Speech:Spring 2015 Krista Cleary Log


 * Home
 * Semesters
 * Spring 2015
 * Proposal
 * Report
 * Information - General Project Information
 * Experiments - List of speech experiments

Week Ending February 3, 2015

 * Task:

1/31:  Read previous semesters logs and start working on proposal for this semesters data team.

2/1:  Write proposal draft and get a more comfortable feel for what this project will entail.

2/2:  Meet with group to discuss next steps for our group. Also discuss the proposal and what needs to be done for class on Wednesday.

2/3:  Meet with group to finalize proposal draft.


 * Results:

1/31:  Read through Spring 2014 Data Team's logs.

2/1:  Read previous semesters proposal to see what their goal was.

2/2:  Met with group and discussed our proposal. Combined different group members proposals, so the draft will be ready for Wednesday.

2/3:  Met with group. Sent proposal back and forth a couple of times to each. Wrote individual timeline for proposal. Our group sent ours to the proposal team tonight so they can post it on the mediawiki page.


 * Plan:


 * Concerns:

I am concerned that we haven't received our logins for Caesar yet, but I'm sure that will be taken care of during the next class.

Week Ending February 10, 2015

 * Task:
 * Add new tasks to the proposal document
 * Download SSH Secure Shell onto personal computer
 * Login to Caesar
 * Start documenting where current .wav files are located and mark which ones need renaming and/or removal


 * Results:


 * Successfully logged into Caesar and Asterix
 * Downloaded and used SSH Secure Shell on my personal computer
 * Looked through corpus file structure


 * Plan:

2/4: This semester the data group has some clear tasks that we want to accomplish. We want to delete .wav files in the experiments prior to Spring 2014 or experiment 0150. Ee want to clean up the corpus files. We want to rename .wav files with more descriptive names and create soft links for experiments group to use this semester. We also have been tasked with getting to the bottom of whether the original switchboard file is 308 hours or 247 hours. We will then document how we came to this conclusion and why. We also need to compare the current dictionaries with any new dictionaries that the Tools group is considering.

I just finished downloading SSH Secure Shell onto my personal computer. I then was able to successfully log into Caesar using my UNH username. I then went and changed my password to my current UNH email password. I then was able to log into Asterix.

2/8: Today I spent some time catching up on this semesters logs to see where other students are in their research and planning. I also read through my team members personal logs. I logged into Caesar and started looking around at the corpus directory.

2/9: Today I looked at the corpus directory and started writing out the current structure of the files.

2/10: Today, our group met and discussed what our individual tasks should be. We are currently unsure of how the tasks should be split as it doesn't appear that the Data group has enough tasks to be split up among four team members. We also discussed the possibility of writing a script to add soft links to previous experiments as creating the soft links is one of our groups main tasks.


 * Concerns:

That there aren't enough tasks for all four of us.

Week Ending February 17, 2015

 * Tasks:


 * Results:

2/12: Updated draft proposal with more objectives that were brought up during class on Wednesday. I updated my personal timeline for the proposal. The rest of our team also updated their personal timelines. After all the personal timelines were updated, we sent the proposal to Trevor to be uploaded into the proposal section of MediaWiki. At this point I would say that our draft is pretty well structured and will most likely not need very many updates after today. Below is my updated personal timeline.

Krista

Week Ending Feb 3rd

Meet with group members to work out draft proposal.

Week Ending Feb 10th

Gain access to Caesar server and familiarize myself with current audio file structure.

Locate .wav files and check for duplicates.

Week Ending Feb 17th

Update/Create file structure documentation.

Work up plan with group for how we want to reorganize all the .wav files.

Document the new structure.

Create new file structure.

Week Ending Feb 24th

Continue reorganization of audio files.

Remove .wav files from previous experiments.

Create soft links to the new file structure.

Week Ending March 3rd

Complete any outstanding work and documentation on the organization system.

Attempt to train and run an experiment.

2/14: Today I read through my group's individual logs to see what they have been up to this week. I also read through various other classmates logs. I read up on some basic Linux commands.

2/16: Today I spent a lot of time learning about soft links and how they work. I also read through Dakota's logs since he seems to really understand what the soft links are all about and how they work.

2/17: Tonight, my group met up via Google Hangouts to assign out tasks for each individual in the group. As of now, we aren't sure how difficult each of the tasks is going to be, so if one task is easier than another we will most likely help each other out with that. We felt our group had 4 main tasks: running an experiment, setting up the soft links, clean up old .wav files, and determine the size of the switchboard corpus. Below are how we have decided to split up the tasks.

Stephen - running an experiment Dakota - soft links Krista - .wav file cleanup Russ - size of switchboard

In class tomorrow, we hope to get some clarification from Professor Jonas about how he would like the soft linking structure set up.


 * Plan:


 * Assign individual tasks
 * Learn more about soft linking
 * Learn more shell commands


 * Concerns:


 * How the soft links are supposed to be set up

Week Ending February 24, 2015

 * Task:


 * Finish final proposal
 * Locate all .wav files prior to Spring 2014
 * Update information wiki page


 * Results:

2/18: Today our team met with the Proposal group to discuss what needed to be updated in our proposal. There were only a couple edits that needed to be made. We need to remove the individual timeline section and add a tasks section that outlines the group objectives and when they will be done and what group members helped. Second we need to remove the bullet points at the beginning of the proposal. We also needed to mention the different corpora. And finally we needed to update the language to show that we will be completing tasks not just attempting to. Today, I worked on updating the proposal language while Russ worked on the tasks. Once the tasks were outlined, we added tentative dates and who is working on what.

2/19: Today I finished updating the proposal. The proposal now has everything that the Proposal team asked for from us. Hopefully there won't be too many more changes. As a team, we are feeling really good about our section of the proposal. Unfortunately, we are not only graded on our section, so hopefully the rest of the teams have been working hard on their respective sections.

