Search

We present an automated, visual question answering based companion – VisQuelle - to facilitate elementary learning of word-object associations. In particular, we attempt to harness the power of machine learning models for object recognition and the understanding of combined processing of images and text data from visual-question answering to provide variety and nuance in the images associated with letters or words presented to the elementary learner. We incorporate elements such as gamification to motivate the learner by recording scores, errors, etc., to track the learner’s progress. Translation is also provided to reinforce word-object associations in the user’s native tongue, if the learner is using VisQuelle to learn a second language.

We present an automated approach to segment key structures of the eye, viz., the iris, pupil and sclera in images obtained using an Augmented Reality (AR)/ Virtual Reality (VR) application. This is done using a two-step classifier: In the first step, we use an auto encoder-decoder network to obtain a pixel-wise classification of regions that comprise the iris, sclera and the background (image pixels that are outside the region of the eye). In the second step, we perform a pixel-wise classification of the iris region to delineate the pupil. The images in the study are from the OpenEDS challenge and were used to evaluate both the accuracy and computational cost of the proposed segmentation method. Our approach achieved a score of 0.93 on the leaderboard, outperforming the baseline model by achieving a higher accuracy and using a smaller number of parameters. These results demonstrate the great promise pipelined models hold along with the benefit of using domain-specific processing and feature engineering in conjunction with deep-learning based approaches for segmentation tasks.

Adverse drug reactions (ADRs) are pharmacological events triggered by drug interactions with various sources of origin including drug-drug interactions. While there are many computational studies that explore models to predict ADRs originating from single drugs, only a few of them explore models that predict ADRs from drug combinations. Further, as far as we know, none of them have developed models using transcriptomic data, specifically the LINCS L1000 drug induced gene expression data to predict ADRs for drug combinations. In this study we use the TWOSIDES database as a source of ADRs originating from two-drug combinations. 34,549 common drug pairs between these two databases were used to train an artificial neural network (ANN), to predict 243 ADRs that were induced by at least 10% of the drug pairs. Our model predicts the occurrence of these ADRs with an average accuracy of 82% across a multi fold cross validation.Source Code and input dataset used in this study can be found at: https://bitbucket.org/ishita98/prediction-of-adr/src/master/

In recent years, the role of big data analytics has exponentially grown and is now slowly making its way into the education industry. Several attempts are being made in this sphere in order to improve the quality of education being provided to students and while many collaborations have been carried out before, automated scoring of answers has been explored to a rather limited extent. One of the biggest hurdles to choosing constructed-response assessments over multiple-choice assessments is the effort and large cost that comes with their evaluation and this is precisely the issue that this project aims to solve. The aim is to accept raw-input from the student in the form of their answer, preprocess the answer, and automatically score the answer. In addition, we have made this a real-time system that captures "snapshots" of the writer's progress with respect to the answer, allowing us to unearth trends with respect to the way a student thinks, and how the student has arrived at their final answer., Comment: This paper was originally written in mid 2018

The ability to estimate the probability of a drug to receive approval in clinical trials provides natural advantages to optimizing pharmaceutical research workflows. Success rates of clinical trials have deep implications for costs, duration of development, and under pressure due to stringent regulatory approval processes. We propose a machine learning approach that can predict the outcome of the trial with reliable accuracies, using biological activities, physicochemical properties of the compounds, target-related features, and NLP-based compound representation. In the above list, biological activities have never been used as an independent variable towards the prediction of clinical trial outcomes. We have extracted the drug-disease pair from clinical trials and mapped target(s) to that pair using multiple data sources. Empirical results demonstrate that ensemble learning outperforms independently trained, small-data ML models. We report results and inferences derived from a Random forest classifier with an average accuracy of 93%, and an F1 score of 0.96 for the 'Pass' class. 'Pass' refers to one of the two classes (Pass/ Fail) of all clinical trials and the model performed well in predicting the "Pass" category. Through the analysis of feature contributions to predictive capability, we have demonstrated that bioactivity plays a statistically significant role in predicting clinical trial outcome. A significant effort has gone into the production of the dataset that, for the first time, integrates clinical trial information with protein targets. Cleaned, organized, integrated data and code to map these entities, - created as a part of this work, are available open-source. This reproducibility and the freely available code ensure that researchers with access to deep curated and proprietary clinical trial databases (we only use open-source data in this study) can further expand the scope of the results. This article is protected by copyright. All rights reserved.

We present an automated, visual question answering based companion – Vis Quelle - to facilitate elementary learning of word-object associations. In particular, we attempt to harness the power of machine learning models for object recognition and the understanding of combined processing of images and text data from visual-question answering to provide variety and nuance in the images associated with letters or words presented to the elementary learner. We incorporate elements such as gamification to motivate the learner by recording scores, errors, etc., to track the learner’s progress. Translation is also provided to reinforce word-object associations in the user’s native tongue, if the learner is using Vis Quelle to learn a second language. Keywords: Visual question answering; object recognition; question generation; question answering; word-object association.

In recent times, the need for virtual screening of chemical compounds has grown with the advent of computational synthesis and de-novo generation of drugs. The state-of-the-art benchmarks of virtual screening today, incorporate chemical and physiological properties, binding affinities, along with the targets of known chemical compounds. However, benchmarks for the classification of drugs into overarching functional groups based purely on the structure of the compound are yet to be explored. In this paper, we introduce VIRTECS: a tool that leverages the simplified molecular-input line-entry system (SMILES) – a structural representation of a drug – to enable virtual screening of large scale chemical databases, based on the therapeutic classes of drugs. The only input required by the system is the SMILES representation, one that is readily available with most computational generation approaches. The experimental results on multiple datasets demonstrate the potency of structural information in determining the functional groups of chemical compounds. VIRTECS holds enormous potential in yielding insights into various properties of novel molecules when an embedding of the SMILES input is used and paired with an apposite graph algorithm, and tested with known molecules. We present a framework that allows for multiple combinations of the input (SMILES with or without the embedding) and a choice of models and databases that can be tested based on the desired output: insight to the function or potential therapeutic value of a chemical compound.

This paper focuses on improving the conversational ability of a robo receptionist. In particular, we seek to improve the ability to retrieve information specific to an organization through the design of a named-entity-recognition module. We accentuate the chatbot’s sensitivity to a user’s comment and the tone of a conversation through designing a fine-grained sentiment analysis module. And, finally, we have ensured the output of the self-learning chatbot is positive and pleasant through a toxic-comment classifier that improves upon a dictionary-based profanity detection module. Improving the core components of the chatbot, viz., the named entity recognition, sentiment analysis and toxic comment classification modules, reflect as an improvement in the performance of the chatbot. The performance of these modules in comparison with predecessor approaches and the code to reproduce the results have also been included to facilitate further improvements in these directions.

Time-series data pertaining to several related domains plays an important role as reinforcement for understanding the behaviour of that ecosystem. This paper intends to analyse one such system through the inspection of the rise/fall of stock prices for a duration in relation to buzzwords trending over the social media during the same time. We present a pipeline to elicit these insights and facilitate prediction of stock prices. In summary, the pooled data is subjected to time-series operations such as differencing transformations, smoothing and filtering to transform the data to a format amenable for further processing. Then, the trend corresponding to the keywords is illustrated by modelling their popularity using a counting-based algorithm customized for the application. This is followed by an attempt to forecast stock prices. The entire system is bundled into a software application that effectively delivers the results and visualizations in an intuitive fashion for a naive user. Throughout the pipeline, we have used a comparative paradigm that has helped us assess the proposed solution against alternatives.