Speech:Spring 2012 Brice Rader Log

From Openitware
Jump to: navigation, search


Week ending February 6, 2012

Task
  • Install Suse via virtual machine and install Sphinx 3 onto computer as well.
  • Familiarize myself with methusalix which is Aaron's and my server connected to Caesar.
    • cat /etc/hosts
Results

Result - 1/31:

  • Created spring 12 wiki for the class
    • Formatted it a little bit
    • Created my page, added tasks
  • Messed around with the server

Result - 2/2:

  • Aaron J and I created this install tutorial. VB/SUSE/Sphinx Install Tutorial
    • Aaron initially installed on his computer. Then we met on Skype and created the tutorial.
    • Installed Virtual Box and openSUSE, having problems with Sphinx though.

Result - 2/3:

  • Read logs, seems as though Brandon figured out how to install Sphinx so hopefully he can give Aaron and I a hand.
  • Messed around with our server.

COMMENT: Brandon M.

I had a typo on my end, I meant to type installing Sphinx. I am working on figuring it out though.

Result - 2/6:

  • Read logs
  • Played around in open suse - kind of frustrating how slow it is.
Plan
  • Really need to find a Sphinx 3.rpm so we can install it on our openSUSE virtual box so we can familiarize ourselves with it.
Concerns


Week ending February 13, 2012

Task
  • Format wiki page to how Mike would like us to have it formatted.
  • Find sphinx3.rpm to install onto virtual box
  • Install Sphinx 3 onto openSUSE
  • Search for speech tools
  • Mess around with Sphinx 3 if able to install
Results

Result 2/7

  • Formatted wiki page
  • Searched for a Sphinx 3 .rpm, still no luck
    • Messaged the person who runs CMU Sphinx on LinkedIn, Aaron is putting a message in their IRC channel so hopefully between the two of us we'll get a reply.
  • Read everyone's logs to familiarize with what everyone else is at

Result 2/8

  • Contacted a LinkedIn group about the .rpm file
  • Read logs

Result 2/12

Result 2/13

  • Attempted to install pocketsphinx onto my virtual box
    • The second link above has the download file
    • The first link was useless for me
  • Read logs
    • I saw that a lot of people said they installed the VPN. I don't know if they just mean they are using it or actually installed something but all you need to do is what this wiki page says and you're on it. http://foss.unh.edu/opencomputing/index.php/VPN
Plan

Get sphinx 3 installed and running.

Concerns

Not finding sphinx 3 or being able to get it installed and working properly

Week ending February 20, 2012

Task
  • Install sphinx 3
    • Revise installation guide
  • Meet with Jonas and the rest of the group on Friday at 3 pm
  • Figure out what Aaron and I are doing and what Chad and Ted are doing (decoding and trainer)
  • Figure out how to use sphinx 3
  • Fix our server so we can access it
Results

Result 2/14

  • Aaron figured out how to install sphinx 3 so we both did that
  • While I was installing it Aaron wrote what I was doing then I took that word document and spiced it up and added it to our install guide so that is now completely finished. I'm going to add that to the project information to be public.
  • Aaron and I are focusing on the decoding part of sphinx while Ted and Chad are responsible for doing the training part
  • Logged into caesar then went to super user and then ssh over to our server then mounted it so we can now access the server under our own username.

Result 2/17

  • Joined Capstone Google Group
  • Met with Chad, Aaron and Prof. Jonas on Skype for about an hour.
    • Unix commands we will be using
      • cd /root/speechtools/SphinxTrain-1.0/train1/etc
      • sed "s/ /\n/g" train1_train.trans | sort | uniq | grep -v ms98 | more
      • sed "s/ /\n/g" train1_train.trans | sort | uniq | grep -v ms98 | grep -v "<" | more
      • sed "s/ /\n/g" train1_train.trans | sort | uniq | grep -v ms98 | grep -v "<" > my.dict
      • sed "s/ /\n/g" train1_train.trans | sort | uniq | grep -v ms98 | grep -v "<" > my.dict
      • /run_decode.pl train1
      • tail -f decode.log
      • cd /media/data/Switchboard/transcripts/ICSI Transcriptions/trans/icsi
      • more ms98_icsi_word.text