2/22: Today I spent the day going through all the experiment directories on Caesar to determine which experiments still have .wav files directly located within them. I was able to find seven experiments that still have .wav files in them. Interestingly all of them are located in the 130-142 range, which are the experiments that are not documented in the Experiment wiki page. I'm going to attempt to figure out what the experiments were trying to do in hope of being able to bridge that gap. I will be asking for clarification from Professor Jonas about whether I can just delete the .wav files or if I need to replace them with soft links since he was a little unclear on that part of the task.

2/24: Last night I met with my group to discuss what we've been working on this week. Today, I read through all of Marcel's logs to gather more information on the NOAA corpus so we can create an information page for it.


 * Plan:

2/18 - 2/19 Work on completing proposal.

2/20 - 2/21 Work on locating all .wav files.

2/22 - 2/24 Update information wiki page.


 * Concerns:

I'm concerned that other groups won't pull together to make a great proposal.

Week Ending March 3, 2015

 * Task:
 * Update soft links found in /mnt/main/corpus/switchboard/dist/flat
 * Add NOAA information page to mediawiki


 * Results:

2/25: Today during our meeting, Professor Jonas said we no longer need to delete the remaining .wav files prior to Spring 2014. Over the weekend I had found that the experiments listed below still have .wav files existing in them.

0135    0136     0137     0138     0139     0140     0142

Since my task in the Data group was to remove the .wav files and that is no longer necessary, I will be helping Dakota correct the soft links. To correct a soft link, you use the command ln - fs (where the link should point to) (where the soft link is located). Below is an example:

ln -fs /mnt/main/corpus/switchboard/dist/disk2/swb1/sw02064.sph /mnt/main/corpus/switchboard/dist/flat/sw02064.sph

Today, I was able to fix soft links from 02064 through 02154. Hopefully, I will be able to complete even more tomorrow.

2/26: This morning, I started working on the soft links located in /mnt/main/corpus/switchboard/dist/flat/. I was going through and changing each cmd line one at a time for each link, but I thought there must be an easier way. I ended up figuring out that if I separated each line with a semicolon, I could paste in more than one line at a time. So I went to the directory I was fixing. I did ls. Then I copied the list of softlinks into excel. I then did a text to column to get each soft link name in it's own cell. I deleted the soft links that were already fixed. I then combined all the columns into one and concatenated "ln -fs /mnt/main/corpus/switchboard/dist/disk#/swb1/", soft link name, " /mnt/main/corpus/switchboard/dist/flat/", soft link name, and ";". I was then able to copy my list from excel into notepad. I then copy and pasted about 30 at a time into SSH Secure Shell. This made the entire process go by much quicker. Since I was able to complete /mnt/main/corpus/switchboard/dist/flat/ so quickly, I then moved on to the soft links in /mnt/main/corpus/switchboard/full/clean/audio/conv/.

2/27: Dakota emailed me today asking if I could fix the 256hr directory as when he was working on it, it somehow ended up with too many files. He removed all the soft links in the directory so I could start from scratch. It was much easier starting from scratch than trying to figure out where the extra files came from. 256hr now contains 2285 soft links pointing to the flat directory. Since, I worked on the soft links, Dakota said he would start working on the NOAA information page.

3/3: Met with my group on google hangouts tonight. We discussed the progress we've made over the last week. The soft links are all set and fixed. The NOAA data wiki page has been started. Stephen was able to complete his first train and decode.


 * Plan:

2/25-2/27: Update soft links

2/28-3/3: Update NOAA information page


 * Concerns:

I don't have any concerns at this time.

Week Ending March 10, 2015

 * Task:
 * fix switchboard structure
 * standardize the structure so all the directories are the same
 * update NOAA wiki


 * Results:

3/4: Today, we had our status meeting with Professor Jonas. He gave us a new file structure for the file directories in the corpus/switchboard directory. Professor Jonas also wanted all the "clean" directories to be removed. Each directory should be set up like below.
 * test/
 * train/
 * audio/
 * conv
 * utt
 * info/
 * trans/

3/5: Today, we cleaned up the file structure for the below directories. 125hr_3170 256hr first_5hr full

We made the following changes so that the directories would follow the new structure. 256hr Renamed "clean" directory to "train" Added "test" directory Added "info" directory under "train" 125hr_3170 Renamed "clean" directory to "train" first_5hr Renamed "clean" directory to "train" Deleted "mono" directory (only contained soft links back to "clean" directory) Deleted "monoRaw" directory (only contained soft links back to "clean" directory) full Deleted "full/audio/conv" (empty directory) Deleted "full/dev" (empty directory) Deleted "full/eval" (empty directory) Deleted "full/mono/trans" (empty directory) Moved "full_transcript.text" from "full/train2/trans" to "clean/trans" Deleted "full/train2/trans" Deleted "full/train2/wav" (only contained the same soft links as "full/clean/audio/conv"  Deleted "full/train2/audio" (only contained soft links that pointed to "full/train2/wav" Deleted "full/train2" Moved "full/mono/audio/utt" to "full/clean/audio/utt" Deleted "full/mono" Deleted "full/train" (contained same broken soft links as "full/clean"  Renamed "clean" directory to "full" directory

The only thing left to do is to figure out more about utt. The directory "full/audio/utt" has not been deleted yet, because it contains different files than "full/train/audio/utt".

3/8: Today, I spend the day reading through the directions on how to run a train. I'm hoping to make my first attempt at it tomorrow or Tuesday.

3/10: I did not get around to running my first experiment tonight like I had hoped. I read through other students logs today. I'm hoping we will be doing a boot camp for the experiments.


 * Plan:


 * Concerns:

Week Ending March 24, 2015

 * Task:
 * Meet with Patriots team
 * Work on completing remaining Data tasks
 * Try to run a train


 * Results:

3/16: Today the Data team met via google Hangouts to discuss the remaining tasks. We divided up the tasks. Russ and Stephen are working on removing the "Feat" and "Wav" directories from the individual experiment directories. I am going to work on the re-structuring of the NOAA directory so it matches the switchboard directory we just finished re-structuring. Dakota is going to work on the remaining tasks.

3/20: Today I read through some other logs because Caesar is down.

3/21: Caesar is still down, so I am still unable to work on my tasks. This is setting back the data team from completing our tasks for this week.

