Papers by Preethi Jyothi

Adversarial Training for Low-Resource Disfluency Correction (2023.findings-acl)

Challenge:	Using recurrent neural networks to build language models for code-switched text is an important problem with implications to downstream applications such as speech recognition and machine translation.
Approach:	They propose a novel recurrent neural network unit with dual components that focus on each language in the code-switched text separately and a generative model estimated using the training data.
Outcome:	The proposed techniques yield significant reductions in perplexity on Mandarin-English task and improve on baseline models.

Copied to clipboard

Vineet Bhat, Preethi Jyothi, Pushpak Bhattacharyya

Responsible & Trustworthy NLP

Challenge:	Disfluencies can be introduced in conversational speech due to the conversational nature of speech and/or speech impairments such as stuttering.
Approach:	They propose an adversarial sequence-tagging model for Disfluency Correction . they evaluate it in Bengali, Hindi, and Marathi languages and use it to correct stuttering disfluencies .
Outcome:	The proposed technique improves in Bengali, Hindi, and Marathi languages . it also removes stuttering disfluencies in ASR transcripts introduced by speech impairments .

SrcMix: Mixing of Related Source Languages Benefits Extremely Low-resource Machine Translation (2026.findings-eacl)

Copied to clipboard

Sanjeev Kumar, Preethi Jyothi, Pushpak Bhattacharyya

Boosting Zero-Shot Crosslingual Performance using LLM-Based Augmentations with Effective Data Selection (2024.findings-acl)

Challenge:	Multilingual models are widely used for machine translation, but their effectiveness for extremely low-resource languages (ELRLs) is dependent on how related languages are incorporated during fine-tuning.
Approach:	They propose a source-side mixing strategy that combines related ELRLs during fine-tuning while constraining the decoder to a single target language.
Outcome:	The proposed approach improves performance in high-resource to ELRL translations and in mid-resourced to MT translations.

Copied to clipboard

Barah Fazili, Ashish Agrawal, Preethi Jyothi

Low-Resource NLP

Challenge:	Large language models generate task-specific data via zero-shot prompting and promote cross-lingual transfer for low-resource target languages.
Approach:	They propose to use a teacher model to label LLM generations and use their label probabilities to identify a representative subset of diverse generations that boost zero-shot accuracies while being efficient.
Outcome:	The proposed models generate task-specific data via zero-shot prompting and promote cross-lingual transfer for low-resource target languages.

CoSTA: Code-Switched Speech Translation using Aligned Speech-Text Interleaving (2025.coling-main)

Copied to clipboard

Bhavani Shankar P S V N, Preethi Jyothi, Pushpak Bhattacharyya

Improving Language Identification for Code-Switched Speech: The Pivotal Role of Accented English (2026.findings-eacl)

Challenge:	More than half of the world's population is presumed to be bilingual . spoken translation of code-switched speech has been under-explored .
Approach:	They propose an end-to-end model architecture CoSTA that scaffolds on pretrained ASR and MT modules.
Outcome:	The proposed model outperforms existing models by 3.5 BLEU points in spoken translation of code-switched speech.

Copied to clipboard

Adyasha Patra, Dhiraj Kumar Sah, Preethi Jyothi

CoCoa: An Encoder-Decoder Model for Controllable Code-switched Generation (2022.emnlp-main)

Challenge:	Existing models fail to identify English spoken with the accent of the matrix (dominant) language.
Approach:	They propose to fine tune existing LID models with accented English to improve code-switched LID . they use a metric that captures relative ranking of identified languages often overlooked by traditional metrics.
Outcome:	The proposed model can be fine tuned with small amounts of accented English without degrading performance on monolingual speech.

