Sunita Sarawagi

Vaidehi Patil, Partha Talukdar, Sunita Sarawagi

Pre-trained multilingual language models such as mBERT and XLM-R have demonstrated great potential for zero-shot cross-lingual transfer to low web-resource languages (LRL). However, due to limited model capacity, the large difference in the sizes of available monolingual corpora between high web-resource languages (HRL) and LRLs does not provide enough scope of co-embedding the LRL with the HRL, thereby affecting downstream task performance of LRLs. In this paper, we argue that relatedness among languages in a language family along the dimension of lexical overlap may be leveraged to overcome some of the corpora limitations of LRLs. We propose Overlap BPE (OBPE), a simple yet effective modification to the BPE vocabulary generation algorithm which enhances overlap across related languages. Through extensive experiments on multiple NLP tasks and datasets, we observe that OBPE generates a vocabulary that increases the representation of LRLs via tokens shared with HRLs. This results in improved zero-shot transfer from related HRLs to LRLs without reducing HRL representation and accuracy. Unlike previous studies that dismissed the importance of token-overlap, we show that in the low-resource related language setting, token overlap matters. Synthetically reducing the overlap to zero can cause as much as a four-fold drop in zero-shot transfer accuracy.

Mayank Kothyari, Dhruva Dhingra, Sunita Sarawagi et al.

Existing Text-to-SQL generators require the entire schema to be encoded with the user text. This is expensive or impractical for large databases with tens of thousands of columns. Standard dense retrieval techniques are inadequate for schema subsetting of a large structured database, where the correct semantics of retrieval demands that we rank sets of schema elements rather than individual elements. In response, we propose a two-stage process for effective coverage during retrieval. First, we instruct an LLM to hallucinate a minimal DB schema deemed adequate to answer the query. We use the hallucinated schema to retrieve a subset of the actual schema, by composing the results from multiple dense retrievals. Remarkably, hallucination $\unicode{x2013}$ generally considered a nuisance $\unicode{x2013}$ turns out to be actually useful as a bridging mechanism. Since no existing benchmarks exist for schema subsetting on large databases, we introduce three benchmarks. Two semi-synthetic datasets are derived from the union of schemas in two well-known datasets, SPIDER and BIRD, resulting in 4502 and 798 schema elements respectively. A real-life benchmark called SocialDB is sourced from an actual large data warehouse comprising 17844 schema elements. We show that our method1 leads to significantly higher recall than SOTA retrieval-based augmentation methods.

Abhijeet Awasthi, Ashutosh Sathe, Sunita Sarawagi

Text-to-SQL parsers typically struggle with databases unseen during the train time. Adapting parsers to new databases is a challenging problem due to the lack of natural language queries in the new schemas. We present ReFill, a framework for synthesizing high-quality and textually diverse parallel datasets for adapting a Text-to-SQL parser to a target schema. ReFill learns to retrieve-and-edit text queries from the existing schemas and transfers them to the target schema. We show that retrieving diverse existing text, masking their schema-specific tokens, and refilling with tokens relevant to the target schema, leads to significantly more diverse text queries than achievable by standard SQL-to-Text generation methods. Through experiments spanning multiple databases, we demonstrate that fine-tuning parsers on datasets synthesized using ReFill consistently outperforms the prior data-augmentation methods.

Abhijeet Awasthi, Nitish Gupta, Bidisha Samanta et al.

Despite cross-lingual generalization demonstrated by pre-trained multilingual models, the translate-train paradigm of transferring English datasets across multiple languages remains to be a key mechanism for training task-specific multilingual models. However, for many low-resource languages, the availability of a reliable translation service entails significant amounts of costly human-annotated translation pairs. Further, translation services may continue to be brittle due to domain mismatch between task-specific input text and general-purpose text used for training translation models. For multilingual semantic parsing, we demonstrate the effectiveness and flexibility offered by large language models (LLMs) for translating English datasets into several languages via few-shot prompting. Through extensive comparisons on two public datasets, MTOP and MASSIVE, spanning 50 languages and several domains, we show that our method of translating data using LLMs outperforms a strong translate-train baseline on 41 out of 50 languages. We study the key design choices that enable more effective multilingual data translation via prompted LLMs.

Abhijeet Awasthi, Soumen Chakrabarti, Sunita Sarawagi

Inference-time adaptation methods for semantic parsing are useful for leveraging examples from newly-observed domains without repeated fine-tuning. Existing approaches typically bias the decoder by simply concatenating input-output example pairs (cases) from the new domain at the encoder's input in a Seq-to-Seq model. Such methods cannot adequately leverage the structure of logical forms in the case examples. We propose StructCBR, a structured case-based reasoning approach, which leverages subtree-level similarity between logical forms of cases and candidate outputs, resulting in better decoder decisions. For the task of adapting Text-to-SQL models to unseen schemas, we show that exploiting case examples in a structured manner via StructCBR offers consistent performance improvements over prior inference-time adaptation methods across five different databases. To the best of our knowledge, we are the first to attempt inference-time adaptation of Text-to-SQL models, and harness trainable structured similarity between subqueries.

Vinit S. Unni, Ashish Mittal, Preethi Jyothi et al.

RNN-Transducers (RNN-Ts) have gained widespread acceptance as an end-to-end model for speech to text conversion because of their high accuracy and streaming capabilities. A typical RNN-T independently encodes the input audio and the text context, and combines the two encodings by a thin joint network. While this architecture provides SOTA streaming accuracy, it also makes the model vulnerable to strong LM biasing which manifests as multi-step hallucination of text without acoustic evidence. In this paper we propose LookAhead that makes text representations more acoustically grounded by looking ahead into the future within the audio input. This technique yields a significant 5%-20% relative reduction in word error rate on both in-domain and out-of-domain evaluation sets.

Soumya Chatterjee, Sunita Sarawagi, Preethi Jyothi

Online alignment in machine translation refers to the task of aligning a target word to a source word when the target sequence has only been partially decoded. Good online alignments facilitate important applications such as lexically constrained translation where user-defined dictionaries are used to inject lexical constraints into the translation model. We propose a novel posterior alignment technique that is truly online in its execution and superior in terms of alignment error rates compared to existing methods. Our proposed inference technique jointly considers alignment and token probabilities in a principled manner and can be seamlessly integrated within existing constrained beam-search decoding algorithms. On five language pairs, including two distant language pairs, we achieve consistent drop in alignment error rates. When deployed on seven lexically constrained translation tasks, we achieve significant improvements in BLEU specifically around the constrained positions.

Ashish Mittal, Darshan Prabhu, Sunita Sarawagi et al.

Harnessing pre-trained LLMs to improve ASR systems, particularly for low-resource languages, is now an emerging area of research. Existing methods range from using LLMs for ASR error correction to tightly coupled systems that replace the ASR decoder with the LLM. These approaches either increase decoding time or require expensive training of the cross-attention layers. We propose SALSA, which couples the decoder layers of the ASR to the LLM decoder, while synchronously advancing both decoders. Such coupling is performed with a simple projection of the last decoder state, and is thus significantly more training efficient than earlier approaches. A challenge of our proposed coupling is handling the mismatch between the tokenizers of the LLM and ASR systems. We handle this mismatch using cascading tokenization with respect to the LLM and ASR vocabularies. We evaluate SALSA on 8 low-resource languages in the FLEURS benchmark, yielding substantial WER reductions of up to 38%.

Adithya Bhaskar, Tushar Tomar, Ashutosh Sathe et al.

Research in Text-to-SQL conversion has been largely benchmarked against datasets where each text query corresponds to one correct SQL. However, natural language queries over real-life databases frequently involve significant ambiguity about the intended SQL due to overlapping schema names and multiple confusing relationship paths. To bridge this gap, we develop a novel benchmark called AmbiQT with over 3000 examples where each text is interpretable as two plausible SQLs due to lexical and/or structural ambiguity. When faced with ambiguity, an ideal top-$k$ decoder should generate all valid interpretations for possible disambiguation by the user. We evaluate several Text-to-SQL systems and decoding algorithms, including those employing state-of-the-art LLMs, and find them to be far from this ideal. The primary reason is that the prevalent beam search algorithm and its variants, treat SQL queries as a string and produce unhelpful token-level diversity in the top-$k$. We propose LogicalBeam, a new decoding algorithm that navigates the SQL logic space using a blend of plan-based template generation and constrained infilling. Counterfactually generated plans diversify templates while in-filling with a beam-search that branches solely on schema names provides value diversity. LogicalBeam is up to $2.5$ times more effective than state-of-the-art models at generating all candidate SQLs in the top-$k$ ranked outputs. It also enhances the top-$5$ Exact and Execution Match Accuracies on SPIDER and Kaggle DBQA.

Deeptanshu Malu, Deevyanshu Malu, Aditya Nemiwal et al.

We investigate training strategies that co-develop in-context learning (ICL) and in-weights learning (IWL), and the ability to switch between them based on context relevance. Although current LLMs exhibit both modes, standard task-specific fine-tuning often erodes ICL, motivating IC-Train - fine-tuning with in-context examples. Prior work has shown that emergence of ICL after IC-Train depends on factors such as task diversity and training duration. In this paper we show that the similarity structure between target inputs and context examples also plays an important role. Random context leads to loss of ICL and IWL dominance, while only similar examples in context causes ICL to degenerate to copying labels without regard to relevance. To address this, we propose a simple Contrastive-Context which enforces two types of contrasts: (1) mix of similar and random examples within a context to evolve a correct form of ICL, and (2) varying grades of similarity across contexts to evolve ICL-IWL mixtures. We present insights on the importance of such contrast with theoretical analysis of a minimal model. We validate with extensive empirical evaluation on four LLMs and several tasks. Diagnostic probes confirm that contrasted contexts yield stable ICL-IWL mixtures, avoiding collapse into pure ICL, IWL, or copying.

Prateek Garg, Lokesh Nagalapatti, Sunita Sarawagi

Algorithmic Recourse provides recommendations to individuals who are adversely impacted by automated model decisions, on how to alter their profiles to achieve a favorable outcome. Effective recourse methods must balance three conflicting goals: proximity to the original profile to minimize cost, plausibility for realistic recourse, and validity to ensure the desired outcome. We show that existing methods train for these objectives separately and then search for recourse through a joint optimization over the recourse goals during inference, leading to poor recourse recommendations. We introduce GenRe, a generative recourse model designed to train the three recourse objectives jointly. Training such generative models is non-trivial due to lack of direct recourse supervision. We propose efficient ways to synthesize such supervision and further show that GenRe's training leads to a consistent estimator. Unlike most prior methods, that employ non-robust gradient descent based search during inference, GenRe simply performs a forward sampling over the generative model to produce minimum cost recourse, leading to superior performance across multiple metrics. We also demonstrate GenRe provides the best trade-off between cost, plausibility and validity, compared to state-of-art baselines. Our code is available at: https://github.com/prateekgargx/genre.

Ashutosh Sathe, Sunita Sarawagi

Maximizing the likelihood of the next token is an established, statistically sound objective for pre-training language models. In this paper we show that we can train better models faster by pre-aggregating the corpus with a collapsed $n$-gram distribution. Previous studies have proposed corpus-level $n$-gram statistics as a regularizer; however, the construction and querying of such $n$-grams, if done naively, prove to be costly and significantly impede training speed, thereby limiting their application in modern large language model pre-training. We introduce an alternative compact representation of the next token distribution that, in expectation, aligns with the complete $n$-gram distribution while markedly reducing variance across mini-batches compared to the standard next-token loss. Empirically, we demonstrate that both the $n$-gram regularized model and our approximation yield substantial improvements in model quality and convergence rate compared to existing methods. Furthermore, our approximation facilitates scalability of gains to larger datasets and models compared to the straightforward $n$-gram regularization method.

