As deep learning methods form a critical part in commercially important applications such as autonomous driving and medical diagnostics, it is important to reliably detect out-of-distribution (OOD) inputs while employing these algorithms. In this work, we propose an OOD detection algorithm which comprises of an ensemble of classifiers. We train each classifier in a self-supervised manner by leaving out a random subset of training data as OOD data and the rest as in-distribution (ID) data. We propose a novel margin-based loss over the softmax output which seeks to maintain at least a margin m between the average entropy of the OOD and in-distribution samples. In conjunction with the standard cross-entropy loss, we minimize the novel loss to train an ensemble of classifiers. We also propose a novel method to combine the outputs of the ensemble of classifiers to obtain OOD detection score and class prediction. Overall, our method convincingly outperforms Hendrycks et al.[7] and the current state-of-the-art ODIN [13] on several OOD detection benchmarks.
Abstract-Recent studies show that concept-based approaches to opinion mining perform better than more canonical methods based on keyword spotting or word co-occurrence frequencies. SenticNet 1.0 is one of the most widely used publicly available resources for concept-based opinion mining. It gives polarity scores for a large number of single-and multi-word common sense concepts. However, developing high-quality opinion mining and sentiment analysis systems also requires affective information associated with the concepts. In this work, we present a methodology for enriching SenticNet concepts with affective information by assigning to them an emotion label. The created resource is freely available for academic use.
In this paper, emotion analysis on blog texts has been carried out for a less privileged language like Bengali. Ekman's six basic emotion types have been selected for reliable and semi automatic word level annotation. An automatic classifier has been applied for recognizing six basic emotion types for different words in a sentence. Application of different scoring strategies to identify sentence level emotion tag based on the acquired word level emotion constituents have produced satisfactory performance.
Abstract-SenticNet 1.0 is one of the most widely used freelyavailable resources for concept-level opinion mining, containing about 5700 common sense concepts and their corresponding polarity scores. Specific affective information associated to such concepts, however, is often desirable for tasks such as emotion recognition. In this work, we propose a method for assigning emotion labels to SenticNet concepts based on a semi-supervised classifier trained on WordNet-Affect emotion lists with features extracted from various lexical resources.
Sentiment Analysis in Twitter has been considered as a vital task for a decade from various academic and commercial perspectives. Several works have been performed on Twitter sentiment analysis or opinion mining for English in contrast to the Indian languages. Here, we summarize the objectives and evaluation of the sentiment analysis task in tweets for three Indian languages namely Bengali, Hindi and Tamil. This is the first attempt to sentiment analysis task in the context of Indian language tweets. The main objective of this task was to classify the tweets into positive, negative and neutral polarity. For training and testing purpose, the tweets from each language were provided. Each of the participating teams was asked to submit two systems, constrained and unconstrained systems for each of the languages. We ranked the systems based on the accuracy of the systems. Total of six teams submitted the results and the maximum accuracy achieved for Bengali, Hindi and Tamil are 43.2 %, 55.67 %, and 39.28 % respectively.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.