In this work, we argue that not all sequence-to-sequence tasks require the strong inductive biases of autoregressive (AR) models. Tasks like multilingual transliteration, code refactoring, grammatical correction or text normalization often rely on local dependencies where the full modeling capacity of AR models can be overkill, creating a trade-off between their high accuracy and high inference latency. While non-autoregressive (NAR) models offer speed, they typically suffer from hallucinations and poor length control. To explore this trade-off, we focus on the multilingual transliteration task in Indic languages and introduce NADIR, a novel NAR architecture designed to strike a balance between speed and accuracy. NADIR integrates a Differential Transformer and a Mixture-of-Experts mechanism, enabling it to robustly model complex character mappings without sequential dependencies. NADIR achieves over a 13× speed-up compared to the state-of-the-art AR baseline. It maintains a competitive mean Character Error Rate of 15.78%, compared to 14.44% for the AR model and 21.88% for a standard NAR equivalent. Importantly, NADIR reduces Repetition errors by 49.53%, Substitution errors by 24.45%, Omission errors by 32.92%, and Insertion errors by 16.87%. This work provides a practical blueprint for building fast and reliable NAR systems, effectively bridging the gap between AR accuracy and the demands of real-time, large-scale deployment.

Saurabh Srivastava, Mayur Patidar, Sudip Chowdhury, Puneet Agarwal, Indrajit Bhattacharya, Gautam Shroff. Proceedings of the 16th Conference of the European Chapter of the Association for Computational Linguistics: Main Volume. 2021.

In the process of filing patents, attorneys need to ask many questions, to the inventors, to ascertain patent eligibility. We propose to ease up such conversation through a deep learning-based system. This system can automatically check whether all key ingredients required for checking the patent eligibility are present in technical write-up shared by the inventors. If not, the inventors can provide the missing information. We present a trainable model to identify various ingredients such as the objective, motivation, new observation, etc. from research articles. We model this as a sentence classification problem, which is a difficult task because a patent can be filed in any domain, and sentences involved can often be very long. To this end, we propose a dilated LSTM and capsule-based neural network architecture. We present experimental results of the proposed model on a real-world patent dataset covering patent applications in diverse domains in which our organization is carrying out research and innovation activities, and also three publicly available sentence classification datasets. Through empirical analysis, we show that a) Our model performs significantly better than several strong baselines on the patent dataset; b) Performing dilation operation on LSTMs allows us to capture long term dependencies; c) our model is comparable to existing state-of-art approaches on the publicly available datasets; d) Error analysis through LIME shows that the proposed approach can help patent attorneys to interpret the decisions taken by the classifier.

Sequence Labeling is one of the most prominent tasks in NLP. The traditional text classification models do not carry context from one sentence to another and hence may not perform well on these tasks. These models lack a hierarchical structure that can aid them in dissecting the input structure at different levels to allow flow of context between sentences. In this paper, we propose a hierarchical neural network comprising of Bi-LSTMs, Dilated Convolution operation, Capsules and Conditional Random Field (CRF) to understand the discourse/ abstract structure and predict next probable label by using label history. We have performed experiments on 3 publicly available datasets through which we have demonstrated that our model has achieved state-of-art performance on these datasets.

Helpdesk is a key component of any large IT organization, where users can log a ticket about any issue they face related to IT infrastructure, administrative services, human resource services, etc. Normally, users have to assign appropriate set of labels to a ticket so that it could be routed to right domain expert who can help resolve the issue. In practice, the number of labels are very large and organized in form of a tree. It is non-trivial to describe the issue completely and attach appropriate labels unless one knows the cause of the problem and the related labels. Sometimes domain experts discuss the issue with the users and change the ticket labels accordingly, without modifying the ticket description. This results in inconsistent and badly labeled data, making it hard for supervised algorithms to learn from. In this paper, we propose a novel approach of creating a conversational helpdesk system, which will ask relevant questions to the user, for identification of the right category and will then raise a ticket on users' behalf. We use attention based seq2seq model to assign the hierarchical categories to tickets. We use a slot filling model to help us decide what questions to ask to the user, if the top-k model predictions are not consistent. We also present a novel approach to generate training data for the slot filling model automatically based on attention in the hierarchical classification model. We demonstrate via a simulated user that the proposed approach can give us a significant gain in accuracy on ticket-data without asking too many questions to users. Finally, we also show that our seq2seq model is as versatile as other approaches on publicly available datasets, as state of the art approaches.