3/24: Yesterday, when I tried to log into Caesar, it was still down. I received an email from the data team saying that Mohamed was going to look at it. I just checked again to try and work on my tasks, but Caesar is still down. It has been a very unsuccessful break as far as getting tasks completed. Hopefully, there is nothing wrong with the server and it will be up and running for class tomorrow.

I'm concerned that Caesar has been down for so long and that the systems group said they were going to bring it back up, but apparently were unable to.
 * Concerns:

Week Ending March 31, 2015

 * Task:
 * Team task for the Patriots
 * Data Group Tasks

3/26: Yesterday during class I was having issues logging into Caesar. I retried today and was able to successfully log in. While I was logged in, I updated my password because I had been unable to do so before since the switch over to new Caesar.
 * Results:

3/27: I tried to log into Caesar via cisunix.unh.edu. It kept telling me my password is incorrect. I tried to log in with root, but it seems that the password is different for root. I emailed Dakota to try and get into Caesar. He was able to copy and email me the file I was looking for for our team task.

3/30: After our internship class today, part of the data team was able to discuss our tasks and our progress on the tasks. We are hoping that we will be able to accomplish more in this coming week if there aren't as many issues logging into Caesar. After talking with the data group. I wandered down to the server room to discuss my logging in issues with the systems group. Melissa was able to reset my password. Then when I got home on Monday evening, I still got the same error about my password being incorrect. It turns out I am not the only one who had issues logging into Caesar after around 8pm.

3/31: I was finally able to get into Caesar! :) I looked through the corpus directories to try and determine was should be in the info file.


 * Plan:


 * Concerns:

Week Ending April 7, 2015

 * Task:
 * run first train
 * clean up Exp
 * info directories


 * Results:

4/4: I started running my first train this morning. I ran into a couple hiccups while running the train. While it was training, I got kicked off the internet, so I emailed Garrett to see how I could tell if the train was completed. He told me to do the following command to tell if I was still running the script. ps -u

After I ran this, it said that I was currently running a script which meant that my training had completed. I then did the language model. I has no issues with those steps.

4/5: Due to the issues I had yesterday, running the train took longer than anticipated, therefore I started the decoding today. I was able to make the directory and copy the script into the DECODE directory. I ran into an issue running the below command. nohup run_decode5.pl 008 0266/008 1000

It kept giving me the following error. run_decode5.pl: Command not found.

I emailed Garrett about this issue. He told me that I could go back to the main directory and run the below. nohup DECODE/run_decode5.pl 008 0266/008 1000 &

I was then able to move the decode.log file to the DECODE directory and from there I had no other issues. I ended up with about a 41% error rate.

4/6: Today I caught up on reading logs of my team mates to see what kind of progress everyone is making.

4/7: Today I started reading through my assigned scripts for my team.


 * Plan:


 * Concerns:

Week Ending April 14, 2015

 * Task:
 * Run another train
 * change some parameters in a train


 * Results:

4/11: Read up on classmates logs.

4/12: Today I ran another train. I change some parameters and it didn't work out so well. My results ended up much higher. I'm going to start a 125hr train tomorrow.

SYSTEM SUMMARY PERCENTAGES by SPEAKER

,-.     |                            hyp.trans                            | |-|      | SPKR    | # Snt # Wrd | Corr    Sub    Del    Ins    Err  S.Err | |-+-+-|     | sw2001b |   18    163 | 81.0   14.7    4.3   39.9   58.9  100.0 | |-+-+-|     | sw2001a |   14    101 | 77.2   20.8    2.0   57.4   80.2  100.0 | |-+-+-|     | sw2005a |   35    683 | 84.3   10.5    5.1   11.4   27.1   97.1 | |-+-+-|     | sw2005b |   62    569 | 66.3   23.7   10.0   28.6   62.4  100.0 | |=================================================================|     | Sum/Avg |  129   1516 | 76.7   16.6    6.7   24.0   47.3   99.2 | |=================================================================|           |  Mean   | 32.3  379.0 | 77.2   17.4    5.4   34.3   57.1   99.3 | | S.D.   | 21.8  290.1 |  7.9    5.9    3.4   19.3   22.1    1.4 | | Median | 26.5  366.0 | 79.1   17.8    4.7   34.3   60.6  100.0 | `-'

4/13: Today I attempted to run my first 125 hour train. It didn't work out so well. From what I understand everyone is having issues with running trains today, but I got the below error. I'm not really sure what it means, but it seems like maybe one of the scripts got messed up somehow?

[kje222@caesar 010]$ nohup scripts_pl/RunAll.pl. & [1] 29183 [kje222@caesar 010]$ MODULE: 00 verify training files O.S. is case sensitive ("A" != "a"). Phones will be treated as case sensitive. Phase 1: DICT - Checking to see if the dict and filler dict agrees with the phonelist file. Found 6 words using 4 phones Phase 2: DICT - Checking to make sure there are not duplicate entries in the dictionary Can not open listoffiles (/mnt/main/Exp/0266/010/etc/010_train.fileids) at /mnt/main/Exp/0266/010/scripts_pl/00.verify/verify_all.pl line 203. Something failed: (/mnt/main/Exp/0266/010/scripts_pl/00.verify/verify_all.pl)

4/14: After some emails from the Patriots team, I have deleted my 010 experiment and restarted it. So far I have not received any errors. The below step is still running. It's been 45 minutes since I started it. I emailed Garrett and Dakota to find out if this normal. I'm hoping that this is normal and it is just taking so long due to being for 125hr_3170.

