Speech:Spring 2013 Bego Terzimustafic Log


 * Home
 * Semesters
 * Spring 2013
 * Proposal
 * Report

Week Ending February 5th, 2013
Feb 3: Read logs and notes. Will plan on doing research on various features of Sphinx and see if it would be worthwhile to upgrade.
 * Task:


 * Results:


 * Plan:


 * Concerns:

Week Ending February 12, 2013
My task for the current week was to make comparisons between Sphinx 3.x and Sphinx 4 in order to make a decision on which one we should be using.
 * Task:

Feb 10 - Read user logs and notes. Did more research on the differences between Sphinx 3 and 4.
 * Results:

Feb 11 - After reading several different sources and doing some research online. I have come to the conclusion that we should continue using Sphinx. The Sphinx 4 engine was completely rewritten in Java. It included multi-threading capabilities that Sphinx 3 is lacking. The question was whether it would be viable to upgrade Sphinx to utilize the hardware more effectively to provide faster, and more accurate results. I looked at several different sources as well as at Mike Bianci's log to help me find an answer to this.

I read a paper in regards to Sphinx 4 and the various advantages and this is a small excerpt: “Another interesting aspect of the performance study shows us that raw computing speed is not our biggest concern when it comes to RT performance. For the 2 CPU results in this table, we used a Scorer that equally divided the scoring task across the available CPUs. While the increase in speed is noticeable, it is not as dramatic as we expected. Further analysis helped us determine that only about 30 percent of the CPU time is spent doing the actual scoring of the acoustic model states. The remaining 70 percent is spent doing non-scoring activity, such as growing and pruning the ActiveList. Our empirical results also show that the Java platform’s garbage collection mechanism only accounts for 2-3 percent of the overall CPU usage.” Source - http://cmusphinx.sourceforge.net/sphinx4/doc/Sphinx4Whitepaper.pdf

It states that despite the increase in processors, the differences in performance wasn't as dramatic as they considered. Looking at charts as well on http://cmusphinx.sourceforge.net/sphinx4/#speed_and_accuracy reflects on that as well. Mike Brown states on our shared folder on Google Drive that the dictionary size is 127128 words, Sphinx 4 hasn't been properly tested on a dictionary of that size.

Feb 12: I forgot to mention that I had done work after class in regards to the sphinx research. Tomorrow before class I will update the links on the current version of the various parts of the sphinx package. Today I continued reading notes and logs of other individuals.


 * Plan:

I'm a bit worried about the direction that we're heading. I'm not completely sure what my next task will be, but I'll talk to my teammates tomorrow and figure it all out with them. My other concern is just making sure that I understood everything that I read and that I was able to make a sound decision based on that, I'll talk to my teammates about that as well.
 * Concerns:

Week Ending February 19, 2013
Write a proposal for upgrading the current version of Sphinx to the newest one.
 * Task:

February 17: Read notes and comments
 * Results:

February 18: I worked a bit on making the Proposal for upgrading Sphinx to version 4. My current format includes an introduction of both versions, their key features, performance, and possible installation time requirements. I looked over the sources that I used and I plan on providing citations for information that I gather. I also continued reading logs.

February 19: I'm finished with the first draft of the proposal. I don't where to upload it exactly, but I'll bring it to class tomorrow and show the entirety of it to Jonas, as well as my group.

Spend some time reading and writing a proposal in regards to upgrading Sphinx. Formatting of the proposal is the only concern I have. There's no real guideline to how I should make it, or all the questions that Professor Jonas will ask in regards to it. I plan on being informed as much as I can in regards to the subject. The proposal is not overly verbose, I did cite all my sources that have all the information in much greater detail if there is a need to through more information.
 * Plan:
 * Concerns:

Week Ending February 26, 2013
My task for this week is to do research on Sphinx 3.8 and see if that would be a more viable solution than Sphinx 4.
 * Task:

February 24: Read other's notes and logs.
 * Results:


 * Plan:


 * Concerns:

Week Ending March 5, 2013
My task for this week is to continue doing research on Sphinx 3.8. We had no class last week.
 * Task:

March 3: I have required extra time on doing research on Sphinx 3.8 since I have had trouble finding any major information on it and what kind of impact it would make. I haven't been able to find release notes on it either. I may have to speak to Lee and see if we can try a quick upgrade. In addition, the snow day has caused some trouble and there is a little bit of confusion in regards to what is happening. I have also continued reading other's notes and seeing what they were up to. I had to make a quick adjustment to my previous log as well.
 * Results:

March 4: Spent some time reading other's logs.

