With over 5000 satellites and 50000 debris currently in low-earth orbit, assessing collision risk between these entities is a problem of growing importance. With about 2 tracking points per day reported by Celestrak for each tracked entity, intermediate orbit states are “propagated” from these measurements using perturbation calculations of which SGP4 is the one in dominant use. Perturbation methods can accumulate errors of 10s of km over a week, and many attempts to use established machine learning techniques, time series analysis and deep neural networks (DNN) to predict based on past measurements have been made so far. We focus on the problem of predicting the orbit of an entity with higher accuracy over 7 days into the future in order to enable better collision risk assessment and ample time to make corrective maneuvers. Towards this, we show that training a physics informed Neural Ordinary Differential Equation (NeuraIODE) over a few measurements can help predict the near future with reduced propagation error compared to SGP4 in current use for this purpose, much further into the future than known ML based methods. We discuss training costs, and highlight challenges with scaling up this approach for the large number of debris and satellites in orbit today.

Intent detection is one of the crucial Natural Language Understanding(NLU) tasks studied extensively. Misclassification of intents impacts the overall performance of the conversational systems as natural language understanding is the first mean of interaction between a user and a virtual agent. The traditional approach of intent detection is limited only to textual features of user utterances and overlooks other semantic features. The sentiment and emotional state of the speaker are two such semantic features that have essential impacts on intent detection, as they implicitly express user intention conveyed through the user’s message. Depending on the context, these features can help the dialogue agent respond to the same intent with varying degrees of sentiment and emotion. Thus, investigating the role of sentiment and emotion on intent detection is a matter of great interest. The current work investigates the impact of utilizing sentiment and emotion information on intent detection tasks and proposes emotion and sentiment aided intent detection models. We also investigate the impact of sentiment and emotion using three different multitasking frameworks and present a joint model that utilizes the co-relation information across these tasks to correctly identify all these NLU aspects (intent, sentiment, and emotion). The obtained experimental results by the proposed models outperform several baselines and state-of-the-art intent detection models on multiple datasets by a significant margin of 1.8% - 3%, demonstrating the significant role of sentiment and emotion features in intent detection1.

Movies reflect society and also hold power to transform opinions. Social biases and stereotypes present in movies can cause extensive damage due to their reach. These biases are not always found to be the need of storyline but can creep in as the author’s bias. Movie production houses would prefer to ascertain that the bias present in a script is the story’s demand. Today, when deep learning models can give human-level accuracy in multiple tasks, having an AI solution to identify the biases present in the script at the writing stage can help them avoid the inconvenience of stalled release, lawsuits, etc. Since AI solutions are data intensive and there exists no domain specific data to address the problem of biases in scripts, we introduce a new dataset of movie scripts that are annotated for identity bias. The dataset contains dialogue turns annotated for (i) bias labels for seven categories, viz., gender, race/ethnicity, religion, age, occupation, LGBTQ, and other, which contains biases like body shaming, personality bias, etc. (ii) labels for sensitivity, stereotype, sentiment, emotion, emotion intensity, (iii) all labels annotated with context awareness, (iv) target groups and reason for bias labels and (v) expert-driven group-validation process for high quality annotations. We also report various baseline performances for bias identification and category detection on our dataset.

Building conversation agents requires considerable manual effort in creating training data for intents / entities as well as mapping out extensive conversation flows.In this demonstration, we present ICM (Intent and Conversation Mining), a tool which can make the BOT build and update process much faster.ICM can be used to analyze existing conversation logs and help a bot designer to cluster, visualize and analyze customer intents; train custom intent models; and also to map and optimize conversation flows.The tool can be used for first time deployment or subsequent conversational flow updates in chatbots.

The formulation of a claim rests at the core of argument mining. To demarcate between a claim and a non-claim is arduous for both humans and machines, owing to latent linguistic variance between the two and the inadequacy of extensive definition-based formalization. Furthermore, the increase in the usage of online social media has resulted in an explosion of unsolicited information on the web presented as informal text. To account for the aforementioned, in this paper, we propose DESYR. It is a framework that intends on annulling the said issues for informal web-based text by leveraging a combination of hierarchical representation learning (dependency-inspired Poincaré embedding), definition-based alignment, and feature projection. We do away with fine-tuning compute-heavy language models in favor of fabricating a more domain-centric but lighter approach. Experimental results indicate that DESYR builds upon the state-of-the-art system across four benchmark claim datasets, most of which were constructed with informal texts. We see an increase of 3 claim-F1 points on the LESA-Twitter dataset, an increase of 1 claim-F1 point and 9 macro-F1 points on the Online Comments (OC) dataset, an increase of 24 claim-F1 points and 17 macro-F1 points on the Web Discourse (WD) dataset, and an increase of 8 claim-F1 points and 5 macro-F1 points on the Micro Texts (MT) dataset. We also perform an extensive analysis of the results. We make a 100-D pre-trained version of our Poincaré-variant along with the source code.