/mnt/main/scripts/user/generateFeats.pl


 * Plan:


 * Concerns:

Week Ending April 21, 2015

 * Task:
 * Run another 5hr train with different parameters
 * Run a 125hr train

4/15: Today I ran a 5hr train. After running it, we found some different parameters to test out. So I changed said parameters (documented in email form) and got the below results. Still not as good as we want it to be.
 * Results:

,-.     |                            hyp.trans                            | |-|     | SPKR    | # Snt # Wrd | Corr    Sub    Del    Ins    Err  S.Err | |-+-+-|     | sw2001b |   18    163 | 77.9   17.8    4.3   40.5   62.6  100.0 | |-+-+-|     | sw2001a |   14    101 | 85.1   12.9    2.0   55.4   70.3  100.0 | |-+-+-|     | sw2005a |   39    701 | 82.5   11.6    6.0   13.7   31.2   94.9 | |-+-+-|     | sw2005b |   67    613 | 60.2   27.2   12.6   29.4   69.2  100.0 | |-+-+-|     | sw2006b |   29    618 | 71.4   19.4    9.2    7.8   36.4  100.0 | |-+-+-|     | sw2006a |   33    455 | 84.2   13.6    2.2   14.9   30.8  100.0 | |-+-+-|     | sw2007a |   61    614 | 75.2   16.4    8.3   18.1   42.8   90.2 | |-+-+-|     | sw2007b |   60    861 | 81.8   14.2    4.1    7.8   26.0   86.7 | |-+-+-|     | sw2008a |   24    260 | 79.2   18.5    2.3   24.6   45.4  100.0 | |-+-+-|     | sw2008b |   26    257 | 81.7   13.2    5.1   26.1   44.4   88.5 | |-+-+-|     | sw2009b |   23    181 | 77.9   14.9    7.2   32.0   54.1   95.7 | |-+-+-|     | sw2009a |   34    473 | 65.5   28.1    6.3   20.3   54.8  100.0 | |-+-+-|     | sw2010b |   22    404 | 68.8   20.0   11.1   10.4   41.6  100.0 | |-+-+-|     | sw2010a |   27    284 | 79.6   13.7    6.7   22.2   42.6   96.3 | |-+-+-|     | sw2012a |   45    838 | 82.0   12.4    5.6   13.2   31.3   95.6 | |-+-+-|     | sw2012b |   28    464 | 80.2   14.4    5.4   14.9   34.7  100.0 | |-+-+-|     | sw2013a |   35    377 | 65.8   28.6    5.6   20.4   54.6  100.0 | |-+-+-|     | sw2013b |   69    942 | 53.4   31.2   15.4   11.6   58.2   97.1 | |-+-+-|     | sw2014a |    9     79 | 78.5   16.5    5.1   34.2   55.7  100.0 | |-+-+-|     | sw2014b |   13    174 | 80.5   14.9    4.6   16.7   36.2   92.3 | |-+-+-|     | sw2015a |   21    375 | 73.9   16.0   10.1    3.5   29.6   85.7 | |-+-+-|     | sw2015b |   32    542 | 76.9   12.7   10.3   10.3   33.4  100.0 | |-+-+-|     | sw2017b |   38    658 | 82.2   12.2    5.6   14.3   32.1   97.4 | |-+-+-|     | sw2017a |   40    332 | 76.5   19.3    4.2   28.0   51.5   97.5 | |-+-+-|     | sw2018a |   49    505 | 59.0   35.0    5.9   42.6   83.6  100.0 | |-+-+-|     | sw2018b |   42    586 | 73.7   19.5    6.8   16.7   43.0  100.0 | |-+-+-|     | sw2019a |   51    600 | 75.8   17.0    7.2   15.5   39.7   96.1 | |-+-+-|     | sw2019b |   53    457 | 84.7   10.9    4.4   26.0   41.4  100.0 | |-+-+-|     | sw2020a |   43    373 | 73.7   22.0    4.3   46.1   72.4  100.0 | |-+-+-|     | sw2020b |   31    464 | 78.4   15.5    6.0   19.4   40.9  100.0 | |-+-+-|     | sw2022a |   29    508 | 68.9   22.0    9.1   12.8   43.9   96.6 | |-+-+-|     | sw2022b |   32    316 | 72.2   22.8    5.1   32.9   60.8   93.8 | |-+-+-|     | sw2023a |   52    795 | 69.8   21.9    8.3   12.8   43.0   94.2 | |-+-+-|     | sw2023b |   69    730 | 72.5   21.1    6.4   17.5   45.1   97.1 | |-+-+-|     | sw2024a |   25    227 | 84.1   12.3    3.5   22.5   38.3   92.0 | |-+-+-|     | sw2024b |   26    391 | 74.7   18.9    6.4   10.5   35.8  100.0 | |-+-+-|     | sw2025a |   25    403 | 55.1   36.2    8.7   14.6   59.6   96.0 | |-+-+-|     | sw2025b |   34    476 | 81.1   14.1    4.8   21.6   40.5   97.1 | |-+-+-|     | sw2027a |   37    862 | 76.0   18.6    5.5   11.7   35.7  100.0 | |-+-+-|     | sw2027b |   30    485 | 78.8   14.0    7.2   24.1   45.4   96.7 | |-+-+-|     | sw2028a |   27    224 | 80.4   13.8    5.8   21.9   41.5   85.2 | |-+-+-|     | sw2028b |   38    397 | 85.6   10.8    3.5   11.6   25.9   84.2 | |=================================================================|     | Sum/Avg | 1500  19565 | 74.3   18.7    6.9   18.0   43.6   96.3 | |=================================================================|     |  Mean   | 35.7  465.8 | 75.4   18.2    6.4   20.7   45.4   96.3 | | S.D.   | 14.9  216.2 |  7.9    6.3    2.7   11.1   13.3    4.7 | | Median | 32.5  460.5 | 77.4   16.5    5.9   17.8   42.7   97.2 | `-'

After running the 5hr train, I started running my first 125hr train. It was still training when I went to bed.

4/16: This morning, I checked on my 125hr train and the training had completed. I was able to create the language model and start decoding before I left for work. When I got home, the decoding had completed. Unfortunately, the parameters I changed with this train didn't help much with the error rate. My results are below.