March 5: I spent some more time trying to find some form of changelog or information on Sphinx 3.8. I downloaded the version itself and looked through some text files and found nothing. The changelog file that was included in the root folder also was totally outdated. I continued to read other's logs as well during this time.

I need to discuss further things with my partners and make sure that we stay on track. I need to provide a good way to display my proposal for an upgrade as well. We lost quite a bit, having a snow day last weekend. But I have a feeling that with a lot of discussion my team and I will be able to get right back on track and make progress.
 * Plan:
 * Concerns:

Week Ending March 12, 2013

 * Task:


 * Results:


 * Plan:


 * Concerns:

Week Ending March 26, 2013

 * Task:

March 25: I read some notes from other students. I also have spoken to Tommy in regards to what I should attempt to do for next week, he has instructed me to look at a tutorial that the group has attempted on Sunday. I was not present for that due to a prior obligation that I had, but I made sure to let my team members know about that.
 * Results:


 * Plan:


 * Concerns:

Week Ending April 2, 2013

 * Task:

March 31: Spend a google hangout with my team and we discussed our current situations. I made it clear that I may require some help during class time. The hangout was spent mostly discussing the training that had to be accomplished.
 * Results:


 * Plan:


 * Concerns:

Week Ending April 9, 2013

 * Task:

April 7: I've gotten in touch with Tommy and have continued to work on getting the mini train to run. So far I've been having some form of connectivity issues, but everything has been set up and the only part that is left is just getting the decode to go through. There shouldn't be too many other issues for me other than that. Last class I was able to properly catch up and now I only have to deal with the finishing touches. I'll keep in contact with Tommy as I make progress in regards to this.
 * Results:

April 9: I forgot to make a log for April 8. I went to school and was able to get the decoding to go through! I was having trouble at my house, for some reason I was continuously disconnected after 10-15 minutes, and it would never finish. I came up with another error while trying to score. I mentioned this to Tommy, went home, read some notes and forgot to write a log about it. I was going to make another attempt today.

As for what I have done on April 9. I have tried once more to get the scoring to work. I am getting an, "sclite: Error, Hypthesis file 'hyp.trans' does not exist" error message. I have contacted Tommy and Tyler for help. I am currently communicating with Tyler about this and Tommy, and see if they make any headway.


 * Plan:


 * Concerns:

Week Ending April 16, 2013

 * Task:

April 13: Read some logs, kept up with the e-mails between the two groups. It looks like we'll be doing a google hangout possibly tomorrow. I will get in contact with Tommy and see if he will need anything done by me.
 * Results:

April 14: Spoke to Tommy and told him my availability for a google hang out if one is going to be organized. Read some more notes.

April 15: We did a google hangout and worked on the project. It looks like we're making progress and according to Tyler we'd be able to run the decoder by tomorrow morning. I had to leave the meeting after around 10:30.


 * Plan:


 * Concerns:

Week Ending April 23, 2013

 * Task:

April 21: Read logs and kept up with the e-mail conversation
 * Results:


 * Plan:


 * Concerns:

Week Ending April 30, 2013

 * Task:

April 27: Read logs.
 * Results:


 * Plan:


 * Concerns:

Week Ending May 7, 2013

 * Task:

May 4: Read notes and logs. May 5: Read notes and logs. Will be working on the final paper tomorrow as well. May 6: Continued to read notes and logs. I also worked on the paper and I posted a question on there. I'll probably work on fine-tuning the paper tomorrow and continue doing so after Wednesday. May 7: Read more notes and logs.
 * Results:
 * Plan:


 * Concerns: