REU Site: Machine Learning in Natural Language Processing and Computer Vision

REU 网站：自然语言处理和计算机视觉中的机器学习

• 批准号: 1659788
• 负责人: Jugal Kalita
• 金额: $ 37.99万
• 依托单位: University of Colorado at Colorado Springs
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2017
• 资助国家: 美国
• 起止时间: 2017-06-01 至 2020-05-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1659788&HistoricalAwards=false
• 关键词: REU; Site; Machine; Learning; Natural

The proposal seeks to increase the number of American citizens and permanent resident undergraduates who are attracted to careers in research and advanced studies in Computer Science. Training in theoretical and empirical machine learning will enable participants to contribute to ubiquitous software-based technologies at the highest levels in innovative ways. The proposal will also focus almost exclusively on training future computer scientists from institutions with limited research opportunities, women and under-represented minorities. The research experience will encourage these students to be productive researchers in academic and non-academic environments during their future careers. Computer vision, through applications such as face recognition and autonomous vehicles, has profound implications for society in terms of future of national security and transportation. Natural language processing, through applications such as efficient large volume semantic analysis and summarization of textual documents from disparate sources, also has implications for national security in addition to numerous other possibilities that can be exploited by industry. The objective of this proposal is to expose bright and motivated undergraduates who want to pursue advanced careers in Computer Science to active research experience early in their careers. This proposal seeks to develop an REU site to broaden the intellectual horizon of participants through exposure to opportunities available in university research. Students will be involved in research projects in machine learning techniques and in emerging applications that exploit machine learning. The applications of interest are in the fields of natural language processing and computer vision. Students will have the opportunity to utilize a combination of theoretical reading, analysis, and research within laboratory environments. The students will be introduced to the topics and helped to obtain in-depth understanding of selected topics through introductory presentations. Then, they will perform hands-on research on novel problems, conduct experiments, and communicate their results through written papers and presentations. The REU students will be involved in research in machine learning and its applications in diverse and emergent areas alongside faculty mentors and graduate students in a university environment. The research will involve undergraduate students in cutting-edge research where they will write software to solve interesting and timely problems and write papers for publication.

该提案旨在增加美国公民和永久居民本科生的数量，他们被计算机科学的研究和高级研究所吸引。理论和经验机器学习的培训将使参与者能够以创新的方式在最高水平上为无处不在的基于软件的技术做出贡献。该提案还将几乎完全侧重于培训未来的计算机科学家，他们来自研究机会有限的机构、妇女和代表性不足的少数民族。研究经验将鼓励这些学生在未来的职业生涯中成为学术和非学术环境中的富有成效的研究人员。通过人脸识别和自动驾驶汽车等应用，计算机视觉在未来的国家安全和交通方面对社会产生了深远的影响。自然语言处理，通过应用程序，如有效的大容量语义分析和总结文本文件从不同的来源，也有影响国家安全，除了许多其他可能性，可以利用行业。 该提案的目的是让那些希望在计算机科学领域从事高级职业的聪明而有动力的本科生在职业生涯早期获得积极的研究经验。这项建议旨在开发一个可再生能源大学网站，通过接触大学研究中提供的机会，拓宽参与者的知识视野。 学生将参与机器学习技术的研究项目和利用机器学习的新兴应用。 感兴趣的应用是在自然语言处理和计算机视觉领域。 学生将有机会在实验室环境中利用理论阅读，分析和研究的组合。 学生将被介绍给主题，并通过介绍性演讲帮助他们深入了解选定的主题。然后，他们将对新问题进行实践研究，进行实验，并通过书面论文和演示来传达他们的结果。REU学生将参与机器学习及其在不同和新兴领域的应用研究，与大学环境中的教师导师和研究生一起。这项研究将涉及本科生在前沿研究，他们将编写软件来解决有趣和及时的问题，并撰写论文发表。

项目成果

Parallel Attention Mechanisms in Neural Machine Translation

• DOI: 10.1109/icmla.2018.00088
• 发表时间: 2018-10
• 期刊: 2018 17th IEEE International Conference on Machine Learning and Applications (ICMLA)
• 作者: Julian R. Medina;J. Kalita

Improving Computer Generated Dialog with Auxiliary Loss Functions and Custom Evaluation Metrics

• 发表时间: 2021-06
• 作者: T. Conley

Open Set Text Classification using Convolutional Neural Networks

使用卷积神经网络进行开放集文本分类

• 发表时间: 2017
• 作者: Prakhya, Sridhama;Venkataram, Vinodini;Kalita, Jugal
• 通讯作者: Kalita, Jugal

Isolated and Ensemble Audio Preprocessing Methods for Detecting Adversarial Examples against Automatic Speech Recognition

用于检测自动语音识别的对抗性示例的隔离和集成音频预处理方法

• 发表时间: 2018
• 期刊: Conference on Computational Linguistics and Speech Processing (ROCLING
• 作者: Rajaratnam, Krishan;Shah, Kunal;Kalita, Jugal
• 通讯作者: Kalita, Jugal

Noise Flooding for Detecting Audio Adversarial Examples Against Automatic Speech Recognition

• DOI: 10.1109/isspit.2018.8642623
• 发表时间: 2018-12
• 期刊: 2018 IEEE International Symposium on Signal Processing and Information Technology (ISSPIT)
• 作者: K. Rajaratnam;J. Kalita