,-.        |                            hyp.trans                            | |-|        | SPKR    | # Snt # Wrd | Corr    Sub    Del    Ins    Err  S.Err | |-+-+-|        | sw2001a |   32    541 | 53.6   41.0    5.4   25.0   71.3  100.0 | |-+-+-|        | sw2001b |   34    488 | 62.3   31.1    6.6   28.1   65.8  100.0 | |-+-+-|        | sw2005a |   53   1172 | 63.3   28.6    8.1   10.7   47.4   98.1 | |-+-+-|        | sw2005b |   77    817 | 43.3   42.4   14.3   24.2   80.9  100.0 | |-+-+-|        | sw2006a |   40    608 | 62.7   30.4    6.9   15.5   52.8   97.5 | |-+-+-|        | sw2006b |   43   1012 | 45.0   40.2   14.8    5.9   61.0  100.0 | |-+-+-|        | sw2007a |   86   1064 | 62.4   27.6   10.0   12.8   50.4   93.0 | |-+-+-|     | sw2007b |   80   1183 | 53.9   37.6    8.5    9.9   56.0  100.0 | |-+-+-|        | sw2008a |   28    369 | 64.5   30.9    4.6   20.3   55.8  100.0 | |-+-+-|        | sw2008b |   32    436 | 61.7   31.7    6.7   22.5   60.8   93.8 | |-+-+-|        | sw2009a |   37    605 | 41.0   46.6   12.4   10.2   69.3  100.0 | |-+-+-|        | sw2009b |   44    649 | 54.5   36.2    9.2   18.2   63.6  100.0 | |-+-+-|        | sw2010a |   38    528 | 62.9   29.9    7.2   18.8   55.9   97.4 | |-+-+-|     | sw2010b |   33    659 | 48.7   37.2   14.1   12.1   63.4   97.0 | |-+-+-|     | sw2012a |   67   1420 | 52.5   35.2   12.3   11.9   59.4  100.0 | |-+-+-|     | sw2012b |   43    846 | 65.0   27.5    7.4   12.6   47.6  100.0 | |-+-+-|     | sw2013a |   52    766 | 40.2   52.3    7.4   15.7   75.5  100.0 | |-+-+-|     | sw2013b |   88   1526 | 41.2   39.6   19.3    9.0   67.9   97.7 | |-+-+-|     | sw2014a |   23    311 | 53.4   41.2    5.5   40.2   86.8  100.0 | |-+-+-|     | sw2014b |   27    543 | 53.0   36.8   10.1   12.0   58.9   96.3 | |-+-+-|     | sw2015a |   29    611 | 52.0   33.4   14.6    3.4   51.4   93.1 | |-+-+-|     | sw2015b |   41    790 | 61.0   26.6   12.4    9.5   48.5   97.6 | |-+-+-|     | sw2017a |   46    504 | 53.6   38.1    8.3   21.2   67.7  100.0 | |-+-+-|     | sw2017b |   46    974 | 62.6   28.5    8.8    9.9   47.2   97.8 | |-+-+-|     | sw2018a |   55    634 | 37.7   56.6    5.7   36.0   98.3  100.0 | |-+-+-|     | sw2018b |   57    972 | 58.8   33.6    7.5   15.5   56.7  100.0 | |-+-+-|     | sw2019a |   60    759 | 49.7   38.9   11.5   11.3   61.7   96.7 | |-+-+-|     | sw2019b |   65    825 | 67.4   25.2    7.4   13.9   46.5  100.0 | |-+-+-|     | sw2020a |   61    665 | 60.6   33.1    6.3   37.1   76.5  100.0 | |-+-+-|     | sw2020b |   66   1586 | 58.9   35.1    6.1   14.1   55.2  100.0 | |-+-+-|     | sw2022a |   40    780 | 53.5   35.8   10.8   12.6   59.1  100.0 | |-+-+-|     | sw2022b |   43    611 | 42.1   46.0   11.9   19.6   77.6  100.0 | |-+-+-|     | sw2023a |   72   1204 | 42.4   46.3   11.3    9.1   66.7   98.6 | |-+-+-|     | sw2023b |   81    946 | 49.9   40.5    9.6   15.0   65.1   97.5 | |-+-+-|     | sw2024a |   38    442 | 67.0   29.0    4.1   17.0   50.0  100.0 | |-+-+-|     | sw2024b |   45    891 | 43.2   42.0   14.8    6.5   63.3  100.0 | |-+-+-|     | sw2025a |   38    718 | 38.2   46.0   15.9   11.0   72.8  100.0 | |-+-+-|     | sw2025b |   44    829 | 60.6   31.4    8.1   14.1   53.6   97.7 | |-+-+-|     | sw2027a |   51   1181 | 56.2   35.1    8.7   13.3   57.1  100.0 | |-+-+-|     | sw2027b |   55   1185 | 43.0   43.5   13.4   14.9   71.9  100.0 | |-+-+-|     | sw2028a |   88    960 | 61.0   29.9    9.1   16.6   55.5   93.2 | |-+-+-|     | sw2028b |   91    948 | 67.4   26.5    6.1   11.7   44.3   92.3 | |-+-+-|     | sw2032a |   76   1311 | 35.0   51.1   13.9    9.6   74.6  100.0 | |-+-+-|     | sw2032b |   60    741 | 60.1   34.4    5.5   25.1   65.0  100.0 | |-+-+-|     | sw2035a |   41    425 | 48.2   43.3    8.5   17.9   69.6  100.0 | |-+-+-|     | sw2035b |   57    878 | 46.9   41.5   11.6   14.1   67.2  100.0 | |-+-+-|     | sw2036a |   64    775 | 46.5   38.1   15.5    8.8   62.3  100.0 | |-+-+-|     | sw2036b |   48    516 | 53.3   38.0    8.7   16.3   63.0   97.9 | |-+-+-|     | sw2038a |   49    869 | 59.1   34.6    6.2   19.1   60.0  100.0 | |-+-+-|     | sw2038b |   60   1156 | 48.3   41.7   10.0    7.5   59.3  100.0 | |-+-+-|     | sw2039a |   81   1031 | 59.1   33.9    7.0   12.5   53.4   93.8 | |-+-+-|     | sw2039b |   65    600 | 58.3   37.0    4.7   29.3   71.0   92.3 | |-+-+-|     | sw2040a |   61    757 | 72.1   25.1    2.8   12.9   40.8   91.8 | |-+-+-|     | sw2040b |   63   1103 | 68.4   24.1    7.4    6.0   37.5   92.1 | |-+-+-|     | sw2041a |   82   1286 | 47.0   42.5   10.5   19.7   72.6   98.8 | |-+-+-|     | sw2041b |   56   1071 | 59.3   31.7    9.1   13.0   53.7  100.0 | |-+-+-|     | sw2044a |   72   