Copied to clipboard

Sneha Mondal, null Ritika, Shreya Pathak, Preethi Jyothi, Aravindan Raghuveer

Accented Speech Recognition With Accent-specific Codebooks (2023.emnlp-main)

Challenge:	Generating code-switched text with fine-grained control on the degree of code-witching and the lexical choices used to convey formality has been well-explored.
Approach:	They propose to generate code-switched text with fine-grained control on the degree of code-changing and lexical choices used to convey formality.
Outcome:	The proposed model can be invoked at test-time to synthesize code-switched text faithful to syntactic and lexical attributes relevant to code-witching.

Copied to clipboard

Darshan Prabhu, Preethi Jyothi, Sriram Ganapathy, Vinit Unni

Meta-Learning for Effective Multi-task and Multilingual Modelling (2021.eacl-main)

Challenge:	Degradation in performance across underrepresented accents is a severe deterrent to inclusive adoption of ASR.
Approach:	They propose an approach to adapt speech accents to unseen accents by using cross-attention with a trainable set of codebooks.
Outcome:	The proposed approach yields significant performance gains on the seen English accents and unseen accents on the Mozilla Common Voice dataset.

Copied to clipboard

Ishan Tarunesh, Sushil Khyalia, Vishwajeet Kumar, Ganesh Ramakrishnan, Preethi Jyothi

DISCO: A Large Scale Human Annotated Corpus for Disfluency Correction in Indo-European Languages (2023.findings-emnlp)

Challenge:	Existing studies on multitask and multilingual learning have shown that learning cross-lingual embeddings can benefit multiple tasks and languages.
Approach:	They propose a meta-learning approach to learn interactions between tasks and languages . they also investigate the role of different sampling strategies used during meta-learned model .
Outcome:	The proposed model improves on five different tasks and six different languages from the XTREME multilingual benchmark dataset.

Copied to clipboard

Vineet Bhat, Preethi Jyothi, Pushpak Bhattacharyya

Text Generation

Challenge:	Existing research on disfluency correction has primarily focused on English due to the unavailability of large-scale open-source datasets.
Approach:	They propose to use an annotated human-annotated corpus to analyze disfluency correction in four important Indo-European languages to demonstrate the benefits.
Outcome:	The proposed model improves BLEU scores by 5.65 points when used with a state-of-the-art machine translation system.

Improving Pretraining Techniques for Code-Switched NLP (2023.acl-long)

Copied to clipboard

Richeek Das, Sahasra Ranjan, Shreya Pathak, Preethi Jyothi

Semantic Text Processing

Challenge:	Multilingual pretraining models for code-switched inputs are a key component of NLP applications.
Approach:	They propose to use masked language modeling techniques to mask code-switched text that are cognizant of language boundaries prior to masking.
Outcome:	The proposed techniques improve performance on two downstream tasks, Question Answering (QA) and Sentiment Analysis (SA), compared to standard pretraining techniques.

DITTO: Data-efficient and Fair Targeted Subset Selection for ASR Accent Adaptation (2023.acl-long)

Copied to clipboard

Suraj Kothawade, Anmol Mekala, D.Chandra Sekhara Hetha Havya, Mayank Kothyari, Rishabh Iyer, Ganesh Ramakrishnan, Preethi Jyothi

Translation Errors Significantly Impact Low-Resource Languages in Cross-Lingual Learning (2024.eacl-short)

Challenge:	State-of-the-art automatic speech recognition systems exhibit disparate performance on varying speech accents.
Approach:	They propose to use submodular mutual information to find the most informative set of utterances matching a target accent within a fixed budget.
Outcome:	The proposed model is 3-5 times more label-efficient on the Indic-TTS and L2 datasets than other methods.

Copied to clipboard

Ashish Agrawal, Barah Fazili, Preethi Jyothi

Automatic Speech Recognition in Sanskrit: A New Speech Corpus and Modelling Insights (2021.findings-acl)

Challenge:	XNLI benchmarks use parallel versions of English evaluation sets in multiple target languages . a recent study found that translation errors exist in some low-resource languages resulting in incorrect estimates of cross-lingual transfer .
Approach:	They propose to measure the gap in performance between zero-shot evaluations on human-translated and machine-transcribed target text across multiple target languages.
Outcome:	The proposed benchmarks show that translation errors exist for Hindi and Urdu . the results corroborate previous studies that found translation errors in Hindi and urdu despite translation errors.

