This is my first presentation for my lecture and talk about Deep Learning. It is focused on Word Embedding. I tried to well organize the contents of word embedding techniques in my presentation.
http://web.donga.ac.kr/yjko/talks/DL-WordEmbedding(Youngjoong%20Ko).pdf
Wednesday, September 9, 2015
Wednesday, June 10, 2015
My accepted paper in ACL 2015
My paper related to SLU was accepted in ACL 2015. Its abstract is as follows:
The intelligent personal assistant software such as the Apple’s Siri and Samsung’s S-Voice has been issued these days. This paper introduces a novel Spoken Language Understanding (SLU) module to predict user’s intention for determining system actions of the intelligent personal assistant software. The SLU module usually consists of several connected recognition tasks on a pipeline framework. The proposed SLU module has four recognition tasks including named entity, speech-act, target and operation recognition, and it is developed to simultaneously recognize the four tasks on a recognition framework using Conditional Random Fields (CRF). In the experiments, the new simultaneous recognition method achieves the higher performance of 4% and faster speed of about 25% than other method using pipeline. By a significance test, this improvement is considered to be statistically significant as a p-value of smaller than 0.05.
ACL 2015 will be held in Beijing, China, July 26-31, 2015. I will post the camera-ready version of my paper when it is ready to be published.
The intelligent personal assistant software such as the Apple’s Siri and Samsung’s S-Voice has been issued these days. This paper introduces a novel Spoken Language Understanding (SLU) module to predict user’s intention for determining system actions of the intelligent personal assistant software. The SLU module usually consists of several connected recognition tasks on a pipeline framework. The proposed SLU module has four recognition tasks including named entity, speech-act, target and operation recognition, and it is developed to simultaneously recognize the four tasks on a recognition framework using Conditional Random Fields (CRF). In the experiments, the new simultaneous recognition method achieves the higher performance of 4% and faster speed of about 25% than other method using pipeline. By a significance test, this improvement is considered to be statistically significant as a p-value of smaller than 0.05.
ACL 2015 will be held in Beijing, China, July 26-31, 2015. I will post the camera-ready version of my paper when it is ready to be published.
Friday, May 15, 2015
topic modeling (LDA)
Before Deep Learning becomes so popular, most researchers had a great interest in topic modeling or LDA. But because LDA requires some knowledge of statistics, many researchers have thought that it is not easy to understand or learn. So I tried to make a ppt material including the fundamental knowledge of both statistics and LDA. Actually, this has been used in my NLP class.
Please check out the following link.
http://web.donga.ac.kr/yjko/usefulthings/TopicModeling.pdf
And my next topic is word embedding for NLP as a deep learning technique.
Please check out the following link.
http://web.donga.ac.kr/yjko/usefulthings/TopicModeling.pdf
And my next topic is word embedding for NLP as a deep learning technique.
Subscribe to:
Posts (Atom)