1371 | 59.6   32.1    8.3   14.7   55.1  100.0 | |-+-+-|     | sw2044b |   51    909 | 70.4   24.6    5.0   16.5   46.1  100.0 | |-+-+-|     | sw2045a |   62    970 | 53.1   38.1    8.8    5.8   52.7   98.4 | |-+-+-|     | sw2045b |   61    506 | 51.8   37.2   11.1   16.6   64.8   95.1 | |-+-+-|     | sw2050a |   68    915 | 70.9   22.6    6.4   19.6   48.6   97.1 | |-+-+-|     | sw2050b |   66   1084 | 56.3   36.3    7.5   13.7   57.4  100.0 | |-+-+-|     | sw2051a |   64    477 | 50.5   43.4    6.1   19.7   69.2   89.1 | |-+-+-|     | sw2051b |  139   1483 | 38.4   46.3   15.2   11.5   73.1   97.1 | |-+-+-|     | sw2053a |   58    575 | 50.3   40.5    9.2   17.6   67.3   98.3 | |-+-+-|     | sw2053b |   61    720 | 51.1   39.6    9.3   15.4   64.3   95.1 | |-+-+-|     | sw2054a |  110   1371 | 58.5   34.2    7.3   12.1   53.6   96.4 | |-+-+-|     | sw2054b |  100    968 | 47.1   41.8   11.1   13.7   66.6  100.0 | |-+-+-|     | sw2055a |   42    458 | 55.9   36.5    7.6   20.7   64.8  100.0 | |-+-+-|     | sw2055b |   32    560 | 49.5   36.6   13.9   10.4   60.9  100.0 | |-+-+-|     | sw2056a |   45    813 | 56.2   36.7    7.1    8.6   52.4   97.8 | |-+-+-|     | sw2056b |   54    492 | 68.3   23.6    8.1   21.1   52.8   96.3 | |-+-+-|     | sw2057a |   71   1614 | 65.5   28.4    6.1   12.1   46.7  100.0 | |-+-+-|     | sw2057b |   36    477 | 60.6   32.1    7.3   19.7   59.1  100.0 | |-+-+-|     | sw2060a |   47    608 | 59.7   35.2    5.1   14.3   54.6   97.9 | |-+-+-|     | sw2060b |   48    463 | 60.3   31.7    8.0   18.1   57.9   93.8 | |-+-+-|     | sw2061a |  114   1742 | 50.4   37.8   11.8   11.5   61.1   99.1 | |-+-+-|     | sw2061b |  115   1072 | 35.5   51.9   12.6   20.4   84.9  100.0 | |-+-+-|     | sw2062a |   40    755 | 52.2   39.3    8.5   10.1   57.9   97.5 | |-+-+-|     | sw2062b |   71   1478 | 57.8   33.4    8.7   10.1   52.3  100.0 | |-+-+-|     | sw2064a |   74    889 | 51.3   35.8   12.9   14.1   62.8  100.0 | |-+-+-|     | sw2064b |   67    860 | 57.9   35.0    7.1   12.8   54.9  100.0 | |-+-+-|     | sw2065a |   77   1282 | 51.8   36.9   11.3   13.9   62.1   98.7 | |-+-+-|     | sw2065b |   78   1276 | 66.5   25.1    8.4   18.8   52.3   96.2 | |-+-+-|     | sw2067a |   23    408 | 51.5   43.6    4.9   20.6   69.1  100.0 | |-+-+-|     | sw2067b |   38    582 | 44.3   49.3    6.4   19.1   74.7  100.0 | |-+-+-|     | sw2071a |   40    550 | 61.6   30.5    7.8   12.0   50.4   97.5 | |-+-+-|     | sw2071b |   46    561 | 72.2   21.2    6.6   13.9   41.7   95.7 | |-+-+-|     | sw2072a |   50    617 | 49.4   44.2    6.3   25.9   76.5  100.0 | |-+-+-|     | sw2072b |   36    647 | 36.5   48.8   14.7   12.4   75.9  100.0 | |-+-+-|     | sw2073a |   55    918 | 55.7   33.4   10.9   10.3   54.7   98.2 | |-+-+-|     | sw2073b |   64    797 | 57.0   33.5    9.5   11.0   54.1   98.4 | |-+-+-|     | sw2078a |  118   1252 | 41.7   48.6    9.7   18.7   77.0  100.0 | |-+-+-|     | sw2078b |  129   1362 | 43.5   40.7   15.9   16.4   73.0   97.7 | |-+-+-|     | sw2079a |   27    435 | 36.3   58.2    5.5   17.7   81.4  100.0 | |-+-+-|     | sw2079b |   32    612 | 37.4   42.6   19.9    4.2   66.8  100.0 | |-+-+-|     | sw2080a |   68    980 | 55.2   37.4    7.3   12.2   57.0   98.5 | |-+-+-|     | sw2080b |   59    690 | 35.1   49.3   15.7    9.1   74.1  100.0 | |-+-+-|     | sw2082a |   38    558 | 49.1   38.2   12.7   17.4   68.3  100.0 | |-+-+-|     | sw2082b |   36    637 | 67.2   24.3    8.5    9.7   42.5  100.0 | |-+-+-|     | sw2083a |   73    750 | 33.2   48.1   18.7   19.2   86.0  100.0 | |-+-+-|     | sw2083b |   66    952 | 50.7   37.4   11.9   10.0   59.2   97.0 | |-+-+-|     | sw2085a |   71    741 | 49.0   41.8    9.2   17.5   68.6  100.0 | |-+-+-|     | sw2085b |  112   1373 | 41.4   44.6   13.9    9.2   67.7  100.0 | |-+-+-|     | sw2086a |   75   1171 | 56.8   30.4   12.8   10.2   53.5   98.7 | |-+-+-|     | sw2086b |   69    701 | 46.6   45.1    8.3   39.1   92.4  100.0 | |-+-+-|     | sw2087a |   39    823 | 65.0   27.0    8.0    6.6   41.6   97.4 | |-+-+-|     | sw2087b |   38    482 | 52.9   36.5   10.6    9.1   56.2  100.0 | |-+-+-|     | sw2089a |   98   1359 | 47.9   37.5   14.6   18.6   70.7   93.9 | |-+-+-|     | sw2089b |   45    600 | 57.3   31.0   11.7   21.7   64.3  100.0 | |=================================================================|     | Sum/Avg | 6500  93823 | 53.2   36.8   10.0   14.4   61.2   98.2 | |=================================================================|     |  Mean   | 59.1  852.9 | 53.5   36.9    9.6   15.3   61.8   98.3 | | S.D.   | 23.4  324.0 |  9.4    7.6    3.5    6.7   11.4    2.4 | | Median | 56.5  793.5 | 53.4   36.7    8.7   14.0   60.9  100.0 | `-'