Result 2/19

  • Mike was nice enough and composed a list of things that we need to accomplish and I created a page for it on the wiki and put the main link on our group page
  • Emailed group to try to figure out how we're going to split up work. As of right now Aaron and I are working on the decoding while Ted and Chad are working on the Training
    • We also need to take that and contribute to the proposal
  • Played around in Caesar some more under root access
  • Read logs

Result 2/20

  • Played around in Unix
  • Added a small amount to the proposal
  • Read logs
Plan
  • Talk to the group tomorrow as communication again this week wasn't good, better than previous weeks but not good
    • Divide up "to-do's"
  • On Tuesday figure out Sphinx!
Concerns
  • Having a difficult time with Sphinx

Week ending February 27, 2012

Plan
  • Have proposal finished before next class
    • 4-5 Paragraphs on how the perl scripts for training and decoding work
    • How the trainer and decoder actually work
Results

Results 2/21

  • Aaron and I started our proposal
  • Played around in Unix in lab today and after lab.
    • Searched around for some of the scripts we might need to use for for the trainer
    • Already found the decoding script and we worked on in class how that works a little bit but we still need to figure it out a little more probably

Results 2/23

  • Aaron and I worked on the proposal, he did most of the work. I did editing and added a few things. I owe him though. It's almost ready to be added to the Wiki
  • Read logs

Results 2/26

  • Aaron and I continued to work on the proposal today. We have what dates we propose to have our tasks done by.

Results 2/27

  • Read logs
  • Aaron uploaded the proposal
Concerns

Week Ending March 5, 2012

Plan
  • Start working on tasks described in proposal
Results

Result 2/28

  • Aaron and I went through the trainer guide and installed new modules in the directory 'train2'
  • Reworked proposal to not include 'we' and to have results completed by at the bottom instead of embedded in the paragraphs

Result 2/29

  • Read logs

Result 3/4

  • Ran decoding with Aaron via Skype Screen Share
    • This created a decode log. This shows what the script run_decode does along with show it's parameters. We don't quite get how the script works yet but we're working on it
  • We also discussed how we are going to organize and show what scripts are being used in the trainer and what they do.
    • genTrans.pl
    • genPhones.sh
    • make_feats.pl
    • Runall.pl

Result 3/5

  • Read logs

Week Ending March 19, 2012

Plan
  • Work on tasks that have been described in the proposal
    • Find all the scripts
    • list what other scripts they use
    • Get their directory
    • Attempt to figure out what they do
Results

Result 3/6

  • Worked on our task, on Aaron's page there is a table which shows the start of our work
    • Found all scripts

Result 3/15

  • Read logs

Result 3/18

  • Attempted to play with the scripts to see what they do, what the output is and how they are used in the sequence of training

Result 3/19

  • Read logs

Week Ending March 26, 2012

Plan
Results

Result 3/20

  • Aaron and I created a table that lists all the scripts, where they are located and what parameters or scripts the script is using. We put the table in our part of the "information" section of Speech

Results 3/22

  • Read logs

Results 3/25

  • I messed around with the scripts, inspected each one to figure out how they work to trace them.
    • Tried running some of them individually but that really didn't do anything because they are only one piece of the pie
  • Looked around the web for some information on how the trainer works. Found some information but not as much as I like
    • Posted a link in our Steps for Building and Acoustic model

Results 3/26

  • Read logs
  • Looked around the web some more
Concerns
  • I don't know if we're going to get everything done that we would have liked to accomplish. We have a much better understanding of how it works and I feel like have a pretty good trace on how it works but replicating it on our own with our own information and matching up text with .wav files just might not be obtainable for us this semester.