Copied to clipboard

Devaraja Adiga, Rishabh Kumar, Amrith Krishna, Preethi Jyothi, Ganesh Ramakrishnan, Pawan Goyal

How Accents Confound: Probing for Accent Information in End-to-End Speech Recognition Systems (2020.acl-main)

Challenge:	In this paper, we propose the first large scale study of automatic speech recognition in Sanskrit . we focus on the impact of unit selection in San's ASR systems .
Approach:	They propose a large scale study of automatic speech recognition in Sanskrit . they propose syllable level unit selection that captures character sequences .
Outcome:	The proposed model captures character sequences from one vowel in the word to the next vowela.

Copied to clipboard

Archiki Prasad, Preethi Jyothi

AMPS: ASR with Multimodal Paraphrase Supervision (2025.naacl-short)

Challenge:	a new study examines how accent information is encoded and propagated in an end-to-end ASR system.
Approach:	They propose to use phone probes to analyze phonetic content of representations at each layer.
Outcome:	The proposed model is based on a large amount of US-accented English speech and is compared with other models using phone probes.

Copied to clipboard

Abhishek Gupta, Amruta Parulekar, Sameep Chattopadhyay, Preethi Jyothi

Text Generation

Challenge:	Spontaneous or conversational multilingual speech presents many challenges for state-of-the-art automatic speech recognition systems.
Approach:	They propose a technique that augments a multilingual multimodal ASR system with paraphrase-based supervision for improved conversational ASR in multiple languages.
Outcome:	The proposed technique reduces word error rates by up to 5% on a state-of-the-art multimodal model .

LASER: An LLM-based ASR Scoring and Evaluation Rubric (2025.emnlp-main)

Copied to clipboard

Amruta Parulekar, Preethi Jyothi

Revisiting the Importance of Encoding Logic Rules in Sentiment Classification (D18-1)

Challenge:	Standard ASR evaluation metrics like word error rate penalize morphological and syntactic nuances that do not significantly alter sentence semantics.
Approach:	They propose an LLM-based scoring rubric LASER that leverages state-of-the-art LLMs’ in-context learning abilities to learn from prompts with detailed examples.
Outcome:	The proposed scoring rubric combines state-of-the-art learning capabilities with a high correlation score with human annotations.

Copied to clipboard

Kalpesh Krishna, Preethi Jyothi, Mohit Iyyer

Information Retrieval

Challenge:	Neural models that explicitly encode word order, syntax and semantic features are unequipped to deal with complex syntactic structures that affect sentiment, such as contrastive conjunctions.
Approach:	They propose to use contextualized ELMo embeddings instead of logic rules to compare models on syntactically complex sentences.
Outcome:	The proposed models outperform baseline models even on sentences with ambiguous sentiment labels.

Part-of-speech Tagging for Extremely Low-resource Indian Languages (2024.findings-acl)

Copied to clipboard

Sanjeev Kumar, Preethi Jyothi, Pushpak Bhattacharyya

Syntactic Text Processing

Challenge:	Modern natural language processing systems thrive when given access to large datasets, but a large fraction of the world’s languages are not privy to such benefits due to sparse documentation and inadequate digital representation.
Approach:	They propose a parallel part-of-speech evaluation dataset for Angika, Magahi, Bhojpuri and Hindi.
Outcome:	The proposed approach improves F1 scores by up to 8% on Angika, Magahi, Bhojpuri and Hindi while ignoring the tokenization challenge.

