Abstract

Automatic categorization of computer science research papers using just the abstracts, is a hard problem to solve. This is due to the short text length of the abstracts. Also, abstracts are a general discussion of the topic with few domain specific terms. These reasons make it hard to generate good representations of abstracts which in turn leads to poor categorization performance. To address this challenge, external Knowledge Bases (KB) like Wikipedia, Freebase etc. can be used to enrich the representations for abstracts, which can aid in the categorization task. In this work, we propose a novel method for enhancing classification performance of research papers into ACM computer science categories using knowledge extracted from related Wikipedia articles and Freebase entities. We use state-of-the-art representation learning methods for feature representation of documents, followed by learning to rank method for classification. Given the abstracts of research papers from the Citation Network Dataset containing 0.24 M papers, our method of using KB, outperforms a baseline method and the stateof-the-art deep learning method in classification task by 13.25% and 5.41% respectively, in terms of accuracy. We have also open-sourced the implementation of the project4.

Abstract

We describe our submission at TAC 2017 for extracting entities and relations of interest from drug labels. We employ an end-to-end relation extraction system which jointly extracts both entities and relations. The task of end-to-end relation extraction consists of identifying boundaries of entity mentions, entity types of these mentions and appropriate semantic relation for each pair of mentions. Based on our earlier work (Pawar et al., 2017), a single neural network model (“All Word Pairs” model ie AWP-NN) is trained to assign an appropriate label to each word pair in a given sentence for performing end-to-end relation extraction. Moreover, we build an ensemble of multiple AWPNN models to achieve better performance than the individual models. We achieved 73.18% and 24.79% F-measures for entity and relation extraction, respectively.

Abstract

The purpose of this research is to address the issue of enhancing the Quality of Service(QoS) in IP networks, to make it more optimal in a given network of networks using a novel approach. In this approach, QoS is provided to the packet even when the packet crosses its access network into the backbone. It proposes a new design or set of features in the routers which maps and translates the QoS provisioning from one network to the other. This paper targets at making the quality of service more dynamic by attempting to provide QoS not just at ingress or egress routers but also in the backbone network. In addition, it discusses the economic aspects for upgraded network infrastructure and also comments on the scenarios in which the proposed schema would be helpful along with the placement of dynamic QoS aware routers in the network of networks.

Abstract

Developers need to comprehend a program before modifying it. In such cases,developers use cues to establish the relevance of a piece of information with a task. Beingfamiliar with different kinds of cues will help the developers to comprehend a programfaster. But unlike the experienced developers, novice developers fail to recognize therelevant cues, because (a) there are many cues and (b) they might be unfamiliar with theartifacts. However, not much is known about the developers’ choice of cue. Hence, weconducted two independent studies to understand the kind of cues used by the developersand how a tool influences their cue selection. First, from our user study on two commoncomprehension tasks, we found that developers actively seek the cues and their cue sourcechoices are similar but task dependent. In our second exploratory study, we investigatedwhether an integrated development environment (IDE) influences a developer’s cuechoices. By observing their interaction history while fixing bugs in Eclipse IDE, we foundthat the IDE’s influence on their cue choices was not statistically significant. Finally, as acase in point, we propose a novel task-specific program summarization approach to aidnovice developers in comprehending an unfamiliar program. Our approach used devel-opers cue choices to synthesize the summaries. A comparison of the synthesized sum-maries with the summaries recorded by the developers shows both had similar content.This promising result encourages us to explore task-specific cue models, which can aidnovice developers to accomplish complex comprehension tasks faster

Abstract

Hate speech detection on Twitter is critical for applications like controversial event extraction, building AI chatterbots, content recommendation, and sentiment analysis. We define this task as being able to classify a tweet as racist, sexist or neither. The complexity of the natural language constructs makes this task very challenging. We perform extensive experiments with multiple deep learning architectures to learn semantic word embeddings to handle this complexity. Our experiments on a benchmark dataset of 16K annotated tweets show that such deep learning methods outperform state-of-the-art char/word n-gram methods by~ 18 F1 points.