Manasi Patwardhan, Ayush Agarwal, Shabbirhussain Bhaisaheb et al.

The performance of Large Language Models (LLMs) for translating Natural Language (NL) queries into SQL varies significantly across databases (DBs). NL queries are often expressed using a domain specific vocabulary, and mapping these to the correct SQL requires an understanding of the embedded domain expressions, their relationship to the DB schema structure. Existing benchmarks rely on unrealistic, ad-hoc query specific textual hints for expressing domain knowledge. In this paper, we propose a systematic framework for associating structured domain statements at the database level. We present retrieval of relevant structured domain statements given a user query using sub-string level match. We evaluate on eleven realistic DB schemas covering diverse domains across five open-source and proprietary LLMs and demonstrate that (1) DB level structured domain statements are more practical and accurate than existing ad-hoc query specific textual domain statements, and (2) Our sub-string match based retrieval of relevant domain statements provides significantly higher accuracy than other retrieval approaches.

Lokesh Nagalapatti, Ashutosh Srivastava, Sunita Sarawagi et al.

Diagnosing the root cause of an anomaly in a complex interconnected system is a pressing problem in today's cloud services and industrial operations. We propose In-Distribution Interventions (IDI), a novel algorithm that predicts root cause as nodes that meet two criteria: 1) **Anomaly:** root cause nodes should take on anomalous values; 2) **Fix:** had the root cause nodes assumed usual values, the target node would not have been anomalous. Prior methods of assessing the fix condition rely on counterfactuals inferred from a Structural Causal Model (SCM) trained on historical data. But since anomalies are rare and fall outside the training distribution, the fitted SCMs yield unreliable counterfactual estimates. IDI overcomes this by relying on interventional estimates obtained by solely probing the fitted SCM at in-distribution inputs. We present a theoretical analysis comparing and bounding the errors in assessing the fix condition using interventional and counterfactual estimates. We then conduct experiments by systematically varying the SCM's complexity to demonstrate the cases where IDI's interventional approach outperforms the counterfactual approach and vice versa. Experiments on both synthetic and PetShop RCD benchmark datasets demonstrate that \our\ consistently identifies true root causes more accurately and robustly than nine existing state-of-the-art RCD baselines. Code is released at https://github.com/nlokeshiisc/IDI_release.

Prathamesh Deshpande, Sunita Sarawagi

Long range forecasts are the starting point of many decision support systems that need to draw inference from high-level aggregate patterns on forecasted values. State of the art time-series forecasting methods are either subject to concept drift on long-horizon forecasts, or fail to accurately predict coherent and accurate high-level aggregates. In this work, we present a novel probabilistic forecasting method that produces forecasts that are coherent in terms of base level and predicted aggregate statistics. We achieve the coherency between predicted base-level and aggregate statistics using a novel inference method based on KL-divergence that can be solved efficiently in closed form. We show that our method improves forecast performance across both base level and unseen aggregates post inference on real datasets ranging three diverse domains. (\href{https://github.com/pratham16cse/AggForecaster}{Project URL})

Arjit Jain, Sunita Sarawagi, Prithviraj Sen

Given two large lists of records, the task in entity resolution (ER) is to find the pairs from the Cartesian product of the lists that correspond to the same real world entity. Typically, passive learning methods on such tasks require large amounts of labeled data to yield useful models. Active Learning is a promising approach for ER in low resource settings. However, the search space, to find informative samples for the user to label, grows quadratically for instance-pair tasks making active learning hard to scale. Previous works, in this setting, rely on hand-crafted predicates, pre-trained language model embeddings, or rule learning to prune away unlikely pairs from the Cartesian product. This blocking step can miss out on important regions in the product space leading to low recall. We propose DIAL, a scalable active learning approach that jointly learns embeddings to maximize recall for blocking and accuracy for matching blocked pairs. DIAL uses an Index-By-Committee framework, where each committee member learns representations based on powerful pre-trained transformer language models. We highlight surprising differences between the matcher and the blocker in the creation of the training data and the objective used to train their parameters. Experiments on five benchmark datasets and a multilingual record matching dataset show the effectiveness of our approach in terms of precision, recall and running time. Code is available at https://github.com/ArjitJ/DIAL

Kartik Khandelwal, Preethi Jyothi, Abhijeet Awasthi et al.

We introduce the problem of adapting a black-box, cloud-based ASR system to speech from a target accent. While leading online ASR services obtain impressive performance on main-stream accents, they perform poorly on sub-populations - we observed that the word error rate (WER) achieved by Google's ASR API on Indian accents is almost twice the WER on US accents. Existing adaptation methods either require access to model parameters or overlay an error-correcting module on output transcripts. We highlight the need for correlating outputs with the original speech to fix accent errors. Accordingly, we propose a novel coupling of an open-source accent-tuned local model with the black-box service where the output from the service guides frame-level inference in the local model. Our fine-grained merging algorithm is better at fixing accent errors than existing word-level combination strategies. Experiments on Indian and Australian accents with three leading ASR models as service, show that we achieve as much as 28% relative reduction in WER over both the local and service models.

Smit Jivani, Sarvam Maheshwari, Sunita Sarawagi

Large language models (LLMs) have revolutionized Text-to-SQL generation, allowing users to query structured data using natural language with growing ease. Yet, real-world deployment remains challenging, especially in complex or unseen schemas, due to inconsistent accuracy and the risk of generating invalid SQL. We introduce Template Constrained Decoding (TeCoD), a system that addresses these limitations by harnessing the recurrence of query patterns in labeled workloads. TeCoD converts historical NL-SQL pairs into reusable templates and introduces a robust template selection module that uses a fine-tuned natural language inference model to match or reject queries efficiently. Once the template is selected, TeCoD enforces it during SQL generation through grammar-constrained decoding, implemented via a novel partitioned strategy that ensures both syntactic validity and efficiency. Together, these components yield up to 36% higher execution accuracy than in-context learning (ICL) and 2.2x lower latency on matched queries.