In this paper, we propose an unsupervised approach for knowledge graph (KG) creation from conversational data. We make use of intent classification and slot-filling, the two important components of any dialogue agent, exploit their interconnectedness, and finally construct a KG. We build a supervised intent classifier to extract the intent classes, and then on top of this we run our occlusion based slot-information extraction algorithm. Our algorithm is able to make use of supervised training of intent classifiers for extracting the relevant slot-information in an unsupervised way. To test the effectiveness of our system, we perform both automatic and manual evaluation of our intent-classifier and slot-filling system on three dialog datasets. Finally, we construct a knowledge graph from the dialogue conversation using an algorithm that makes use of our occlusion based slot-information extraction module. Empirical evaluation shows that our occlusion based method is able to successfully extract slot information from conversations, resulting in a high-quality KG.

Developing an adequate and human-like virtual agent has been one of the primary applications of artificial intelligence. In the last few years, task-oriented dialogue systems have gained huge popularity because of their upsurging relevance and positive outcomes. In real-world, users may not always have a predefined and rigid task goal beforehand; they upgrade/downgrade/change their goal component dynamically depending upon their utility value and agent's serving capability. However, existing virtual agents fail to incorporate this dynamic behavior, leading to either unsuccessful task completion or an ungratified user experience. The paper presents an end to end multimodal dialogue system for dynamic and co-operative goal setting, which incorporates i) a multi-modal semantic state representation in policy learning to deal with multi-modal inputs, ii) a goal manager module in a traditional dialogue manager for handling dynamic and goal unavailability scenarios effectively, iii) an accumulative reward (task/persona/sentiment) for task success, personalized persuasion and user-adaptive behavior, respectively. The obtained experimental results and the comparisons with baselines firmly establish the need and efficacy of the proposed system.

Conversational systems are of primary interest in the AI community.Organizations are increasingly using chatbot to provide round-the-clock support and to increase customer engagement.Many commercial bot building frameworks follow a standard approach that requires one to build and train an intent model to recognize user input.These frameworks require a collection of user utterances and corresponding intent to train an intent model.Collecting a substantial coverage of training data is a bottleneck in the bot building process.In cases where past conversation data is available, the cost of labeling hundreds of utterances with intent labels is time-consuming and laborious.In this paper, we present an intent discovery framework that can mine a vast amount of conversational logs and to generate labeled data sets for training intent models.We have introduced an extension to the DBSCAN (Ester et al., 1996) algorithm and presented a density-based clustering algorithm ITER-DBSCAN for unbalanced data clustering.Empirical evaluation on one conversation dataset, six intent dataset, and one short text clustering dataset show the effectiveness of our hypothesis.We release the datasets and code for future evaluation at

In this paper, we propose a hybrid technique for semantic question matching. It uses a proposed two-layered taxonomy for English questions by augmenting state-of-the-art deep learning models with question classes obtained from a deep learning based question classifier. Experiments performed on three open-domain datasets demonstrate the effectiveness of our proposed approach. We achieve state-of-the-art results on partial ordering question ranking (POQR) benchmark dataset. Our empirical analysis shows that coupling standard distributional features (provided by the question encoder) with knowledge from taxonomy is more effective than either deep learning or taxonomy-based knowledge alone.

We present a goal oriented virtual agent for task planning, inference and user personalization. The dialog system of the agent models the activities in the form of a task graph. Coordination and planning of these tasks are done through associated constraints and interdependencies. The system uses the interaction history of an individual and a basic user preference model to create a personalized travel planning experience, as an example. Sophisticated natural language understanding techniques are used to identify multiple intents from user, and partial user utterances. In this demonstration we show how the virtual agent can be used to help the user organize his/her day by performing a multitude of related activities including scheduling meetings, retrieving flight itineraries, making routing decisions by analyzing traffic patterns, retrieving hotel accommodation options and suggesting restaurants by integrating with multiple external systems as well as a decision support engine and knowledge graph.