4/18: Today, I read through some other team mates logs.

4/21: Today, I started my first 256hr train.


 * Plan:


 * Concerns:

Week Ending April 28, 2015

 * Task:
 * Change parameters and rerun some of my experiments
 * Complete 256 hour train and decode

4/22: Today, I canceled my 256 hour train, it hadn't gotten very far and I forgot to change some the parameters. I restarted it with different configurations as discussed with Patriots on team email.
 * Results:

4/23: Today, I started another 125 hour train with the new parameters since my 256 hour train is still running. Hopefully I will have some results tomorrow.

4/24: I just finished scoring my 5 hour train. Results are below.

,-.     |                            hyp.trans                            | |-|     | SPKR    | # Snt # Wrd | Corr    Sub    Del    Ins    Err  S.Err | |-+-+-|     | sw2001b |   18    163 | 87.7   11.0    1.2   38.0   50.3  100.0 | |-+-+-|         | sw2001a |   14    101 | 93.1    5.9    1.0   49.5   56.4  100.0 | |-+-+-|     | sw2005a |   39    701 | 91.2    5.8    3.0   10.8   19.7   94.9 | |-+-+-|     | sw2005b |   67    613 | 76.7   17.0    6.4   27.4   50.7  100.0 | |-+-+-|     | sw2006b |   29    618 | 85.1    9.2    5.7    6.5   21.4   93.1 | |-+-+-|     | sw2006a |   33    455 | 93.8    5.1    1.1   13.6   19.8   87.9 | |-+-+-|     | sw2007a |   61    614 | 84.2    8.8    7.0   15.6   31.4   80.3 | |-+-+-|     | sw2007b |   60    861 | 92.3    5.7    2.0    9.2   16.8   78.3 | |-+-+-|     | sw2008a |   24    260 | 86.9   10.8    2.3   24.2   37.3  100.0 | |-+-+-|     | sw2008b |   26    257 | 92.2    6.2    1.6   25.7   33.5   88.5 | |-+-+-|     | sw2009b |   23    181 | 87.3    9.4    3.3   29.8   42.5   91.3 | |-+-+-|     | sw2009a |   34    473 | 85.0    9.3    5.7   14.4   29.4   94.1 | |-+-+-|     | sw2010b |   22    404 | 82.9   12.1    5.0    8.7   25.7   95.5 | |-+-+-|     | sw2010a |   27    284 | 88.4    8.5    3.2   22.9   34.5   96.3 | |-+-+-|     | sw2012a |   45    838 | 89.4    7.2    3.5   11.9   22.6   91.1 | |-+-+-|     | sw2012b |   28    464 | 88.4   10.1    1.5   14.4   26.1  100.0 | |-+-+-|     | sw2013a |   35    377 | 89.1    8.8    2.1   14.6   25.5   82.9 | |-+-+-|     | sw2013b |   69    942 | 70.6   19.1   10.3   10.8   40.2   91.3 | |-+-+-|     | sw2014a |    9     79 | 89.9    2.5    7.6   38.0   48.1  100.0 | |-+-+-|     | sw2014b |   13    174 | 89.7    6.9    3.4   16.1   26.4   92.3 | |-+-+-|     | sw2015a |   21    375 | 84.5   10.4    5.1    4.0   19.5   81.0 | |-+-+-|     | sw2015b |   32    542 | 81.4   11.4    7.2   13.3   31.9   93.8 | |-+-+-|     | sw2017b |   38    658 | 91.3    7.1    1.5   12.6   21.3   89.5 | |-+-+-|     | sw2017a |   40    332 | 85.2   10.8    3.9   26.5   41.3   87.5 | |-+-+-|     | sw2018a |   49    505 | 79.6   16.2    4.2   41.2   61.6   98.0 | |-+-+-|     | sw2018b |   42    586 | 90.1    6.5    3.4   17.9   27.8   97.6 | |-+-+-|     | sw2019a |   51    600 | 91.2    5.0    3.8   13.0   21.8   90.2 | |-+-+-|     | sw2019b |   53    457 | 91.7    5.9    2.4   21.7   30.0   96.2 | |-+-+-|     | sw2020a |   43    373 | 86.9   10.5    2.7   45.6   58.7  100.0 | |-+-+-|     | sw2020b |   31    464 | 90.7    6.5    2.8   17.5   26.7   93.5 | |-+-+-|     | sw2022a |   29    508 | 87.6    8.5    3.9   11.0   23.4   93.1 | |-+-+-|     | sw2022b |   32    316 | 83.5   10.8    5.7   29.4   45.9   93.8 | |-+-+-|     | sw2023a |   52    795 | 87.7    7.2    5.2   10.3   22.6   86.5 | |-+-+-|     | sw2023b |   69    730 | 90.0    7.3    2.7   14.5   24.5   87.0 | |-+-+-|     | sw2024a |   25    227 | 88.1    7.5    4.4   20.3   32.2   80.0 | |-+-+-|     | sw2024b |   26    391 | 84.1    7.2    8.7    8.7   24.6   88.5 | |-+-+-|     | sw2025a |   25    403 | 67.0   25.8    7.2   13.9   46.9   92.0 | |-+-+-|     | sw2025b |   34    476 | 84.9   12.6    2.5   20.8   35.9   91.2 | |-+-+-|     | sw2027a |   37    862 | 90.0    7.1    2.9   11.0   21.0   94.6 | |-+-+-|     | sw2027b |   30    485 | 91.3    6.4    2.3   20.8   