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

Ensemble POS taggers are a good choice to integrate and leverage benefits of various types of POS taggers. This can help the large number (6500+) of resource-poor languages which do not have much annotated training data by providing ways to integrate semi-supervised/unsupervised taggers with supervised taggers. In this paper we present our experiments of developing ensemble POS taggers using a decision tree. We integrate a semi-supervised data mining approach that uses context based lists (CBLs) for POS tagging with supervised (1) Support Vector Machine based POS tagger, called SVMTool and (2) Conditional Random Field based POS tagger. The results are enhanced semi-supervised ensemble POS taggers which outperform the base methods. In these POS taggers, we use a decision tree to decide when to rely on the output of supervised tagger, and when to rely on the semi-supervised CBL method. The CBL based tagger uses rich contextual information which helps in tagging both existing and unseen words and uses no domain knowledge while supervised taggers give good performance for words present in the training model and can include domain based features. Hence, these algorithms have complementary strengths and in our ensemble we are able to combine these strengths. Enhanced performance of our new POS taggers over the base methods suggests that integrating these methods combines the qualities of these in the new tagger which enhances the performance. Therefore, these new semi-supervised ensemble taggers are more suitable for resource-poor languages.

Abstract

Predicting the outcome of a match has always been at the center of sports analytics. Indian Premier League (IPL), a professional Twenty20 (T20) cricket league in India, has established itself as one of the biggest tournaments in cricket history. In this paper, we propose a model to predict the winner at the end of each over in the second innings of an IPL cricket match. Our methodology not only incorporates the dynamically updating game context as the game progresses, but also includes the relative strength between the two teams playing the match. Estimating the relative strength between two teams involves modeling the individual participating players’ potentials. To model a player, we use his career as well as recent performance statistics. Using the various dynamic features, we evaluate several supervised learning algorithms to predict the winner of the match. Finally, using the Random Forest Classifier (RFC), we have achieved an accuracy of 65.79% - 84.15% over the course of second innings, with an overall accuracy of 75.68%.

Abstract

Energy utilization can be improved by precise plug load monitoring and control. Plug load energy consumption is nearly 30% of the total building energy consumption. Therefore, plug load identification is a key requirement for energy conservation in buildings. Intrusive load monitoring techniques identify loads precisely but have not been tested widely so far for their performance in changing operating conditions. Hence, the present research proposes a robust low frequency intrusive load monitoring technique to identify load accurately. A smart power strip using proposed load identification technique is designed and developed. Linear regression is applied on the acquired data to capture the behavioral trends of a particular device more explicitly and concisely. Further, weighted K-NN classifier is applied on the transformed data set for device. Experimental results show that the proposed algorithm performs better than the standard classifiers, and can offer tangible savings.

Abstract

Extracting structured information from unstructured text is a critical problem. Over the past few years, various clustering algorithms have been proposed to solve this problem. In addition, various algorithms based on probabilistic topic models have been developed to find the hidden thematic structure from various corpora (i.e publications, blog setc). Both types of algorithms have been transferred to the domain of scientific literature to extract structured information to solve problems like data exploration, expert detection etc. In order to remain domain-agnostic, these algorithms do not exploit the structure present in a scientific publication. Majority of researchers interpret a scientific publication as research conducted to report progress in solving some research problems.Following this interpretation, in this paper we present a different outlook to the same problem by modelling scientific publications around research problems. By associating a scientific publication with a research problem, exploring the scientific literature becomes more intuitive. In this paper, we propose an unsupervised framework to mine research problems from titles and abstracts of scientific literature. Our framework uses weighted frequent phrase mining to generate phrases and filters them to obtain high-quality phrases.These high-quality phrases are then used to segment the scientific publication into meaningful semantic units. After segmenting publications, we apply a number of heuristics to score the phrases and sentences to identify the research problems. In a post-processing step we use a neighborhood based algorithm to merge different representations of the same problems. Experiments conducted on parts of DBLP dataset show promising results.

Abstract

Cricket is one of the most popular team sports in the world.Players have multiple roles in a game of cricket, predominantly as bats-men and bowlers. Over the generations, statistics such as batting and bowling averages, and strike and economy rates have been used to judge the performance of individual players. These measures, however, do not take into consideration the context of the game in which a player per-formed. Furthermore, these types of statistics are incapable of comparing the performance of players across different roles. In this paper, we present an approach to quantitatively assess the performances of individual players in single match of One Day International (ODI) cricket. For this, we have developed a new measure, called the Work Index, which represents the amount of work that is yet to be done by a team to achieve its tar-get. Our approach incorporates game situations and the team strength sto measure the player contributions. This not only helps us in evaluating the individual performances, but also enables us to compare players within and across various roles on a common scale. Using the player performances in a match, we predict the player of the match award for the ODI matches played between 2006 and 2016. We have achieved an accuracy of 86.80% for the top-3 positions, which is superior to base linemodels and previous works, to the best of our knowledge.

Abstract

We present a regression-based feature learning algorithm that generates new features from a set of available features (raw data points). Being data-driven, it requires no domain knowledge and is hence generic. Such a representation is learnt by mining pairwise feature associations, identifying the linear or non-linear relationship between each pair, applying regression and selecting those relationships that are stable.Our experimental evaluation on 20 datasets taken from UC Irvine and Gene Expression, across different domains, provides evidence that the features learnt through our model can improve the overall prediction accuracy, substantially, over the original feature space across 8 different classifiers without any domain knowledge.

Abstract

The problem of ordering documents in a large collection into a sequence that is efficient for learning (both human and machine) is of high practical significance,but has not yet been well-formulated. We formulate this problem as mining a maximally informative simple sequence of documents. The mined sequence should be maximally informative in the sense that the reader learns quickly by reading only a few documents, and its should be simple so that the reader is not over whelmed while trying to learn the content. The task can be posed as: Given that a reader wishes to read (at most)k documents, which documents should be selected from the repository and in what order, so as to provide maximum information. We present the WikiSeq algorithm for this purpose. We also design a metric based on information-gain to help objectively evaluate WikiSeq, and conduct experiments to compare with indicative baselines. Finally, we provide case-studies to subjectively illustrate WikiSeq’s merits.

Abstract

We present Paper2vec, a novel neural network embedding based approach for creating scientific paper representations which make use of both textual and graph-based information. An academic citation network can be viewed as a graph where individual nodes contain rich textual information. With the current trend of open-access to most scientific literature, we presume that this full text of a scientific article contain vital source of information which aids in various recommendation and prediction tasks concerning this domain. To this end, we propose an approach, Paper2vec, which comprises of information from both the modalities and results in a rich representation for scientific papers. Over the recent past representation learning techniques have been studied extensively using neural networks. However, they are modeled independently for text and graph data. Paper2vec leverages recent research in the broader field of unsupervised feature learning from both graphs and text documents. We demonstrate the efficacy of our representations on three real world academic datasets in two tasks - node classification and link prediction where Paper2vec is able to outperform state-of-the-art by a considerable margin.

Abstract

In this work we demonstrate a method to detect controversy on news issues. This is done by performing an analysis of people’s reaction on social media to news articles reporting these issues. Detecting controversial news topics on web is a relevant problem today. It helps to identify the issues upon which people have divided opinion and is specially useful on topics such as a presidential election, government reforms,climate change etc. We use sentiment analysis and word matching to accomplish this task. We show the application of our method for detecting controversial topics during the US Presidential elections 2016