Studying software evolution using topic models

Topic models are generative probabilistic models which have been applied to information retrieval to automatically organize and provide structure to a text corpus. Topic models discover topics in the corpus, which represent real world concepts by frequently co-occurring words. Recently, researchers found topics to be effective tools for structuring various software artifacts, such as source code, requirements documents, and bug reports. This research also hypothesized that using topics to describe the evolution of software repositories could be useful for maintenance and understanding tasks. However, research has yet to determine whether these automatically discovered topic evolutions describe the evolution of source code in a way that is relevant or meaningful to project stakeholders, and thus it is not clear whether topic models are a suitable tool for this task.In this paper, we take a first step towards evaluating topic models in the analysis of software evolution by performing a detailed manual analysis on the source code histories of two well-known and well-documented systems, JHotDraw and jEdit. We define and compute various metrics on the discovered topic evolutions and manually investigate how and why the metrics evolve over time. We find that the large majority (87%–89%) of topic evolutions correspond well with actual code change activities by developers. We are thus encouraged to use topic models as tools for studying the evolution of a software system.     

Static test case prioritization using topic models

Software development teams use test suites to test changes to their source code. In many situations, the test suites are so large that executing every test for every source code change is infeasible, due to time and resource constraints. Development teams need to prioritize their test suite so that as many distinct faults as possible are detected early in the execution of the test suite. We consider the problem of static black-box test case prioritization (TCP), where test suites are prioritized without the availability of the source code of the system under test (SUT). We propose a new static black-box TCP technique that represents test cases using a previously unused data source in the test suite: the linguistic data of the test cases, i.e., their identifier names, comments, and string literals. Our technique applies a text analysis algorithm called topic modeling to the linguistic data to approximate the functionality of each test case, allowing our technique to give high priority to test cases that test different functionalities of the SUT. We compare our proposed technique with existing static black-box TCP techniques in a case study of multiple real-world open source systems: several versions of Apache Ant and Apache Derby. We find that our static black-box TCP technique outperforms existing static black-box TCP techniques, and has comparable or better performance than two existing execution-based TCP techniques. Static black-box TCP methods are widely applicable because the only input they require is the source code of the test cases themselves. This contrasts with other TCP techniques which require access to the SUT runtime behavior, to the SUT specification models, or to the SUT source code.     

Modeling topic control to detect influence in conversations using nonparametric topic models

Identifying influential speakers in multi-party conversations has been the focus of research in communication, sociology, and psychology for decades. It has been long acknowledged qualitatively that controlling the topic of a conversation is a sign of influence. To capture who introduces new topics in conversations, we introduce SITS—Speaker Identity for Topic Segmentation—a nonparametric hierarchical Bayesian model that is capable of discovering (1) the topics used in a set of conversations, (2) how these topics are shared across conversations, (3) when these topics change during conversations, and (4) a speaker-specific measure of “topic control”. We validate the model via evaluations using multiple datasets, including work meetings, online discussions, and political debates. Experimental results confirm the effectiveness of SITS in both intrinsic and extrinsic evaluations.     

A survey on the use of topic models when mining software repositories

Researchers in software engineering have attempted to improve software development by mining and analyzing software repositories. Since the majority of the software engineering data is unstructured, researchers have applied Information Retrieval (IR) techniques to help software development. The recent advances of IR, especially statistical topic models, have helped make sense of unstructured data in software repositories even more. However, even though there are hundreds of studies on applying topic models to software repositories, there is no study that shows how the models are used in the software engineering research community, and which software engineering tasks are being supported through topic models. Moreover, since the performance of these topic models is directly related to the model parameters and usage, knowing how researchers use the topic models may also help future studies make optimal use of such models. Thus, we surveyed 167 articles from the software engineering literature that make use of topic models. We find that i) most studies centre around a limited number of software engineering tasks; ii) most studies use only basic topic models; iii) and researchers usually treat topic models as black boxes without fully exploring their underlying assumptions and parameter values. Our paper provides a starting point for new researchers who are interested in using topic models, and may help new researchers and practitioners determine how to best apply topic models to a particular software engineering task.     