Lokesh Nagalapatti, Akshay Iyer, Abir De et al.

We address the Individualized continuous treatment effect (ICTE) estimation problem where we predict the effect of any continuous-valued treatment on an individual using observational data. The main challenge in this estimation task is the potential confounding of treatment assignment with an individual's covariates in the training data, whereas during inference ICTE requires prediction on independently sampled treatments. In contrast to prior work that relied on regularizers or unstable GAN training, we advocate the direct approach of augmenting training individuals with independently sampled treatments and inferred counterfactual outcomes. We infer counterfactual outcomes using a two-pronged strategy: a Gradient Interpolation for close-to-observed treatments, and a Gaussian Process based Kernel Smoothing which allows us to downweigh high variance inferences. We evaluate our method on five benchmarks and show that our method outperforms six state-of-the-art methods on the counterfactual estimation error. We analyze the superior performance of our method by showing that (1) our inferred counterfactual responses are more accurate, and (2) adding them to the training data reduces the distributional distance between the confounded training distribution and test distribution where treatment is independent of covariates. Our proposed method is model-agnostic and we show that it improves ICTE accuracy of several existing models.

Annie D'souza, Swetha M, Sunita Sarawagi

Handling imbalance in class distribution when building a classifier over tabular data has been a problem of long-standing interest. One popular approach is augmenting the training dataset with synthetically generated data. While classical augmentation techniques were limited to linear interpolation of existing minority class examples, recently higher capacity deep generative models are providing greater promise. However, handling of imbalance in class distribution when building a deep generative model is also a challenging problem, that has not been studied as extensively as imbalanced classifier model training. We show that state-of-the-art deep generative models yield significantly lower-quality minority examples than majority examples. %In this paper, we start with the observation that imbalanced data training of generative models trained imbalanced dataset which under-represent the minority class. We propose a novel technique of converting the binary class labels to ternary class labels by introducing a class for the region where minority and majority distributions overlap. We show that just this pre-processing of the training set, significantly improves the quality of data generated spanning several state-of-the-art diffusion and GAN-based models. While training the classifier using synthetic data, we remove the overlap class from the training data and justify the reasons behind the enhanced accuracy. We perform extensive experiments on four real-life datasets, five different classifiers, and five generative models demonstrating that our method enhances not only the synthesizer performance of state-of-the-art models but also the classifier performance.

Ashwin Ramachandran, Sunita Sarawagi

Calibration is crucial as large language models (LLMs) are increasingly deployed to convert natural language queries into SQL for commercial databases. In this work, we investigate calibration techniques for assigning confidence to generated SQL queries. We show that a straightforward baseline -- deriving confidence from the model's full-sequence probability -- outperforms recent methods that rely on follow-up prompts for self-checking and confidence verbalization. Our comprehensive evaluation, conducted across two widely-used Text-to-SQL benchmarks and multiple LLM architectures, provides valuable insights into the effectiveness of various calibration strategies.

Prateek Garg, Bhavya Kohli, Sunita Sarawagi

Masked Diffusion Models (MDMs) have emerged as one of the most promising paradigms for generative modeling over discrete domains. It is known that MDMs effectively train to decode tokens in a random order, and that this ordering has significant performance implications in practice. This observation raises a fundamental question: can we design a training framework that optimizes for a favorable decoding order? We answer this in the affirmative, showing that the continuous-time variational objective of MDMs, when equipped with multivariate noise schedules, can identify and optimize for a decoding order during training. We establish a direct correspondence between decoding order and the multivariate noise schedule and show that this setting breaks invariance of the MDM objective to the noise schedule. Furthermore, we prove that the MDM objective decomposes precisely into a weighted auto-regressive losses over these orders, which establishes them as auto-regressive models with learnable orders.

Afrin Dange, Sunita Sarawagi

Time Series Foundation Models (TSFMs) have recently achieved state-of-the-art performance in univariate forecasting on new time series simply by conditioned on a brief history of past values. Their success demonstrates that large-scale pretraining across diverse domains can acquire the inductive bias to generalize from temporal patterns in a brief history. However, most TSFMs are unable to leverage covariates -- future-available exogenous variables critical for accurate forecasting in many applications -- due to their domain-specific nature and the lack of associated inductive bias. We propose TFMAdapter, a lightweight, instance-level adapter that augments TSFMs with covariate information without fine-tuning. Instead of retraining, TFMAdapter operates on the limited history provided during a single model call, learning a non-parametric cascade that combines covariates with univariate TSFM forecasts. However, such learning would require univariate forecasts at all steps in the history, requiring too many calls to the TSFM. To enable training on the full historical context while limiting TSFM invocations, TFMAdapter uses a two-stage method: (1) generating pseudo-forecasts with a simple regression model, and (2) training a Gaussian Process regressor to refine predictions using both pseudo- and TSFM forecasts alongside covariates. Extensive experiments on real-world datasets demonstrate that TFMAdapter consistently outperforms both foundation models and supervised baselines, achieving a 24-27\% improvement over base foundation models with minimal data and computational overhead. Our results highlight the potential of lightweight adapters to bridge the gap between generic foundation models and domain-specific forecasting needs.

Mayank Kothyari, Sunita Sarawagi, Soumen Chakrabarti et al.

