Papers by Shrey Desai
Compressive Summarization with Plausibility and Salience Modeling (2020.emnlp-main)
Copied to clipboard
| Challenge: | a new method to learn which compressions to apply is based on syntactic rules for deleting spans . plausibility and salience are the two main criteria for determining which compression to apply . a recent study shows that the plausability model generally selects for grammatical and factual deletions compared to extractive methods . |
| Approach: | They propose to leave the decision about what to delete to two data-driven criteria . they show that plausibility and salience are the most important criteria if a span is deleted . |
| Outcome: | The proposed method achieves strong in-domain results on benchmark datasets and human evaluation shows that plausibility model generally selects for grammatical and factual deletions. |
Retrieve-and-Fill for Scenario-based Task-Oriented Semantic Parsing (2023.eacl-main)
Copied to clipboard
Akshat Shrivastava, Shrey Desai, Anchit Gupta, Ali Elkahky, Aleksandr Livshits, Alexander Zotov, Ahmed Aly
| Challenge: | Task-oriented semantic parsing models have achieved strong results in recent years, but they often face obstacles adapting to novel settings with distinct semantics and scarce data. |
| Approach: | They propose a scenario-based semantic parsing model which isolates coarse-grained and fine-grounded aspects of the task and solves them with off-the-shelf neural modules. |
| Outcome: | The proposed model outperforms previous approaches in high-resource, low-resourced, and multilingual settings, and is modular, differentiable, interpretable, and allows extra supervision from scenarios. |
Span Pointer Networks for Non-Autoregressive Task-Oriented Semantic Parsing (2021.findings-emnlp)
Copied to clipboard
Akshat Shrivastava, Pierce Chuang, Arun Babu, Shrey Desai, Abhinav Arora, Alexander Zotov, Ahmed Aly
| Challenge: | a novel approach to map utterances to semantic frames is based on non-autoregressive parsers that shift the decoding task from text generation to span prediction. |
| Approach: | They propose a non-autoregressive, task-oriented parser which shifts the decoding task from text generation to span prediction and produces endpoints as opposed to text. |
| Outcome: | The proposed model bridges the quality gap between non-autoregressive and autoregressive parsers, achieving 87 EM on TOPv2 and shows a 70% reduction in latency and 83% reduction in memory at beam size 5 compared to prior non-regressives. |
Evaluating Lottery Tickets Under Distributional Shifts (D19-61)
Copied to clipboard
| Challenge: | Recent research suggests deep neural networks are dramatically over-parametrized. |
| Approach: | They propose that large, over-parameterized neural networks consist of small, sparse subnetworks that can be trained in isolation to reach a similar (or better) test accuracy. |
| Outcome: | The proposed models can achieve commensurate performance using the same initialization as the original model. |
Detecting Perceived Emotions in Hurricane Disasters (2020.acl-main)
Copied to clipboard
| Challenge: | Existing methods for emotion detection are limited in disaster-centric domains due to distributional shifts. |
| Approach: | They propose to use a Twitter emotion dataset to analyze emotions in natural disasters . they propose to apply classification tasks to discriminate between coarse-grained emotions . |
| Outcome: | The proposed model achieves only 68% accuracy after pre-training with unlabeled Twitter data. |
Calibration of Pre-trained Transformers (2020.emnlp-main)
Copied to clipboard
| Challenge: | Pre-trained Transformers dominate benchmark tasks but use a large number of self-attention heads across many layers in a way that is difficult to unpack. |
| Approach: | They analyze pre-trained Transformer models' posterior probabilities to determine whether they are calibrated for three tasks: natural language inference, paraphrase detection, and commonsense reasoning. |
| Outcome: | The models are calibrated in-domain and out-of-domain, and their calibration error out-domain can be as much as 3.5x lower. |
Diagnosing Transformers in Task-Oriented Semantic Parsing (2021.findings-acl)
Copied to clipboard
| Challenge: | a recent study shows transformer-based parsers struggle with disambiguating intents/slots and producing syntactically valid frames. |
| Approach: | They propose to use seq2seq transformers to map textual utterances to semantic frames . they propose to model transformer-based parsers across monolingual and multilingual settings . |
| Outcome: | The proposed parsers struggle with disambiguating intents/slots and produce syntactically valid frames. |
Understanding Neural Abstractive Summarization Models via Uncertainty (2020.emnlp-main)
Copied to clipboard
| Challenge: | Recent advances in abstractive summarization have been fueled by the advent of large-scale Transformers pre-trained on autoregressive language modeling objectives. |
| Approach: | They analyze summarization decoders in both blackbox and whitebox ways by studying on the entropy, or uncertainty, of the model’s token-level predictions. |
| Outcome: | The proposed model generates tokens in a free-form manner, but this flexibility makes it difficult to interpret their behavior. |
Adaptive Ensembling: Unsupervised Domain Adaptation for Political Document Analysis (D19-1)
Copied to clipboard
| Challenge: | a new study examines the use of labeled and unlabeled corpora in political science research . large corporata often contain documents of a certain subject or type, but they are often unlabed . a recent study found that labeles with pertinent documents stem from a single source . |
| Approach: | They propose an unsupervised domain adaptation framework that uses a text classification model and time-aware training to ensure it works well with diachronic corpora. |
| Outcome: | The proposed framework outperforms benchmarks on an expert-annotated dataset and is more stable and learns better representations. |