29.5   93.3 | |-+-+-|     | sw2028a |   61    547 | 92.1    5.1    2.7   21.4   29.3   78.7 | |-+-+-|     | sw2028b |   76    758 | 92.5    5.8    1.7   10.6   18.1   73.7 | |-+-+-|     | sw2032a |   55    803 | 88.0    8.8    3.1   13.1   25.0   89.1 | |-+-+-|     | sw2032b |   33    309 | 87.7    9.4    2.9   33.3   45.6   93.9 | |-+-+-|     | sw2035a |   32    263 | 88.6    9.5    1.9   24.7   36.1   90.6 | |-+-+-|     | sw2035b |   39    456 | 88.6    8.6    2.9   20.2   31.6   92.3 | |-+-+-|     | sw2036b |   41    401 | 85.8   10.7    3.5   10.7   24.9   85.4 | |-+-+-|     | sw2036a |   53    552 | 82.1   10.0    8.0   12.5   30.4   88.7 | |-+-+-|     | sw2038b |   41    838 | 82.1   13.6    4.3    5.1   23.0   82.9 | |-+-+-|     | sw2038a |   35    507 | 87.4    6.9    5.7   26.2   38.9  100.0 | |-+-+-|     | sw2039b |   49    342 | 84.8   13.7    1.5   30.7   45.9   91.8 | |-+-+-|     | sw2039a |   66    696 | 87.2    9.3    3.4   11.4   24.1   69.7 | |-+-+-|     | sw2040a |   49    547 | 95.1    4.0    0.9    9.7   14.6   71.4 | |-+-+-|     | sw2040b |   51    750 | 86.7    7.2    6.1    6.8   20.1   66.7 | |-+-+-|     | sw2041a |   68    933 | 82.9   10.6    6.5   19.7   36.9  100.0 | |-+-+-|     | sw2041b |   33    492 | 85.6    8.5    5.9   13.8   28.3   90.9 | |-+-+-|     | sw2044a |   40    487 | 83.2   10.7    6.2   22.8   39.6   95.0 | |-+-+-|     | sw2044b |   39    644 | 93.6    5.1    1.2   14.3   20.7   94.9 | |-+-+-|     | sw2045a |   44    596 | 91.8    5.7    2.5    6.0   14.3   79.5 | |-+-+-|     | sw2045b |   51    352 | 86.4   10.2    3.4   24.4   38.1   86.3 | |-+-+-|     | sw2050b |   50    774 | 88.1    8.3    3.6   10.5   22.4   88.0 | |-+-+-|     | sw2050a |   57    630 | 94.9    2.7    2.4   19.7   24.8   96.5 | |-+-+-|     | sw2051a |   55    361 | 85.6    7.5    6.9   22.7   37.1   83.6 | |-+-+-|     | sw2051b |  124   1283 | 85.2   10.4    4.4   11.7   26.5   75.0 | |-+-+-|     | sw2053a |   50    476 | 87.0    7.6    5.5   18.3   31.3   86.0 | |-+-+-|     | sw2053b |   45    479 | 86.0    9.4    4.6   17.3   31.3   88.9 | |-+-+-|     | sw2054a |   91   1069 | 92.6    5.4    2.0   11.7   19.1   79.1 | |-+-+-|     | sw2054b |   86    729 | 82.3   14.1    3.6   19.3   37.0   86.0 | |-+-+-|     | sw2055a |   35    322 | 82.0   13.4    4.7   23.0   41.0   91.4 | |-+-+-|     | sw2055b |   23    360 | 84.7    9.4    5.8   14.7   30.0   91.3 | |-+-+-|     | sw2056a |   33    481 | 86.1    6.7    7.3   10.0   23.9   75.8 | |-+-+-|     | sw2056b |   48    374 | 89.8    7.8    2.4   24.9   35.0   89.6 | |-+-+-|     | sw2057a |   42    850 | 89.6    7.4    2.9   10.5   20.8   88.1 | |-+-+-|     | sw2057b |   28    269 | 89.2   10.0    0.7   26.4   37.2   96.4 | |-+-+-|     | sw2060a |   33    326 | 87.7    7.1    5.2   17.2   29.4   72.7 | |-+-+-|     | sw2060b |   43    398 | 87.2    5.8    7.0   16.1   28.9   74.4 | |-+-+-|     | sw2061b |   99    780 | 76.7   17.1    6.3   27.8   51.2   93.9 | |-+-+-|     | sw2061a |   89   1302 | 88.2    7.6    4.1   10.3   22.0   87.6 | |-+-+-|     | sw2062b |   55    960 | 86.9    4.8    8.3    9.4   22.5   92.7 | |-+-+-|     | sw2062a |   29    500 | 90.8    6.4    2.8    7.6   16.8   79.3 | |=================================================================|     | Sum/Avg | 3506  42940 | 86.9    8.9    4.2   15.7   28.8   87.8 | |=================================================================|     |  Mean   | 43.8  536.8 | 87.0    8.9    4.1   18.0   31.0   89.0 | | S.D.   | 20.2  247.5 |  4.8    3.7    2.1    9.3   10.6    8.0 | | Median | 40.0  486.0 | 87.6    8.5    3.5   15.2   29.3   91.0 | `-'

4/28: My 256 hour train is having some issues. A lot of the error rates for individual speakers is higher than 100% which doesn't seem possible. I wasn't having this issue until I started using the fileids that Prof. Jonas provided us. I emailed Garrett about the issue and he is having similar issues.


 * Plan:


 * Concerns:

Week Ending May 5, 2015

 * Task:


 * Results:


 * Plan:


 * Concerns: