for easing the creation of new members of the family of languages. We show that developing such language families is feasible and demonstrate the flexibility of our language family by applying it to the development of two variability-management languages.

Relating feature models to other models of a software product line: a comparative study of featuremapper and VML

Software product lines using feature models often require the relation between feature models in problem space and the models used to describe the details of the product line to be expressed explicitly. This is particularly important, where automatic product derivation is required. Different approaches for modelling this mapping have been proposed in the literature. However, a discussion of their relative benefits and drawbacks is currently missing. As a first step towards a better understanding of this field, this paper applies two of these approaches-- FeatureMapper as a representative of declarative approaches and VML* as a representative of operational approaches--to the case study. We show in detail how the case study can be expressed using these approaches and discuss strengths and weaknesses of the two approaches with regard to the case study.

Multi-view composition language for software product line requirements

Composition of requirements models in Software Product Line (SPL) development enables stakeholders to derive the requirements of target software products and, very important, to reason about them. Given the growing complexity of SPL development and the various stakeholders involved, their requirements are often specified from heterogeneous, partial views. However, existing requirements composition languages are very limited to generate specific requirements views for SPL products. They do not provide specialized composition rules for referencing and composing elements in recurring requirements models, such as use cases and activity models. This paper presents a multi-view composition language for SPL requirements, the Variability Modeling Language for Requirements (VML4RE). This language describes how requirements elements expressed in different models should be composed to generate a specific SPL product. The use of VML4RE is illustrated with UML-based requirements models defined for a home automation SPL case study. The language is evaluated with additional case studies from different application domains, such as mobile phones and sales management.

Generating Requirements Analysis Models from Textual Requirements

Use case modeling is a commonly used technique to describe functional requirements in requirements engineering. Typically, use cases are captured from textual requirements documents describing the functionalities the system should meet. Requirements elicitation, analysis and modeling is a time consuming and error-prone activity, which it is not usually supported by automated tools. This paper tackles this problem by taking free-form textual requirements and offering a semi-automatic process for generation of domain models, such as use cases. Our goal is twofold: (i) reduce the time spent to produce requirements artifacts; and (ii) enable future application of model-driven engineering techniques to maintain traceability information and consistency between textual and requirements visual models artifacts.

Streamlining scenario modeling with Model-Driven Development: A case study

Scenario modeling can be realized through different perspectives. In UML, scenarios are often modeled with activity models, in an early stage of development. Later, sequence diagrams are used to detail object interactions. The migration from activity diagrams to sequence diagrams is a repetitive and error-prone task. Model-Driven Development (MDD) can help streamlining this process, through transformation rules. Since the information in the activity model is insufficient to generate the corresponding complete sequence model, manual refinements are required. Our goal is to compare the relative effort of building the sequence diagrams manually with that of building them semi-automatically. Our results show a decrease in the number of operations required to build and refine the sequence model of approximately 64% when using MDD, when compared to the manual approach.

Ranking Linked-Entities in a Sentiment Graph

Reputation analysis is naturally associated to a sentiment analysis task of the targeted named-entities. This analysis leverages on a sentiment lexicon that includes general sentiment words that characterize the general sentiment towards the targeted named-entity. However, in most cases, target entities are themselves part of the sentiment lexicon, creating a loop from which it is difficult to infer an entity reputation. Sometimes, the entity became a reference in the domain and is vastly cited as an example of a highly reputable entity. For example, in the movies domain it is not uncommon to see reviews citing Batman or Anthony Hopkins as esteemed references. In this paper we describe a three-step procedure to perform reputation analysis of linked entities. First, our method jointly extracts named entities reputation and a domain specific sentiment lexicon. Second, an entities graph is created by analyzing cross-citations in subjective sentences. Third, the entities reputation are updated through an iterative optimization that exploits the graph of the linked-entities. The proposed approach closely models real-world domains, where domain specific jargon is common and entities are so popular that they become widely used as sentiment references. The evaluation on a graph with 12,687 vertices, of which 3,177 are linked entities and 9,510 are sentiment words, shows that our approach can improve the correct detection of an entitys reputation.

Reputation analysis with a ranked sentiment-lexicon

Reputation analysis is naturally linked to a sentiment analysis task of the targeted entities. This analysis leverages on a sentiment lexicon that includes general sentiment words and domain specific jargon. However, in most cases target entities are themselves part of the sentiment lexicon, creating a loop from which it is difficult to infer an entity reputation. Sometimes, the entity became a reference in the domain and is vastly cited as an example of a highly reputable entity. For example, in the movies domain it is not uncommon to see reviews citing Batman or Anthony Hopkins as esteemed references. In this paper we describe an unsupervised method for performing a simultaneous-analysis of the reputation of multiple named-entities. Our method jointly extracts named entities reputation and a domain specific sentiment lexicon. The objective is two-fold: (1) named-entities are naturally ranked by our method and (2) we can build a reputation graph of the domains named entities. This framework has immediate applications in terms of visualization or search by reputation.

Increasing Quality in Scenario Modelling with Model-Driven Development

Models, with different levels of detail, share similar abstractions that can be reused by means of model-driven techniques such as transformations. For example, scenarios are a well-known technique in requirements engineering to represent behavioral flows in a software system. When using UML, scenarios are typically represented with activity models in the early stages of software development, while sequence models are used to describe more detailed object interactions as modeling progresses. This paper defines transformation rules to automate the migration from activity to sequence models. We present a case study illustrating the application of our transformation rules. Our preliminary assessment of the impact of the benefits of using these transformations points to: (i) a reduction of around 50% in the effort building sequence models, (ii) increased trace ability among models, and (iii) error prevention when migrating from different scenario notations.

Improving Cold-Start Recommendations with Social-Media Trends and Reputations.

In recommender systems, the cold-start problem is a common challenge. When a new item has no ratings, it becomes difficult to relate it to other items or users. In this paper, we address the cold-start problem and propose to leverage on social-media trends and reputations to improve the recommendation of new items. The proposed framework models the long-term reputation of actors and directors, to better characterize new movies. In addition, movies popularity are deduced from social-media trends that are related to the corresponding new movie. A principled method is then applied to infer cold-start recommendations from these social-media signals. Experiments on a realistic time-frame, covering several movie-awards events between January 2014 and March 2014, showed significant improvements over ratings-only and metadata-only based recommendations.

Analysing the Footprint of Classifiers in Overlapped and Imbalanced Contexts

It is recognised that the imbalanced data problem is aggravated by other difficulty factors, such as class overlap. Over the years, several research works have focused on this problematic, although presenting two major hitches: the limitation of test domains and the lack of a formulation of the overlap degree, which makes results hard to generalise. This work studies the performance degradation of classifiers with distinct learning biases in overlap and imbalanced contexts, focusing on the characteristics of the test domains (shape, dimensionality and imbalance ratio) and on to what extent our proposed overlapping measure (degOver) is aligned with the performance results observed. Our results show that MLP and CART classifiers are the most robust to high levels of class overlap, even for complex domains, and that KNN and linear SVM are the most aligned with degOver. Furthermore, we found that the dimensionality of data also plays an important role in explaining performance results.