Introductory class.

This blog post is about the introductory class by Prof. S. Stober of the Otto-von Guericke University, Magdeburg. I think I really like the concept of the flipped classroom! 

During this class, he explained how the course is going to be conducted and the tools we are going to use. There are a couple of things that stood out in the introductory class:

a. The models learn the features by themselves! This takes quite an amount of load off from the shoulders of the practitioner!

b. The idea of making a blog post is really cool! I already have a blog www.anirbansaha.com and generally, I do make logs of what I study. That's mainly because of the awesome memory the Almighty has gifted me. Writing down helps me remember things.

c. I really liked the phrase "Positive failure culture".

There are a few things I am a little worried about. I'm not sure how much time I would be able to dedicate to this. But since right now, I have nothing else to do except for the implementation of Dr. Fallucchi's work, I think this would be one opportunity I wouldn't get the second time. I can use all the time into doing this. 

In the introductory class, Prof. Stober asked us to set a mini-goal for us to achieve. I think it's a little complicated for me, although fairly simple when seen individually.

On a personal level, I would like to work more towards building chatbots. That would involve a lot of things which include: 

a. Retrieving answers from a chunk of text OR retrieving the most relevant document from a bank of documents. E.g: choose the most relevant blog post based on a user's query.

b. Formulating the response to the user. 

c. Remembering the context of the previous conversation. 

d. Understanding user's messages. 

While this is not an exhaustive list of tasks in a chatbot, each one of them would open to an area of research. Since I still want to build chatbots, I would like to pursue at least one of the previous objectives.

Another objective could be the work at my disposal! I am supposed to extract Document structure from PDF textbooks. While I am able to convert the PDF to XML with the tags in the markup language, I am not sure how to proceed with it. Most researchers have used supervised statistical learning methods. So yea, this too would be my lookout.