In this paper we present the design, architecture and implementation of KNADIA, a conversational dialogue system for intra-enterprise use, providing knowledge-assisted question answering and transactional assistance to employees of a large organization. KNADIA has been deployed in production in TCS, a large organization with over 380,000 employees distributed globally; the system is currently supporting a few thousand active users making hundreds of queries per day. We identify, define and distinguish two distinct classes of use-cases: virtual assistance and knowledge synthesis, which we have found to cover a variety of enterprise needs. KNADIA supports both types of conversational agents, with multiple instances of each, while presenting a common digital persona. KNADIA disambiguates which agent should respond to each query. Further, since individual components use deep learning algorithms giving probabilistic outputs, the confidence with which different components answer is often required before answering or acting. The user's dialogue context also needs to be maintained judiciously. Due to these and many other challenges, the overall architecture of KNADIA is non-trivial. We present instances of HR-assistance and technical knowledge synthesis that are in production use in TCS along with accuracy figures and key performance metrics. Finally, we suggest that many elements of our architecture are also generally applicable to other complex deep learning systems.

Recent proliferation of conversational systems has resulted in an increased demand for more natural dialogue systems, capable of more sophisticated interactions than merely providing factual answers. This is evident from usage pattern of a conversational system deployed within our organization. Users not only expect it to perform co-reference resolution of anaphora, but also of the antecedent or posterior facts presented by users with respect to their query. Presence of such facts in a conversation sometimes modifies the answer of main query, e.g., answer to 'how many sick leave do I get?' would be different when a fact 'I am on contract' is also present. Sometimes there is a need to collectively resolve three or four such facts. In this paper, we propose a novel solution which uses hierarchical neural network, comprising of BiLSTM layer and a maxpool layer that is hierarchically stacked to first obtain a representation of each user utterance and then to obtain a representation for sequence of utterances. This representation is used to identify users' intention. We also improvise this model by using skip connections in the second network to allow better gradient flow. Our model, not only a)~resolves the antecedent and posterior facts, but also b)~performs better even on self-contained queries. It is also c)~faster to train, making it the most promising approach for use in our environment where frequent training and tuning is needed. It slightly outperforms the benchmark on a publicly available dataset, and e)~performs better than obvious baselines approaches on our datasets.

We describe an automated assistant for answering frequently asked questions; our system has been deployed, and is currently answering HR-related queries in two different areas (leave management and health insurance) to a large number of users. The needs of a large global corporate lead us to model a frequently asked question (FAQ) to be an equivalence class of actually asked questions, for which there is a common answer (certified as being consistent with the organization's policy). When a new question is posed to our system, it finds the class of question, and responds with the answer for the class. At this point, the system is either correct (gives correct answer); or incorrect (gives wrong answer); or incomplete (says "I don't know''). We employ a hybrid deep-learning architecture in which a BiLSTM-based classifier is combined with second BiLSTM-based Siamese network in an iterative manner: Questions for which the classifier makes an error during training are used to generate a set of misclassified question-question pairs. These, along with correct pairs, are used to train the Siamese network to drive apart the (hidden) representations of the misclassified pairs. We present experimental results from our deployment showing that our iteratively trained hybrid network: (a) results in better performance than using just a classifier network, or just a Siamese network; (b) performs better than state-of-the art sentence classifiers in the two areas in which it has been deployed, in terms of both accuracy as well as precision-recall tradeoff; and (c) also performs well on a benchmark public dataset. We also observe that using question-question pairs in our hybrid network, results in marginally better performance than using question-to-answer pairs. Finally, estimates of precision and recall from the deployment of our automated assistant suggest that we can expect the burden on our HR department to drop from answering about 6000 queries a day to about 1000.

