Hello, I would like to work on *"**GSoC_2015 : Better Accessibility for Okular"*. I have also mailed the mentors mentioned (*Mr. Albert and Mr. Frederik*) on the ideas page. I have a prior experience of working on a related problem wherein I have created *Hidden Markov Models* for detecting the varied speech word samples with ~*95%* accuracy.The code for the same was developed in C++ language.
This project has fascinated me and I have also worked on deep neural networks, whose recent advancement has enabled Baidu for creating benchmarks in voice related inputs. Hence I am really excited to take up this idea as my GSoC project. I have the following thoughts on how to deal with this project : 1. To preserve the semantics, and other NLP related tasks, one could use the Hidden Markov Model to extend the states for the speech samples. (However, this can have the problem associated with the training of the sample set, which would take relatively longer of time) 2. To employ Deep Learning technique (window-based approach) using Artificial Neural Network to enable the Semantic Role Labeling in a much faster and precise manner, which has a theoretically simple approach but rigorous code has to be written. 3. Detecting the text word size and then predicting if the semantics associated is a heading or a new point, for the matter. This would employ the usage of data structure such that it has the past k word sizes, current word size and color, apart from the following k1 words' sizes (this abstraction would also suffice, which would be a build-up from the usual robotic text-to-speech conversion softwares) I have *cloned* *git://anongit.kde.org/okular <http://anongit.kde.org/okular>* and have successfully* built* the same after resolving the dependencies. Could you please suggest how I should go about in further structuring the problem statement? I have a very good developing experience with C++, Python and Java; and I am thorough and fluent in the same. I have also used system version control for my projects during my internships. I would be obliged if I could be guided to fine-tune my approach for this GSoC project. Regards, Vishal Anand, Ex-Summer Analyst at Goldman Sachs Final year CSE student, IIT Guwahati, India. Resume : www.vishalanand.net/resume [Apologies if you are receiving multiple copies of this email as the previous mail could not be sent out to the complete mailing-list]
>> Visit http://mail.kde.org/mailman/listinfo/kde-devel#unsub to unsubscribe <<