STORiCo: Storytelling TTS for Hindi with Character Voice Modulation (2024.eacl-short)

Copied to clipboard

Pavan Tankala, Preethi Jyothi, Preeti Rao, Pushpak Bhattacharyya

DeFT-X: Denoised Sparse Fine-Tuning for Zero-Shot Cross-Lingual Transfer (2025.findings-emnlp)

Challenge:	Existing datasets for read speech for Hindi lack expressiveness and character voice consistency.
Approach:	They propose to use a Hindi text-to-speech (TTS) dataset to train a multi-speaker model on the single-sector data and propose to improve expressiveness and character voice consistency.
Outcome:	The proposed model improves expressiveness and character voice consistency compared to the baseline single-speaker model.

Copied to clipboard

Sona Elza Simon, Preethi Jyothi

Post-ASR Correction in Hindi: Comparing Language Models and Large Language Models in Low-Resource Scenarios (2026.eacl-short)

Challenge:	Prior studies have explored multiple approaches to combine task knowledge from task-specific data in a (high-resource) source language with language knowledge from unlabeled text in 'low-resourced' target language.
Approach:	They propose a composable sparse fine-tuning approach that learns task-specific and language-specific sparsen masks to select a subset of the pretrained model's parameters.
Outcome:	The proposed approach performs at par or outperforms SFT and other prominent cross-lingual transfer baselines.

Copied to clipboard

Rishabh Kumar, Amrith Krishna, Ganesh Ramakrishnan, Preethi Jyothi

Speech-enriched Memory for Inference-time Adaptation of ASR Models to Word Dictionaries (2023.emnlp-main)

Challenge:	Automatic Speech Recognition (ASR) systems for low-resource languages produce erroneous transcripts due to limited annotated data and linguistic complexity.
Approach:	They compare language models and large language models for post-ASR correction in Hindi . they observe a scaling trend under zero-shot ICL where mid-sized LLMs degrade performance before marginal recovery at extreme scales.
Outcome:	The proposed model outperforms larger models in both fine-tuning and in-context learning settings.

Copied to clipboard

Ashish Mittal, Sunita Sarawagi, Preethi Jyothi, George Saon, Gakuto Kurata

DIMSIM: Distilled Multilingual Critics for Indic Text Simplification (2024.findings-acl)

Challenge:	Existing contextual biasing techniques require additional parameterization . state-of-the-art ASR systems often fail to recognize named entities or critical rare words .
Approach:	They propose an algorithm that uses nearest-neighbor matching to predict ASR models . a list of rare entities is indexed in a memory and then stored the best possible match .
Outcome:	The proposed algorithm improves the prediction of state-of-the-art ASR models on rare words . it prevents spurious matches by restricting to word-level matches .

Copied to clipboard

Sneha Mondal, Ritika Ritika, Ashish Agrawal, Preethi Jyothi, Aravindan Raghuveer

In-context Mixing (ICM): Code-mixed Prompts for Multilingual LLMs (2024.acl-long)

Challenge:	Existing approaches to improve the quality of responses generated by large language models (LLMs) however, these critique-refine steps require multiple expensive LLM calls.
Approach:	They propose to use critique distillation to train critic models that are trained on input-critique pairs generated by an LLM.
Outcome:	The proposed model trains two separate critics that focus on lexical and structure complexity, and is more effective than using an LLM directly as a critic in both 0-shot and few-shot settings.

Copied to clipboard

Bhavani Shankar, Preethi Jyothi, Pushpak Bhattacharyya

Low-Resource NLP

Challenge:	In-context mixing is a prompting technique for effective in-contact learning with multilingual large language models.
Approach:	They propose a prompting technique called in-context mixing for effective in-constext learning with multilingual large language models.
Outcome:	The proposed prompts perform better with multilingual large language models.

Partitioned Gradient Matching-based Data Subset Selection for Compute-Efficient Robust ASR Training (2022.findings-emnlp)

Copied to clipboard

Ashish Mittal, Durga Sivasubramanian, Rishabh Iyer, Preethi Jyothi, Ganesh Ramakrishnan

Responsible & Trustworthy NLP

Challenge:	Existing DSS algorithms for RNN-T have a high cost and performance degradation.
Approach:	They propose a distributable DSS algorithm for RNN-T that can be used to train a subset of training data.
Outcome:	The proposed algorithm achieves between 3x to 6x speedup with only a small accuracy degradation even in settings where the training data is corrupted with noise.