Estimating software readiness using predictive models

In this study, defect tracking is used as a proxy method to predict software readiness. The number of remaining defects in an application under development is one of the most important factors that allow one to decide if a piece of software is ready to be released. By comparing predicted number of faults and number of faults discovered in testing, software manager can decide whether the software is likely ready to be released or not.The predictive model developed in this research can predict: (i) the number of faults (defects) likely to exist, (ii) the estimated number of code changes required to correct a fault and (iii) the estimated amount of time (in minutes) needed to make the changes in respective classes of the application. The model uses product metrics as independent variables to do predictions. These metrics are selected depending on the nature of source code with regards to architecture layers, types of faults and contribution factors of these metrics. The use of neural network model with genetic training strategy is introduced to improve prediction results for estimating software readiness in this study. This genetic-net combines a genetic algorithm with a statistical estimator to produce a model which also shows the usefulness of inputs.The model is divided into three parts: (1) prediction model for presentation logic tier (2) prediction model for business tier and (3) prediction model for data access tier. Existing object-oriented metrics and complexity software metrics are used in the business tier prediction model. New sets of metrics have been proposed for the presentation logic tier and data access tier. These metrics are validated using data extracted from real world applications. The trained models can be used as tools to assist software mangers in making software release decisions.     

Studying just-in-time defect prediction using cross-project models

Unlike traditional defect prediction models that identify defect-prone modules, Just-In-Time (JIT) defect prediction models identify defect-inducing changes. As such, JIT defect models can provide earlier feedback for developers, while design decisions are still fresh in their minds. Unfortunately, similar to traditional defect models, JIT models require a large amount of training data, which is not available when projects are in initial development phases. To address this limitation in traditional defect prediction, prior work has proposed cross-project models, i.e., models learned from other projects with sufficient history. However, cross-project models have not yet been explored in the context of JIT prediction. Therefore, in this study, we empirically evaluate the performance of JIT models in a cross-project context. Through an empirical study on 11 open source projects, we find that while JIT models rarely perform well in a cross-project context, their performance tends to improve when using approaches that: (1) select models trained using other projects that are similar to the testing project, (2) combine the data of several other projects to produce a larger pool of training data, and (3) combine the models of several other projects to produce an ensemble model. Our findings empirically confirm that JIT models learned using other projects are a viable solution for projects with limited historical data. However, JIT models tend to perform best in a cross-project context when the data used to learn them are carefully selected.     

Studying software evolution using artefacts’ shared information content

In order to study software evolution, it is necessary to measure artefacts representative of project releases. If we consider the process of software evolution to be copying with subsequent modification, then, by analogy, placing emphasis on what remains the same between releases will lead to focusing on similarity between artefacts. At the same time, software artefacts-stored digitally as binary strings-are all information. This paper introduces a new method for measuring software evolution in terms of artefacts’ shared information content. A similarity value representing the quantity of information shared between artefact pairs is produced using a calculation based on Kolmogorov complexity. Similarity values for releases are then collated over the software’s evolution to form a map quantifying change through lack of similarity. The method has general applicability: it can disregard otherwise salient software features such as programming paradigm, language or application domain because it considers software artefacts purely in terms of the mathematically justified concept of information content. Three open-source projects are analysed to show the method’s utility. Preliminary experiments on udev and git verify the measurement of the projects’ evolutions. An experiment on ArgoUML validates the measured evolution against experimental data from other studies.     

Prediction of software reliability using connectionist models

The usefulness of connectionist models for software reliability growth prediction is illustrated. The applicability of the connectionist approach is explored using various network models, training regimes, and data representation methods. An empirical comparison is made between this approach and five well-known software reliability growth models using actual data sets from several different software projects. The results presented suggest that connectionist models may adapt well across different data sets and exhibit a better predictive accuracy. The analysis shows that the connectionist approach is capable of developing models of varying complexity     

Solving models with inequalities using standard econometric software

Simultaneous econometric models may contain pairs of complementary inequalities. It is discussed how to reformulate such models and solve them with econometric software which can handle only equalities. Two approaches are applied: the normal map representation and the Fischer-Burmeister NCP function. The latter seems to work best. The software programs TSP, SAS/ETS and EViews are tested. The test model describes two markets for electricity, each with fluctuating demand and an endogenous production capacity; the capacity of the trade link between the regions is also endogenous.     

Syndromic surveillance of Flu on Twitter using weakly supervised temporal topic models

Surveillance of epidemic outbreaks and spread from social media is an important tool for governments and public health authorities. Machine learning techniques for nowcasting the Flu have made significant inroads into correlating social media trends to case counts and prevalence of epidemics in a population. There is a disconnect between data-driven methods for forecasting Flu incidence and epidemiological models that adopt a state based understanding of transitions, that can lead to sub-optimal predictions. Furthermore, models for epidemiological activity and social activity like on Twitter predict different shapes and have important differences. In this paper, we propose two temporal topic models (one unsupervised model as well as one improved weakly-supervised model) to capture hidden states of a user from his tweets and aggregate states in a geographical region for better estimation of trends. We show that our approaches help fill the gap between phenomenological methods for disease surveillance and epidemiological models. We validate our approaches by modeling the Flu using Twitter in multiple countries of South America. We demonstrate that our models can consistently outperform plain vocabulary assessment in Flu case-count predictions, and at the same time get better Flu-peak predictions than competitors. We also show that our fine-grained modeling can reconcile some contrasting behaviors between epidemiological and social models.     

