Abstract

We study an important task of attacking natural language processing models in a black box setting. We propose an attack strategy that crafts semantically similar adversarial examples on text classification and entailment tasks. Our proposed attack finds candidate words by considering the information of both the original word and its surrounding context. It jointly leverages masked language modelling and next sentence prediction for context understanding. In comparison to attacks proposed in prior literature, we are able to generate high quality adversarial examples that do significantly better both in terms of success rate and word perturbation percentage.

Abstract

Team Recommendation has always been a challenging aspect in team sports. Such systems aim to recommend a player combination best suited against the opposition players, resulting in an optimal outcome. In this paper, we propose a semi-supervised statistical approach to build a team recommendation system for cricket by modelling players into embeddings. To build these embeddings, we design a qualitative and quantitative rating system which considers the strength of opposition also for evaluating player performance. The embeddings obtained, describes the strengths and weaknesses of the players based on past performances of the player. We also embark on a critical aspect of team composition, which includes the number of batsmen and bowlers in the team. The team composition changes over time, depending on different factors which are tough to predict, so we take this input from the user and use the player embeddings to decide the best possible team combination with the given team composition.

Abstract

Load identification is an essential step in Non-Intrusive Load Monitoring (NILM), a process of estimating the power consumption of individual appliances using only whole-house aggregate consumption. Such estimates can help consumers and utility companies improve load management and save power. Current state-of-the-art methods for load identification generally use either steady state or transient features for load identification. We hypothesize that these are complementary features and so a hybrid combination of them will result in an improved appliance signature. We propose a novel hybrid combination that has the advantage of being low-dimensional and can thus be easily integrated with existing classification models to improve load identification. Our improved hybrid features are then used for building appliance identification models using Naive Bayes, KNN, Decision Tree and Random Forest classifiers. The proposed NILM methodology is evaluated for robustness in changing environments. An automated data collection setup is established to capture 7 home appliances aggregate data under varying voltages. Experimental results show that our proposed feature fusion based algorithms are more robust and outperform steady state and transient feature-based algorithms by at least +9% and +15% respectively.

Abstract

Within the social system of science, citation practices characterize social functions like the conferral of recognition upon the work of others as well as the acknowledgement of one’s intellectual debt. However, the structure of intellectual influence is misrepresented when only the immediate citations and their cardinality are taken into consideration. Thus, in order to better understand the associative dissemination of influence and approximately construe the anatomy of this structure, complex interactions in the convoluted network of authors and papers need to be probed. Our study aims at understanding these heterogeneous complex interactions. For the bibliographic dataset of authors and publications, we define proxy scores that attempt to determine the associative influence of the cited author over the citing author. In order to harness structural connectivity of the network, we generate author vector representations using these influence scores. Furthermore, with a view to assess the competence of our proposed scores, we evaluate these representations and provide an empirical study of the results obtained with our algorithm against the baseline and also present a qualitative analysis.

Abstract

Variational autoencoders were proven successful in domains such as computer vision and speech processing. Their adoption for modeling user preferences is still unexplored, although recently it is starting to gain attention in the current literature. In this work, we propose a model which extends variational autoencoders by exploiting the rich information present in the past preference history. We introduce a recurrent version of the VAE, where instead of passing a subset of the whole history regardless of temporal dependencies, we rather pass the consumption sequence subset through a recurrent neural network. At each time-step of the RNN, the sequence is fed through a series of fully-connected layers, the output of which models the probability distribution of the most likely future preferences. We show that handling temporal information is crucial for improving the accuracy of the VAE: In fact, our model beats the current state-of-the-art by valuable margins because of its ability to capture temporal dependencies among the user-consumption sequence using the recurrent encoder still keeping the fundamentals of variational autoencoders intact.

Abstract

Clustering has become a very popular machine learning technique for identifying groups of data points with common features in a set of data points. In several applications, there is a need to explain the clusters so that the user can understand the underlying commonalities. One such application is in the area of building energy simulation. There is a need to cluster solutions obtained by parametric energy simulation runs and explain the characteristics of each cluster for human consumption. This paper demonstrates how the axis-aligned hyper-rectangles based clustering, on building energy simulation data, can help identify clusters and describe the governing rules for each cluster. We are calling these rules design strategies. Instead of the distance-based clustering methods that are unable to extract simple rules from the underlying commonalities in each cluster, this method is able to overcome this limitation. This method is applied to identify design strategies from a parametric run of a simple five-zone rectangular building model. Based on a user-given threshold, low energy solutions are selected for clustering. Each axis-aligned hyper-rectangle cluster is a unique design strategy that can be easily communicated to the user.

Abstract

As scientific communities grow and evolve, there is a high demand for improved methods for finding relevant papers, comparing papers on similar topics and studying trends in the research community. All these tasks involve the common problem of extracting structured information from scientific articles. In this paper, we propose a novel, scalable, semi-supervised method for extracting relevant structured information from the vast available raw scientific literature. We extract the fundamental concepts of “aim”,” method” and “result” from scientific articles and use them to construct a knowledge graph. Our algorithm makes use of domain-based word embedding and the bootstrap framework. Our experiments show that our system achieves precision and recall comparable to the state of the art. We also show the domain independence of our algorithm by analyzing the research trends of two distinct communities-computational linguistics and computer vision.

Abstract

Recommender systems are a key component of music sharing platforms, which suggest musical recordings a user might like. People often have implicit preferences while listening to music, though these preferences might not always be the same while they listen to music at different times. For example, a user might be interested in listening to songs of only a particular artist at some time, and the same user might be interested in the top-rated songs of a genre at another time. In this paper we try to explicitly model the short term preferences of the user with the help of Last.fm tags of the songs the user has listened to. With a session defined as a period of activity surrounded by periods of inactivity, we introduce the concept of a subsession, which is that part of the session wherein the preference of the user does not change much. We assume the user preference might change within a session and a session might have multiple subsessions. We use our modelling of the user preferences to generate recommendations for the next song the user might listen to. Experiments on the user listening histories taken from Last.fm indicate that this approach beats the present methodologies in predicting the next recording a user might listen to.

Abstract

Recommender Systems are an integral part of music sharing platforms. Often the aim of these systems is to increase the time, the user spends on the platform and hence having a high commercial value. The systems which aim at increasing the average time a user spends on the platform often need to recommend songs which the user might want to listen to next at each point in time. This is different from recommendation systems which try to predict the item which might be of interest to the user at some point in the user lifetime but not necessarily in the very near future. Prediction of next song the user might like requires some kind of modeling of the user interests at the given point of time. Attentive neural networks have been exploiting the sequence in which the items were selected by the user to model the implicit short-term interests of the user for the task of next item prediction, however we feel that features of the songs occurring in the sequence could also convey some important information about the short-term user interest which only the items cannot. In this direction we propose a novel attentive neural architecture which in addition to the sequence of items selected by the user, uses the features of these items to better learn the user short-term preferences and recommend next song to the user.

Abstract

The purpose of this study is to examine the differences between different types of news stories. Given the huge impact of social networks, online content plays an important role in forming or changing the opinions of people. Unlike traditional journalism where only certain news organizations can publish content, online journalism has given chance even for individuals to publish. This has its own advantages like individual empowerment but has given a chance to a lot of malicious entities to spread misinformation for their own benefit. As reported by many organizations in recent history, this even has influence on major events like the outcome of elections. Therefore, it is of great importance now, to have some sort of automated classification of news stories. In this work, we propose a deep hierarchical attention neural architecture combining sentiment and semantic embeddings for more accurate fine grained classification of news stories. Experimental results show that the sentiment embedding along with semantic information outperform several state-of-the art methods in this task.