Duplicate Bug Detection is the problem of identifying whether a newly reported bug is a duplicate of an existing bug in the system and retrieving the original or similar bugs from the past. This is required to avoid costly rediscovery and redundant work. In typical software projects, the number of duplicate bugs reported may run into the order of thousands, making it expensive in terms of cost and time for manual intervention. This makes the problem of duplicate or similar bug detection an important one in Software Engineering domain. However, an automated solution for the same is not quite accurate yet in practice, in spite of many reported approaches using various machine learning techniques. In this work, we propose a retrieval and classification model using Siamese Convolutional Neural Networks (CNN) and Long Short Term Memory (LSTM) for accurate detection and retrieval of duplicate and similar bugs. We report an accuracy close to 90% and recall rate close to 80%, which makes possible the practical use of such a system. We describe our model in detail along with related discussions from the Deep Learning domain. By presenting the detailed experimental results, we illustrate the effectiveness of the model in practical systems, including for repositories for which supervised training data is not available.

Ontology, the shared formal conceptualization of domain information, has been shown to have multiple applications in modeling, processing and understanding natural language text. In this work, we use distributed word vectors out of various recent language models from Deep Learning for semi-automated domain ontology creation for closed domains. We cover all major aspects of Domain Ontology Induction or Learning like concept identification, attribute identification, taxonomical and non-taxonomical relationship identification using the distributed word vectors. Preliminary results show that simple clustering based methods using distributed word vectors from these language models outperforms methods using models like LSI in ontology learning for closed domains.

We propose a novel cloud-based data storage solution framework named Data Vaporizer (DV). The proposed framework provides many unique features such as storing data over multiple clouds or storage zones, resistance against organized vendor attacks, maintaining data integrity and confidentiality through client-side processing, fault-tolerance against failure of one or more cloud storage locations and avoids vendor lock-in of data. Data Vaporizer is highly configurable to meet various client data encryption requirements, compliance to industry standards and fault tolerance constraints depending on the nature and sensitivity of the data. To enhance the level of security and reliability, especially to protect data against malicious attacks and secure key management in cloud, DV uses advanced techniques of secret sharing of the keys. The architecture and optimality of data placement and efficient key management algorithm of DV ensure that the solution is highly scalable. The data foot print and subsequent cost incurred by our storage solution is minimal, considering the benefits provided. The initial response for the adoption of DV in actual client scenarios is promising.

Management of legal contracts in various business domains such as Real Estate are examples of typical business process outsourcing activity. One of such process is Lease Abstraction, where largely manual inspection and validation of large commercial lease documents made for real estate deals is done by offshore experts and relevant information from the documents is extracted into a structured form. This structured information is further used for aggregate analytics and decision making by large real estate firms. We propose a system based on machine learning techniques to semi automate this process, essentially leading to 50% human effort savings. Our approach weaves together state-of-the-art machine learning techniques like supervised classifier models, sequence modeling techniques and various semi-supervised approaches. We articulate the effectiveness of our solution using the results from the experiments. Our platform is being used in production environment by Accenture Operations and the initial results and user feedback are encouraging.

Different PaaS (Platform as a Service) Clouds offer different set of capabilities and services and have different constraints on types of application that can be hosted on their platforms. Migrating existing enterprise applications to such platforms thus is non-trivial and needs a thorough assessment of the system to be migrated. In this paper, we present a novel approach for automated assessment of applications for migration to a target PaaS platform. We take an approach of systematically studying typical external technical services that different types of applications need in a traditional non-PaaS deployment and evaluate support for each of services in major PaaS environments. We have created rich sets of repositories each for technical capabilities and services used by typical enterprise applications as well as for the different technical services exposed for use by PaaS platforms along with their limitations and caveats. Using these repositories, our approach analyses the source code as well as the configuration files to recursively extract the services it requires and then tries to map them to a target PaaS platform. The approach results in a detailed report of the parts of the system that can migrate as-is, which need some changes, as well as those which can't be migrated at all due to the limitations of the chosen PaaS platform.

Quality of requirements written in natural language has always been a critical concern in software engineering. Poorly written requirements lead to ambiguity and false interpretation in different phases of a software delivery project. Further, incomplete requirements lead to partial implementation of the desired system behavior. In this paper, we present a model for harvesting domain (functional or business) knowledge. Subsequently we present natural language processing and ontology based techniques for leveraging the model to analyze requirements quality and for requirements comprehension. The prototype also provides an advisory to business analysts so that the requirements can be aligned to the expected domain standard. The prototype developed is currently being used in practice, and the initial results are very encouraging.