Abstract

Scientific article recommendation problem deals with recommending similar scientific articles given a query article. It can be categorized as a content based similarity system. Recent advancements in representation learning methods have proven to be effective in modeling distributed representations in different modalities like images, languages, speech, networks etc. The distributed representations obtained using such techniques in turn can be used to calculate similarities. In this paper, we address the problem of scientific paper recommendation through a novel method which aims to combine multimodal distributed representations, which in this case are: 1. distributed representations of paper's content, and 2. distributed representation of the graph constructed from the bibliographic network. Through experiments we demonstrate that our method outperforms the state-of-the-art distributed representation methods in text and graph, by 29.6% and 20.4%, both in terms of precision and mean-average-precision respectively.

Abstract

Research in analysis of microblogging platforms is experiencing a renewed surge with a large number of works applying representation learning models for applications like sentiment analysis, semantic textual similarity computation, hashtag prediction, etc. Although the performance of the representation learning models has been better than the traditional baselines for such tasks, little is known about the elementary properties of a tweet encoded within these representations, or why particular representations work better for certain tasks. Our work presented here constitutes the first step in opening the black-box of vector embeddings for tweets.

Abstract

Automated segmentation of cortical and noncortical human brain structures has been hither to approached using nonrigid registration followed by label fusion. We propose an alternative approach for this using a convolutional neural network (CNN) which classifies a voxel into one of many structures. Four different kinds of two-dimensional and three-dimensional intensity patches are extracted for each voxel, providing local and global (context) information to the CNN. The proposed approach is evaluated on five different publicly available datasets which differ in the number of labels per volume. The obtained mean Dice coefficient varied according to the number of labels, for example, it is0.844
0.031and0.743
0.019for datasets with the least(32) and the most (134) number of labels, respectively. These figures are marginally better or on par with those obtained with the current state-of-the-art methods on nearly all datasets, at a reduced computational time. The consistently good performance of the proposed method across datasets and no requirement for registration make it attractive for many applications where reduced computational time is necessary.

Abstract

Retinal vessel keypoint detection and classification is a fundamental step in tracking the physiological changes that occur in the retina which is linked to various retinal and systemic diseases. In this paper, we propose a novel Vessel Keypoint Detector (VKD) which is derived from the projection of log-polar transformed binary patches around vessel points.VKD is used to design a two stage solution for junction detection and classification. In the first stage, the keypoints detected using VKD are refined using curvature orientation information to extract candidate junctions. True junctions from these candidates are identified in a supervised manner using a Random Forest classifier. In the next stage, a novel combination of local orientation and shape based features is extracted from the junction points and classified using asecond Random Forest classifier. Evaluation results on five datasets show that the designed system is robust to changes in resolution and other variations across datasets, with average values of accuracy/sensitivity/specificity for junction detection being 0.78/0.79/0.75 and for junction classification being 0.87/0.85/0.88. Our system outperforms the state of the art method [1] by at least 11%, on the DRIVE and IOSTAR datasets. These results demonstrate the effectiveness of VKDfor vessel analysis.

Abstract

Retinal Neovascularization (NV) is a critical stage of Diabetic Retinopathy (DR) and its detection is important to prevent blindness. Existing fully supervised frameworks typically take a patch-based approach and report good results only on limited number of images due to sparsity of annotated data.We propose a patch-based semi-supervised framework which paves the way for including unlabeled data in training. In this framework, NV patches are modeled using oriented energy and vesselness based features. These features are fused within a co-training based semi-supervised framework by using neighborhood information in feature space. Rule-based criteria on patch-level neovascularity scores is used to derive the final image-level decision. The proposed approach was evaluated on 1 private and 3 public datasets, both at patch and image level detection on nearly 200,000 patches. An AUCof 0.985 with sensitivity of 96.2% at specificity of 92.6% was obtained for abnormality detection at patch-level, while at the image-level, a sensitivity of 96.76% at a specificity of 91.85%were obtained. The achieved performance on a large number of patches indicates the robustness of our approach.