There exists a class of problems in e-commerce and retail businesses where the shopping behavior of customers is analyzed in order to predict their repeat behavior for products or retail stores. This analysis plays a crucial role in advertisement budgeting, product placement and relevant customer targeting. Researchers have addressed this problem by using standard predictive models, which use ad hoc features. We propose a metamodel that abstracts the different dimensions of data present in transactional datasets. These dimensions can be customer, product, offer, target, marketplace and transactions. Our framework also has abstract functions for comprehensive feature set generation, and includes different machine learning algorithms to learn prediction model. Our framework works end-to-end from feature engineering to reporting repeat probabilities of customers for products (or marketplace, brand, website or storechain). Moreover, the predicted repeat behavior of customers for different products along with their transactional history is used by our offer optimization model i-Prescribe to suggest products to be offered to customers with the goal of maximizing the return on investment of given marketing budget. We prove that our abstract features work on two different data-challenge datasets, by sharing experimental results.

Modern industrial equipments of all kinds are instrumented with a large number of sensors that continuously transmit their readings wirelessly, giving rise to what is often referred to as the `industrial internet'. Such data are often explored by engineers to determine the different usage patterns and behavior of similar machines. In this paper we describe a technique to automatically summarize the usage and behavioral patterns of a collection of similar machines by a small set of rules that nevertheless cover a large fraction of the observed data. We characterize the usage and behavior of a machine over a day, by a collection of single-sensor histograms; thus each day is a point in a high-dimensional space. We first cluster days according to each sensor separately and then combine the clusters using communities in a specially constructed graph that considers common days within clusters of different sensors. In the process some clusters of a single sensor get merged. Finally, we discover rules, each comprising of memberships in clusters of possibly different sensors. Thus, we use the term multi-subspace clustering to describe such a collection of cluster-based rules. Last but not the least, we attempt to cover a large fraction of observed days with a small number of such rules. We present empirical results on voluminous (100s of GBs) real-life sensor data and also compare our technique with related work in subspace clustering and histogram summarization.

Traditional reliability analysis of complex machinery involves statistical modeling of historical data on part failures from warranty claims, using distributions from exponential family such as the Weibull or log-normal distribution. When observed failures (in one or more parts) across a population of machines exceed the number expected based on such a model, this may serve as an early warning of a potential systemic problem with the population. Of course, such early warnings rely on some exceptionally high failures having actually occurred. However, modern connected vehicles, engines and machines of all kinds are equipped with on-board electronics that transmit alerts, referred to as `diagnostic trouble codes' or DTCs over the network, whenever abnormal conditions are detected. Such DTC signals should also be able to serve as early-warning indicators, typically before actual failures are observed in large numbers. In this paper, we develop a graphical Bayesian model that augments standard reliability analysis with early-warning indicators such as DTC signals observed over the industrial Internet. We demonstrate that our augmented model can detect of potential problems earlier than that using traditional reliability analysis. Going further, we note that significant deviations from expected failure counts might often occur only in some unknown subset of the population, e.g., a particular batch, or machines manufactured at a particular plant. In such cases, deviations from expected numbers are insignificant across the full population. We present a rule mining technique that discovers such subsets efficiently even when the number of dimensions across which a subset may be defined is large. We term our approach as reliability mining since it combines the use of a Bayesian reliability model with subgroup discovery using data mining techniques. We present experimental results using synthetically simulated scenarios as well as real-life data from a major global automobile manufacturer.

In distributed side-by-side software development, a pair of distributed team members are assigned a single task and allowed to (a) work concurrently on two different computers and (b) see each others' displays. They can control when they communicate with each other, view each others' actions, and in

Collaborative software development models, inspired by the open source community, are also being considered for development and deployment of business applications. We describe current work and future directions towards a hosted model to support collaboration during software development. We submit that such a platform can enable rapid application development, facilitate globally located virtual teams spanning locations to work as a single unit, and also encourage re-use, especially of technical frameworks. Going forward, we foresee extending such environments to providing an end-to-end development process accessible to globally distributed virtual teams.

Abstract To improve response time of a Web site, one replicates the site on multiple servers. The effectiveness of a replicated server system will depend on how the incoming requests are distributed among replicas. A large number of load‐balancing strategies for Web server systems have been proposed. In this paper we describe a testbed that can be used to evaluate the performance of different load‐balancing strategies. The testbed uses a general architecture which allows different load‐balancing approaches to be supported easily. It emulates a typical World Wide Web scenario and allows variable load generation and performance measurement. We have performed some preliminary experiments to measure the performance of a few policies for load balancing using this testbed. Copyright © 2003 John Wiley & Sons, Ltd.