We present a method for the automatic extraction of glossary terms from unconstrained natural language requirements. The glossary terms are identified in two steps - a) compute units (which are candidates for glossary terms) b) disambiguate between the mutually exclusive units to identify terms. We introduce novel linguistic techniques to identify process nouns, abstract nouns and auxiliary verbs. The identification of units also handles co-ordinating conjunctions and adjectival modifiers. This requires solving co-ordination ambiguity and adjectival modifier ambiguity. The identification of terms among the units adapts an in-document statistical metric. We present an evaluation of our method over a real-life set of software requirements' documents and compare our results with that of a base algorithm. The intricate linguistic classification and the tackling of ambiguity result in superior performance of our approach over the base algorithm.

Source code clustering is an important technique used in software development and maintenance to understand the modular structure of code. An array of algorithms are available for clustering like simulated annealing based search. Source code have different kinds of features such as structural or textual features. The collection of these different types of source code features and computation of relevant feature metrics is a difficult task. Further, the clustering algorithms can run on metrics based on different types of source code features or their combinations. This flexibility makes it non-trivial to test effectiveness of clustering algorithms on a source code. In this paper, we present a highly configurable clustering workbench that allows the user to collect the various source code features and then to select the code features used for clustering, the clustering algorithm and its various parameters. Clustering quality metrics are computed. They allow comparison of algorithm output based on different combinations of code-features and algorithms. We also present the specific contribution made in multi-dimensional feature analysis and clustering. The tool hides the algorithm complexity from the user, thus allowing complete focus on understanding the 'effect' of the configuration choices. We have also applied this tool in real-life maintenance projects, where the users found it useful to tweak the clustering techniques for the source-code peculiarities.

Maintaining the state of applications and user sessions is difficult in large scale web-based software systems. This problem is particularly accentuated in the context of Cloud computing as Cloud providers, especially Platform as a Service (PaaS) vendors, do not explicitly support state management infrastructure - such as clustering. In a PaaS environment, a user has little or no access and control over the server platform and session management layer. Additionally, the platform tiers are generally loosely coupled and service-oriented. These make traditional session-state management techniques non-usable. In this work, we present ReLoC - a session-state management architecture for Cloud that uses loosely-coupled services and platform agnostic scalable messaging technology to propagate and save session states. Preliminary experiments show a very high level of tolerance to failures of the platform tiers without corresponding disruptions in user sessions. We argue that, in the context of PaaS Clouds, ReLoC architecture will be more scalable compared to traditional clustering environments.

Cloud Computing is increasingly becoming popular as many enterprise applications and data are moving into cloud platforms. However, a major barrier for cloud adoption is real and perceived lack of security. In this paper, we take a holistic view of cloud computing security - spanning across the possible issues and vulnerabilities connected with virtualization infrastructure, software platform, identity management and access control, data integrity, confidentiality and privacy, physical and process security aspects, and legal compliance in cloud. We present our findings from the points of view of a cloud service provider, cloud consumer, and third-party authorities such as Govt. We also discuss important research directions in cloud security in areas such as Trusted Computing, Information Centric Security and Privacy Preserving Models. Finally, we sketch a set of steps that can be used, at a high level, to assess security preparedness for a business application to be migrated to cloud.

Enterprises are looking at grid computing as a technology of enormous potential. One such issue is the issue of grid application migration, where legacy applications are migrated to the grid environment. The grid application migration framework (GAMF), is an attempt to provide a systematic approach for handling the migration process. The framework consists of three independently deployable components: grid code analyzer (GCA), grid task generator (GTG), and grid simulator (GS). The GCA component analyzes the different applications and generates a directed acyclic graph (DAG) based on the information obtained from analysis. The DAG is based on both static and runtime information available from application profiling. The DAG information is then analyzed and grid tasks are generated based on different DAG reducing and clustering algorithms. The simulator takes the generated DAG as input and provides and analyzes the application performance with respect to grid infrastructure and policies

A grid is described as a distributed network which comprises of heterogeneous and non-dedicated elements. The heterogeneity of a grid is not only defined in terms of computing elements and operating systems but also in terms of implementation of policies, policy decisions and the environment. Currently, most grid solutions are targeted at the high performance community for e-sciences and hence are not adequate for enterprises. In this paper, we present an enterprise level workflow and grid management system called the unified grid management and data architecture (UGanDA) which takes care of most needs of enterprises and can be used in traditional e-sciences high performance community as well. It consists of two components: a grid workflow manager called GridWorM and a grid infrastructure manager called MAGI. The two independently deployable components orchestrate to provide the final goal. Representative results are discussed to prove the effectiveness of the system.