From Machine Translation to Code-Switching: Generating High-Quality Code-Switched Text (2021.acl-long)

Copied to clipboard

Ishan Tarunesh, Syamantak Kumar, Preethi Jyothi

Accurate Online Posterior Alignments for Principled Lexically-Constrained Decoding (2022.acl-long)

Challenge:	a computational model for code-switching text is lacking in the corpus of real text.
Approach:	They propose a neural machine translation model to generate Hindi-English code-switched sentences using monolingual Hindi sentences.
Outcome:	The proposed model reduces perplexity on a language modeling task and improves on linguistic inference tasks.

Copied to clipboard

Soumya Chatterjee, Sunita Sarawagi, Preethi Jyothi

Machine Translation

Challenge:	Existing methods of offline alignment use only the entire target sentence.
Approach:	They propose a posterior alignment technique that is truly online in its execution and superior in terms of alignment error rates compared to existing methods.
Outcome:	The proposed technique is online in execution and superior in alignment error rates compared to existing methods.

Aligning Multilingual Embeddings for Improved Code-switched Natural Language Understanding (2022.coling-1)

Copied to clipboard

Barah Fazili, Preethi Jyothi

Zero-shot Cross-lingual Transfer With Learned Projections Using Unlabeled Target-Language Data (2023.acl-short)

Challenge:	a recent study has shown that multilingual models can be effective on monolingual data but need additional training to work well with code-switched text.
Approach:	They propose to train multilingual models with alignment objectives using parallel text . they find such an explicit alignment step improves performance on code-switched NLP tasks .
Outcome:	The proposed model improves on Hindi-English Sentiment Analysis, Named Entity Recognition and Question Answering tasks.

Copied to clipboard

Ujan Deb, Ridayesh Parab, Preethi Jyothi

Responsible & Trustworthy NLP

Challenge:	Zero-shot cross-lingual transfer is enabled by pairing the language adapter in the target language with an appropriate task adapter within a source language.
Approach:	They propose to use unlabeled text to enhance zero-shot transfer by pairing language adapters with task adapters in a target language.
Outcome:	The proposed framework improves on three cross-lingual tasks with up to 11% relative improvement in Named Entity Recognition (NER), Question Answering (QA) and Natural Language Inference (NLI).

LexGen: Domain-aware Multilingual Lexicon Generation (2025.acl-long)

Copied to clipboard

Ayush Maheshwari, Atul Kumar Singh, N J Karthika, Krishnakant Bhatt, Preethi Jyothi, Ganesh Ramakrishnan

Zero-shot Disfluency Detection for Indian Languages (2022.coling-1)

Challenge:	Lexicon generation is a key task in specialized domains due to infrequent usage of terms . a new model is proposed to generate dictionary words for 6 Indian languages .
Approach:	They propose a model to generate dictionary words for 6 Indian languages in the multi-domain setting.
Outcome:	The proposed model generalizes to unseen domains and unsealed languages.

Copied to clipboard

Rohit Kundu, Preethi Jyothi, Pushpak Bhattacharyya

LoFTI: Localization and Factuality Transfer to Indian Locales (2025.findings-acl)

Challenge:	Disfluency correction models can help alleviate this problem, but the unavailability of labeled data in low-resource languages impairs progress.
Approach:	They propose to use a pretrained multilingual model to detect zero-shot disfluency in Indian languages.
Outcome:	The proposed model achieves F1 scores of 75 and higher on five disfluency types across four languages.

Copied to clipboard

Sona Elza Simon, Soumen Kumar Mondal, Abhishek Singhania, Sayambhu Sen, Preethi Jyothi