LLMs are increasingly used as seq2seq translators from natural language utterances to structured programs, a process called semantic interpretation. Unlike atomic labels or token sequences, programs are naturally represented as abstract syntax trees (ASTs). Such structured representation raises novel issues related to the design and selection of in-context examples (ICEs) presented to the LLM. We focus on decomposing the pool of available ICE trees into fragments, some of which may be better suited to solving the test instance. Next, we propose how to use (additional invocations of) an LLM with prompted syntax constraints to automatically map the fragments to corresponding utterances. Finally, we adapt and extend a recent method for diverse ICE selection to work with whole and fragmented ICE instances. We evaluate our system, SCUD4ICL, on popular diverse semantic parsing benchmarks, showing visible accuracy gains from our proposed decomposed diverse demonstration method. Benefits are particularly notable for smaller LLMs, ICE pools having larger labeled trees, and programs in lower resource languages.

Lokesh Nagalapatti, Pranava Singhal, Avishek Ghosh et al.

Treatment effect estimation involves assessing the impact of different treatments on individual outcomes. Current methods estimate Conditional Average Treatment Effect (CATE) using observational datasets where covariates are collected before treatment assignment and outcomes are observed afterward, under assumptions like positivity and unconfoundedness. In this paper, we address a scenario where both covariates and outcomes are gathered after treatment. We show that post-treatment covariates render CATE unidentifiable, and recovering CATE requires learning treatment-independent causal representations. Prior work shows that such representations can be learned through contrastive learning if counterfactual supervision is available in observational data. However, since counterfactuals are rare, other works have explored using simulators that offer synthetic counterfactual supervision. Our goal in this paper is to systematically analyze the role of simulators in estimating CATE. We analyze the CATE error of several baselines and highlight their limitations. We then establish a generalization bound that characterizes the CATE error from jointly training on real and simulated distributions, as a function of the real-simulator mismatch. Finally, we introduce SimPONet, a novel method whose loss function is inspired from our generalization bound. We further show how SimPONet adjusts the simulator's influence on the learning objective based on the simulator's relevance to the CATE task. We experiment with various DGPs, by systematically varying the real-simulator distribution gap to evaluate SimPONet's efficacy against state-of-the-art CATE baselines.

Lokesh Nagalapatti, Pranava Singhal, Avishek Ghosh et al.

Given a dataset of individuals each described by a covariate vector, a treatment, and an observed outcome on the treatment, the goal of the individual treatment effect (ITE) estimation task is to predict outcome changes resulting from a change in treatment. A fundamental challenge is that in the observational data, a covariate's outcome is observed only under one treatment, whereas we need to infer the difference in outcomes under two different treatments. Several existing approaches address this issue through training with inferred pseudo-outcomes, but their success relies on the quality of these pseudo-outcomes. We propose PairNet, a novel ITE estimation training strategy that minimizes losses over pairs of examples based on their factual observed outcomes. Theoretical analysis for binary treatments reveals that PairNet is a consistent estimator of ITE risk, and achieves smaller generalization error than baseline models. Empirical comparison with thirteen existing methods across eight benchmarks, covering both discrete and continuous treatments, shows that PairNet achieves significantly lower ITE error compared to the baselines. Also, it is model-agnostic and easy to implement.

Vinit Unni, Shreya Khare, Ashish Mittal et al.

RNN-Transducer (RNN-T) models have become synonymous with streaming end-to-end ASR systems. While they perform competitively on a number of evaluation categories, rare words pose a serious challenge to RNN-T models. One main reason for the degradation in performance on rare words is that the language model (LM) internal to RNN-Ts can become overconfident and lead to hallucinated predictions that are acoustically inconsistent with the underlying speech. To address this issue, we propose a lightweight adaptive LM discounting technique AdaptLMD, that can be used with any RNN-T architecture without requiring any external resources or additional parameters. AdaptLMD uses a two-pronged approach: 1) Randomly mask the prediction network output to encourage the RNN-T to not be overly reliant on it's outputs. 2) Dynamically choose when to discount the implicit LM (ILM) based on rarity of recently predicted tokens and divergence between ILM and implicit acoustic model (IAM) scores. Comparing AdaptLMD to a competitive RNN-T baseline, we obtain up to 4% and 14% relative reductions in overall WER and rare word PER, respectively, on a conversational, code-mixed Hindi-English ASR task.

Vihari Piratla, Praneeth Netrapalli, Sunita Sarawagi

We consider the problem of training a classification model with group annotated training data. Recent work has established that, if there is distribution shift across different groups, models trained using the standard empirical risk minimization (ERM) objective suffer from poor performance on minority groups and that group distributionally robust optimization (Group-DRO) objective is a better alternative. The starting point of this paper is the observation that though Group-DRO performs better than ERM on minority groups for some benchmark datasets, there are several other datasets where it performs much worse than ERM. Inspired by ideas from the closely related problem of domain generalization, this paper proposes a new and simple algorithm that explicitly encourages learning of features that are shared across various groups. The key insight behind our proposed algorithm is that while Group-DRO focuses on groups with worst regularized loss, focusing instead, on groups that enable better performance even on other groups, could lead to learning of shared/common features, thereby enhancing minority performance beyond what is achieved by Group-DRO. Empirically, we show that our proposed algorithm matches or achieves better performance compared to strong contemporary baselines including ERM and Group-DRO on standard benchmarks on both minority groups and across all groups. Theoretically, we show that the proposed algorithm is a descent method and finds first order stationary points of smooth nonconvex functions.

Anshul Nasery, Soumyadeep Thakur, Vihari Piratla et al.

In several real world applications, machine learning models are deployed to make predictions on data whose distribution changes gradually along time, leading to a drift between the train and test distributions. Such models are often re-trained on new data periodically, and they hence need to generalize to data not too far into the future. In this context, there is much prior work on enhancing temporal generalization, e.g. continuous transportation of past data, kernel smoothed time-sensitive parameters and more recently, adversarial learning of time-invariant features. However, these methods share several limitations, e.g, poor scalability, training instability, and dependence on unlabeled data from the future. Responding to the above limitations, we propose a simple method that starts with a model with time-sensitive parameters but regularizes its temporal complexity using a Gradient Interpolation (GI) loss. GI allows the decision boundary to change along time and can still prevent overfitting to the limited training time snapshots by allowing task-specific control over changes along time. We compare our method to existing baselines on multiple real-world datasets, which show that GI outperforms more complicated generative and adversarial approaches on the one hand, and simpler gradient regularization methods on the other.

Vihari Piratla, Soumen Chakrabarty, Sunita Sarawagi

Our goal is to evaluate the accuracy of a black-box classification model, not as a single aggregate on a given test data distribution, but as a surface over a large number of combinations of attributes characterizing multiple test data distributions. Such attributed accuracy measures become important as machine learning models get deployed as a service, where the training data distribution is hidden from clients, and different clients may be interested in diverse regions of the data distribution. We present Attributed Accuracy Assay (AAA)--a Gaussian Process (GP)--based probabilistic estimator for such an accuracy surface. Each attribute combination, called an 'arm', is associated with a Beta density from which the service's accuracy is sampled. We expect the GP to smooth the parameters of the Beta density over related arms to mitigate sparsity. We show that obvious application of GPs cannot address the challenge of heteroscedastic uncertainty over a huge attribute space that is sparsely and unevenly populated. In response, we present two enhancements: pooling sparse observations, and regularizing the scale parameter of the Beta densities. After introducing these innovations, we establish the effectiveness of AAA in terms of both its estimation accuracy and exploration efficiency, through extensive experiments and analysis.

Yash Khemchandani, Sarvesh Mehtani, Vaidehi Patil et al.

Recent research in multilingual language models (LM) has demonstrated their ability to effectively handle multiple languages in a single model. This holds promise for low web-resource languages (LRL) as multilingual models can enable transfer of supervision from high resource languages to LRLs. However, incorporating a new language in an LM still remains a challenge, particularly for languages with limited corpora and in unseen scripts. In this paper we argue that relatedness among languages in a language family may be exploited to overcome some of the corpora limitations of LRLs, and propose RelateLM. We focus on Indian languages, and exploit relatedness along two dimensions: (1) script (since many Indic scripts originated from the Brahmic script), and (2) sentence structure. RelateLM uses transliteration to convert the unseen script of limited LRL text into the script of a Related Prominent Language (RPL) (Hindi in our case). While exploiting similar sentence structures, RelateLM utilizes readily available bilingual dictionaries to pseudo translate RPL text into LRL corpora. Experiments on multiple real-world benchmark datasets provide validation to our hypothesis that using a related language as pivot, along with transliteration and pseudo translation based data augmentation, can be an effective way to adapt LMs for LRLs, rather than direct training or pivoting through English.

Abhijeet Awasthi, Aman Kansal, Sunita Sarawagi et al.

We consider the task of personalizing ASR models while being constrained by a fixed budget on recording speaker-specific utterances. Given a speaker and an ASR model, we propose a method of identifying sentences for which the speaker's utterances are likely to be harder for the given ASR model to recognize. We assume a tiny amount of speaker-specific data to learn phoneme-level error models which help us select such sentences. We show that speaker's utterances on the sentences selected using our error model indeed have larger error rates when compared to speaker's utterances on randomly selected sentences. We find that fine-tuning the ASR model on the sentence utterances selected with the help of error models yield higher WER improvements in comparison to fine-tuning on an equal number of randomly selected sentence utterances. Thus, our method provides an efficient way of collecting speaker utterances under budget constraints for personalizing ASR models.

Parikshit Bansal, Prathamesh Deshpande, Sunita Sarawagi

We present DeepMVI, a deep learning method for missing value imputation in multidimensional time-series datasets. Missing values are commonplace in decision support platforms that aggregate data over long time stretches from disparate sources, and reliable data analytics calls for careful handling of missing data. One strategy is imputing the missing values, and a wide variety of algorithms exist spanning simple interpolation, matrix factorization methods like SVD, statistical models like Kalman filters, and recent deep learning methods. We show that often these provide worse results on aggregate analytics compared to just excluding the missing data. DeepMVI uses a neural network to combine fine-grained and coarse-grained patterns along a time series, and trends from related series across categorical dimensions. After failing with off-the-shelf neural architectures, we design our own network that includes a temporal transformer with a novel convolutional window feature, and kernel regression with learned embeddings. The parameters and their training are designed carefully to generalize across different placements of missing blocks and data characteristics. Experiments across nine real datasets, four different missing scenarios, comparing seven existing methods show that DeepMVI is significantly more accurate, reducing error by more than 50% in more than half the cases, compared to the best existing method. Although slower than simpler matrix factorization methods, we justify the increased time overheads by showing that DeepMVI is the only option that provided overall more accurate analytics than dropping missing values.

Prathamesh Deshpande, Kamlesh Marathe, Abir De et al.

In recent years, marked temporal point processes (MTPPs) have emerged as a powerful modeling machinery to characterize asynchronous events in a wide variety of applications. MTPPs have demonstrated significant potential in predicting event-timings, especially for events arriving in near future. However, due to current design choices, MTPPs often show poor predictive performance at forecasting event arrivals in distant future. To ameliorate this limitation, in this paper, we design DualTPP which is specifically well-suited to long horizon event forecasting. DualTPP has two components. The first component is an intensity free MTPP model, which captures microscopic or granular level signals of the event dynamics by modeling the time of future events. The second component takes a different dual perspective of modeling aggregated counts of events in a given time-window, thus encapsulating macroscopic event dynamics. Then we develop a novel inference framework jointly over the two models % for efficiently forecasting long horizon events by solving a sequence of constrained quadratic optimization problems. Experiments with a diverse set of real datasets show that DualTPP outperforms existing MTPP methods on long horizon forecasting by substantial margins, achieving almost an order of magnitude reduction in Wasserstein distance between actual events and forecasts.

Sahil Shah, Vihari Piratla, Soumen Chakrabarti et al.

