Getting to Production with Few-shot Natural Language Generation Models
- P. Heidari, Arash Einolghozati, Michael White
- Computer ScienceSIGDIAL Conferences
- 2021
This paper introduces a system consisting of iterative self-training and an extensible mini-template framework that textualizes the structured input data into semi-natural text to fully take advantage of pre-trained language models to enable few-shot Natural Language Generation.
Structure-to-Text Generation with Self-Training, Acceptability Classifiers and Context-Conditioning for the GEM Shared Task
- Shreyan Bakshi, Soumya Batra, P. Heidari, A. Arun, Shashank Jain, Michael White
- Computer ScienceIEEE Games Entertainment Media Conference
- 2021
It is found that self-training with reconstruction matching along with acceptability classifier filtering can improve semantic correctness, though gains are limited in the full-data setting.
Best Practices for Data-Efficient Modeling in NLG:How to Train Production-Ready Neural Models with Less Data
- A. Arun, Soumya Batra, Michael White
- Computer ScienceInternational Conference on Computational…
- 8 November 2020
A family of sampling and modeling techniques to attain production quality with light-weight neural network models using only a fraction of the data that would be necessary otherwise, and results show that domain complexity dictates the appropriate approach to achieve high data efficiency.
Interactive Dance Lessons through Human Body Pose Estimation and Skeletal Topographies Matching
- S. Deb, Alpana Sharan, Shivangi Chaturvedi, A. Arun, Aayush Gupta
- Computer Science
- 2018
This work device a learning game for teaching simple dance moves to children which encourages them to perform a pose to match the given dance move, leveraging the power of Deep Learning to perform the same task.
Building Adaptive Acceptability Classifiers for Neural NLG
- Soumya Batra, Shashank Jain, Michael White
- Computer ScienceConference on Empirical Methods in Natural…
- 2021
It is shown that building acceptability classifiers using data that resembles the generation model outputs followed by a validation framework outperforms the existing techniques, achieving state-of-the-art results.
A Survey on Current Semantic level Algorithms for improving Performance in CBIR
- A. Arun, P. Nirmaladevi
- Computer Science
- 1 February 2021
A complete review of recent topical methodological attainments in the investigation region of CBIR related to extraction of content features in different sized of the vectors, distance matrices, connect bridge between complexities and added with objective of the applications improvement such as precision and recall rate.
Semantic Role Labeling for Process Recognition Questions
- Samuel Louvan, Chetan Naik, Veronica E. Lynn, A. Arun, Niranjan Balasubramanian, Peter Clark
- Computer Science
- 2015
Empirical evaluation shows that manually generated roles provide a 12% relative improvement in accuracy over a simpler bag-of-words representation, but automatic role identification is noisy and doesn’t provide gains even with distant supervision and domain adaptation modifications to account for the limited training data.
Improving Faithfulness of Abstractive Summarization by Controlling Confounding Effect of Irrelevant Sentences
- Asish Ghoshal, Arash Einolghozati, Asli Celikyilmaz
- Computer ScienceArXiv
- 19 December 2022
This paper gives a principled characterization of data distributions where such confounding can be large thereby necessitating the use of human annotated relevant sentences to generate factual summaries and designs a simple multi-task model to control such confounding.