Unifying design and runtime software adaptation using aspect models

Software systems are seen more and more as evolutive systems. At the design phase, software is constantly in adaptation by the building process itself, and at runtime, it can be adapted in response to changing conditions in the executing environment such as location or resources. Adaptation is generally difficult to specify because of its cross-cutting impact on software. This article introduces an approach to unify adaptation at design and at runtime based on Aspect Oriented Modeling. Our approach proposes a unified aspect metamodel and a platform that realizes two different weaving processes to achieve design and runtime adaptations. This approach is used in a Dynamic Software Product Line which derives products that can be configured at design time and adapted at runtime in order to dynamically fit new requirements or resource changes. Such products are implemented using the Service Component Architecture and Java. Finally, we illustrate the use of our approach based on an adaptive e-shopping scenario. The main advantages of this unification are: a clear separation of concerns, the self-contained aspect model that can be weaved during the design and execution, and the platform independence guaranteed by two different types of weaving.     

Using topic models for OCR correction

Despite several decades of research in document analysis, recognition of unconstrained handwritten documents is still considered a challenging task. Previous research in this area has shown that word recognizers perform adequately on constrained handwritten documents which typically use a restricted vocabulary (lexicon). But in the case of unconstrained handwritten documents, state-of-the-art word recognition accuracy is still below the acceptable limits. The objective of this research is to improve word recognition accuracy on unconstrained handwritten documents by applying a post-processing or OCR correction technique to the word recognition output. In this paper, we present two different methods for this purpose. First, we describe a lexicon reduction-based method by topic categorization of handwritten documents which is used to generate smaller topic-specific lexicons for improving the recognition accuracy. Second, we describe a method which uses topic-specific language models and a maximum-entropy based topic categorization model to refine the recognition output. We present the relative merits of each of these methods and report results on the publicly available IAM database.     

Unifying design and runtime software adaptation using aspect models

Software systems are seen more and more as evolutive systems. At the design phase, software is constantly in adaptation by the building process itself, and at runtime, it can be adapted in response to changing conditions in the executing environment such as location or resources. Adaptation is generally difficult to specify because of its cross-cutting impact on software. This article introduces an approach to unify adaptation at design and at runtime based on Aspect Oriented Modeling. Our approach proposes a unified aspect metamodel and a platform that realizes two different weaving processes to achieve design and runtime adaptations. This approach is used in a Dynamic Software Product Line which derives products that can be configured at design time and adapted at runtime in order to dynamically fit new requirements or resource changes. Such products are implemented using the Service Component Architecture and Java. Finally, we illustrate the use of our approach based on an adaptive e-shopping scenario. The main advantages of this unification are: a clear separation of concerns, the self-contained aspect model that can be weaved during the design and execution, and the platform independence guaranteed by two different types of weaving.     

Probabilistic topic models for sequence data

Probabilistic topic models are widely used in different contexts to uncover the hidden structure in large text corpora. One of the main (and perhaps strong) assumption of these models is that generative process follows a bag-of-words assumption, i.e. each token is independent from the previous one. We extend the popular Latent Dirichlet Allocation model by exploiting three different conditional Markovian assumptions: (i) the token generation depends on the current topic and on the previous token; (ii) the topic associated with each observation depends on topic associated with the previous one; (iii) the token generation depends on the current and previous topic. For each of these modeling assumptions we present a Gibbs Sampling procedure for parameter estimation. Experimental evaluation over real-word data shows the performance advantages, in terms of recall and precision, of the sequence-modeling approaches.     

Robust supervised topic models under label noise

Machine Learning - Recently, some statistical topic modeling approaches have been widely applied in the field of supervised document classification. However, there are few researches on these...     

Statistical topic models for multi-label document classification

Machine learning approaches to multi-label document classification have to date largely relied on discriminative modeling techniques such as support vector machines. A?drawback of these approaches is that performance rapidly drops off as the total number of labels and the number of labels per document increase. This problem is amplified when the label frequencies exhibit the type of highly skewed distributions that are often observed in real-world datasets. In this paper we investigate a class of generative statistical topic models for multi-label documents that associate individual word tokens with different labels. We investigate the advantages of this approach relative to discriminative models, particularly with respect to classification problems involving large numbers of relatively rare labels. We compare the performance of generative and discriminative approaches on document labeling tasks ranging from datasets with several thousand labels to datasets with tens of labels. The experimental results indicate that probabilistic generative models can achieve competitive multi-label classification performance compared to discriminative methods, and have advantages for datasets with many labels and skewed label frequencies.