State-of-the-art NLP inference uses enormous neural architectures and models trained for GPU-months, well beyond the reach of most consumers of NLP. This has led to one-size-fits-all public API-based NLP service models by major AI companies, serving large numbers of clients. Neither (hardware deficient) clients nor (heavily subscribed) servers can afford traditional fine tuning. Many clients own little or no labeled data. We initiate a study of adaptation of centralized NLP services to clients, and present one practical and lightweight approach. Each client uses an unsupervised, corpus-based sketch to register to the service. The server uses an auxiliary network to map the sketch to an abstract vector representation, which then informs the main labeling network. When a new client registers with its sketch, it gets immediate accuracy benefits. We demonstrate the success of the proposed architecture using sentiment labeling, NER, and predictive language modeling

Sriram Balasubramanian, Naman Jain, Gaurav Jindal et al.

We evaluate named entity representations of BERT-based NLP models by investigating their robustness to replacements from the same typed class in the input. We highlight that on several tasks while such perturbations are natural, state of the art trained models are surprisingly brittle. The brittleness continues even with the recent entity-aware BERT models. We also try to discern the cause of this non-robustness, considering factors such as tokenization and frequency of occurrence. Then we provide a simple method that ensembles predictions from multiple replacements while jointly modeling the uncertainty of type annotations and label predictions. Experiments on three NLP tasks show that our method enhances robustness and increases accuracy on both natural and adversarial datasets.

Abhijeet Awasthi, Sabyasachi Ghosh, Rasna Goyal et al.

In many applications labeled data is not readily available, and needs to be collected via pain-staking human supervision. We propose a rule-exemplar method for collecting human supervision to combine the efficiency of rules with the quality of instance labels. The supervision is coupled such that it is both natural for humans and synergistic for learning. We propose a training algorithm that jointly denoises rules via latent coverage variables, and trains the model through a soft implication loss over the coverage and label variables. The denoised rules and trained model are used jointly for inference. Empirical evaluation on five different tasks shows that (1) our algorithm is more accurate than several existing methods of learning from a mix of clean and noisy supervision, and (2) the coupled rule-exemplar supervision is effective in denoising rules.

Vihari Piratla, Praneeth Netrapalli, Sunita Sarawagi

Domain generalization refers to the task of training a model which generalizes to new domains that are not seen during training. We present CSD (Common Specific Decomposition), for this setting,which jointly learns a common component (which generalizes to new domains) and a domain specific component (which overfits on training domains). The domain specific components are discarded after training and only the common component is retained. The algorithm is extremely simple and involves only modifying the final linear classification layer of any given neural network architecture. We present a principled analysis to understand existing approaches, provide identifiability results of CSD,and study effect of low-rank on domain generalization. We show that CSD either matches or beats state of the art approaches for domain generalization based on domain erasure, domain perturbed data augmentation, and meta-learning. Further diagnostics on rotated MNIST, where domains are interpretable, confirm the hypothesis that CSD successfully disentangles common and domain specific components and hence leads to better domain generalization.

Oishik Chatterjee, Ganesh Ramakrishnan, Sunita Sarawagi

Scarcity of labeled data is a bottleneck for supervised learning models. A paradigm that has evolved for dealing with this problem is data programming. An existing data programming paradigm allows human supervision to be provided as a set of discrete labeling functions (LF) that output possibly noisy labels to input instances and a generative modelfor consolidating the weak labels. We enhance and generalize this paradigm by supporting functions that output a continuous score (instead of a hard label) that noisily correlates with labels. We show across five applications that continuous LFs are more natural to program and lead to improved recall. We also show that accuracy of existing generative models is unstable with respect to initialization, training epochs, and learning rates. We give control to the data programmer to guide the training process by providing intuitive quality guides with each LF. We propose an elegant method of incorporating these guides into the generative model. Our overall method, called CAGE, makes the data programming paradigm more reliable than other tricks based on initialization, sign-penalties, or soft-accuracy constraints.

Abhijeet Awasthi, Sunita Sarawagi, Rasna Goyal et al.

We present a Parallel Iterative Edit (PIE) model for the problem of local sequence transduction arising in tasks like Grammatical error correction (GEC). Recent approaches are based on the popular encoder-decoder (ED) model for sequence to sequence learning. The ED model auto-regressively captures full dependency among output tokens but is slow due to sequential decoding. The PIE model does parallel decoding, giving up the advantage of modelling full dependency in the output, yet it achieves accuracy competitive with the ED model for four reasons: 1.~predicting edits instead of tokens, 2.~labeling sequences instead of generating sequences, 3.~iteratively refining predictions to capture dependencies, and 4.~factorizing logits over edits and their token argument to harness pre-trained language models like BERT. Experiments on tasks spanning GEC, OCR correction and spell correction demonstrate that the PIE model is an accurate and significantly faster alternative for local sequence transduction.

Prathamesh Deshpande, Sunita Sarawagi

We present ARU, an Adaptive Recurrent Unit for streaming adaptation of deep globally trained time-series forecasting models. The ARU combines the advantages of learning complex data transformations across multiple time series from deep global models, with per-series localization offered by closed-form linear models. Unlike existing methods of adaptation that are either memory-intensive or non-responsive after training, ARUs require only fixed sized state and adapt to streaming data via an easy RNN-like update operation. The core principle driving ARU is simple --- maintain sufficient statistics of conditional Gaussian distributions and use them to compute local parameters in closed form. Our contribution is in embedding such local linear models in globally trained deep models while allowing end-to-end training on the one hand, and easy RNN-like updates on the other. Across several datasets we show that ARU is more effective than recently proposed local adaptation methods that tax the global network to compute local parameters.

Vihari Piratla, Sunita Sarawagi, Soumen Chakrabarti

Given a small corpus $\mathcal D_T$ pertaining to a limited set of focused topics, our goal is to train embeddings that accurately capture the sense of words in the topic in spite of the limited size of $\mathcal D_T$. These embeddings may be used in various tasks involving $\mathcal D_T$. A popular strategy in limited data settings is to adapt pre-trained embeddings $\mathcal E$ trained on a large corpus. To correct for sense drift, fine-tuning, regularization, projection, and pivoting have been proposed recently. Among these, regularization informed by a word's corpus frequency performed well, but we improve upon it using a new regularizer based on the stability of its cooccurrence with other words. However, a thorough comparison across ten topics, spanning three tasks, with standardized settings of hyper-parameters, reveals that even the best embedding adaptation strategies provide small gains beyond well-tuned baselines, which many earlier comparisons ignored. In a bold departure from adapting pretrained embeddings, we propose using $\mathcal D_T$ to probe, attend to, and borrow fragments from any large, topic-rich source corpus (such as Wikipedia), which need not be the corpus used to pretrain embeddings. This step is made scalable and practical by suitable indexing. We reach the surprising conclusion that even limited corpus augmentation is more useful than adapting embeddings, which suggests that non-dominant sense information may be irrevocably obliterated from pretrained embeddings and cannot be salvaged by adaptation.

Aviral Kumar, Sunita Sarawagi

We study the calibration of several state of the art neural machine translation(NMT) systems built on attention-based encoder-decoder models. For structured outputs like in NMT, calibration is important not just for reliable confidence with predictions, but also for proper functioning of beam-search inference. We show that most modern NMT models are surprisingly miscalibrated even when conditioned on the true previous tokens. Our investigation leads to two main reasons -- severe miscalibration of EOS (end of sequence marker) and suppression of attention uncertainty. We design recalibration methods based on these signals and demonstrate improved accuracy, better sequence-level calibration, and more intuitive results from beam-search.

Shiv Shankar, Vihari Piratla, Soumen Chakrabarti et al.

We present CROSSGRAD, a method to use multi-domain training data to learn a classifier that generalizes to new domains. CROSSGRAD does not need an adaptation phase via labeled or unlabeled data, or domain features in the new domain. Most existing domain adaptation methods attempt to erase domain signals using techniques like domain adversarial training. In contrast, CROSSGRAD is free to use domain signals for predicting labels, if it can prevent overfitting on training domains. We conceptualize the task in a Bayesian setting, in which a sampling step is implemented as data augmentation, based on domain-guided perturbations of input instances. CROSSGRAD parallelly trains a label and a domain classifier on examples perturbed by loss gradients of each other's objectives. This enables us to directly perturb inputs, without separating and re-mixing domain signals while making various distributional assumptions. Empirical evaluation on three different applications where this setting is natural establishes that (1) domain-guided perturbation provides consistently better generalization to unseen domains, compared to generic instance perturbation methods, and that (2) data augmentation is a more stable and accurate method than domain adversarial training.

Srayanta Mukherjee, Devashish Shankar, Atin Ghosh et al.

Accurate demand forecasts can help on-line retail organizations better plan their supply-chain processes. The challenge, however, is the large number of associative factors that result in large, non-stationary shifts in demand, which traditional time series and regression approaches fail to model. In this paper, we propose a Neural Network architecture called AR-MDN, that simultaneously models associative factors, time-series trends and the variance in the demand. We first identify several causal features and use a combination of feature embeddings, MLP and LSTM to represent them. We then model the output density as a learned mixture of Gaussian distributions. The AR-MDN can be trained end-to-end without the need for additional supervision. We experiment on a dataset of an year's worth of data over tens-of-thousands of products from Flipkart. The proposed architecture yields a significant improvement in forecasting accuracy when compared with existing alternatives.

Shiv Shankar, Sunita Sarawagi

Augmenting a neural network with memory that can grow without growing the number of trained parameters is a recent powerful concept with many exciting applications. We propose a design of memory augmented neural networks (MANNs) called Labeled Memory Networks (LMNs) suited for tasks requiring online adaptation in classification models. LMNs organize the memory with classes as the primary key.The memory acts as a second boosted stage following a regular neural network thereby allowing the memory and the primary network to play complementary roles. Unlike existing MANNs that write to memory for every instance and use LRU based memory replacement, LMNs write only for instances with non-zero loss and use label-based memory replacement. We demonstrate significant accuracy gains on various tasks including word-modelling and few-shot learning. In this paper, we establish their potential in online adapting a batch trained neural network to domain-relevant labeled data at deployment time. We show that LMNs are better than other MANNs designed for meta-learning. We also found them to be more accurate and faster than state-of-the-art methods of retuning model parameters for adapting to domain-specific labeled data.

Pavel Sountsov, Sunita Sarawagi

Encoder-decoder networks are popular for modeling sequences probabilistically in many applications. These models use the power of the Long Short-Term Memory (LSTM) architecture to capture the full dependence among variables, unlike earlier models like CRFs that typically assumed conditional independence among non-adjacent variables. However in practice encoder-decoder models exhibit a bias towards short sequences that surprisingly gets worse with increasing beam size. In this paper we show that such phenomenon is due to a discrepancy between the full sequence margin and the per-element margin enforced by the locally conditioned training objective of a encoder-decoder model. The discrepancy more adversely impacts long sequences, explaining the bias towards predicting short sequences. For the case where the predicted sequences come from a closed set, we show that a globally conditioned model alleviates the above problems of encoder-decoder models. From a practical point of view, our proposed model also eliminates the need for a beam-search during inference, which reduces to an efficient dot-product based search in a vector-space.

Aman Madaan, Sunita Sarawagi

The problem of collecting reliable estimates of occurrence of entities on the open web forms the premise for this report. The models learned for tagging entities cannot be expected to perform well when deployed on the web. This is owing to the severe mismatch in the distributions of such entities on the web and in the relatively diminutive training data. In this report, we build up the case for maximum mean discrepancy for estimation of occurrence statistics of entities on the web, taking a review of named entity disambiguation techniques and related concepts along the way.