Week Ending April 2, 2012

Plan
  • Aaron and I are going to attempt to run a sample train on new data
    • We are going to use the last 100 utterances of the sample data
Results

Result 3/27

  • Aaron and I spent a few hours trying to figure out how to grab the last 100 utterances and bringing them over to our new train4 folder where will we be running the new sample train
    • It was easy to get the output for the last 100 utterances using the tail -100 command but grabbing that output and putting it where we want it was more difficult.
    • We created a text file in the train4/wavTemp/ directory and then copied the output into that text document
    • We will attempt to run the new train on that

Result 4/1

  • Aaron worked on the train 4 today but he forgot where a few files were that we need to complete it so he forwarded me what he had done and I continued to work on it
  • Moved the .sph files to the wavTemp folder
    • Made sure to grab new .sph files and not ones that weren't used before

Result 4/2

  • Finished up messing around with the directories to get them how we needed
  • Tried finishing up the trainer but for whatever reason it's not recognizing the genTrans.pl script even though I'm in the correct directory and so is the location of the script...
    • Need help on this...

Week Ending April 9, 2012

Plan
  • Aaron and I need to finish up tasks from previous weeks
    • Finish train4
    • Add to our table in the Class Notes::Model Building::Building & Verifying models section
      • Find out what executable files are being called in the Runall.pl script
Results

Results 4/2

  • Started the organization for our poster presentation for UNH URC
    • Posted the first parts of it on the wiki
  • Worked on train4
    • Not much success, still got errors that prevented the train from working

Results 4/5

  • Prepared for meeting with my teammates tomorrow at 2:30 via Skype
    • Created 'train10' and moved over some files in preparation

Results 4/6

  • Lead meeting with team via Skype; presented screen via TeamViewer
    • Mini train 100% successful after a lot of tweaks
      • Prof. Jonas edited some scripts and added a couple of words to the dicionary
      • He also helped me add to the 'train#.phone' dictionary
  • Created a tutorial on how to use TeamViewer for both the presenter and participant

Results 4/8

  • Added to Aaron and my table: Buliding & Verifying Models
    • Put a location for each script because there were duplicate scripts in different locations.
    • Added a little information about the table
  • Aaron and I split up each of the scripts in Runall.pl to search for executable files
    • I was only able to find one of them which was in /04.vtln_align/slave_align.pl
      • This had sphinx3_align executable
  • Added information to my team for this week, Team A
    • Information includes what we discussed in our meeting, what Prof. Jonas and I changed, and some notes

Results 4/9

  • Read logs
    • Answered Matt's e-mail
  • Worked on Decode, got an error, will go back to it later today if I have the time.

Week Ending April 16, 2012

Plan
  • Meet on Skype to discuss what we need to do
    • Move scripts over to new server and have them run on /mnt/main/
  • Complete poster
    • Contact everyone that needs to give me information
Results

Result 4/13

  • Skype meeting
    • Discussed what we were going to do
    • Played around in our new environment

Result 4/14

  • Created a thread on the Google group notifying everyone that our poster is due this Monday.
  • Worked on poster

Result 4/15

  • Added more information to Aaron and my page
    • Added a couple of paragraphs of what we're actually doing to build and verify an acoustic model
    • Added information for the genTrans.pl script
  • Coordinated with Johnny and Bethany for the poster

Result 4/16

  • Worked on poster
    • Created system flow diagram
    • Created section for training and decoding
    • Edited abstract

Week Ending April 23, 2012

Plan
  • Train and Decode on our server
Results

Result 4/17

  • Worked on training and decoding on our server
    • We moved the scripts, transcripts, and linked the wav files to our server
    • Got partially through the train
  • Skype meeting scheduled for Friday

Result 4/20

  • Worked more on the training
    • Edited sphinx_train.cfg so the link to the directories was set up properly
    • Walked through the steps with Bethany and John
    • We have an error at the moment
      • We had Bethany post an question on the Google Group to see if anyone else had the same issue
  • Skype meeting

Result 4/22

  • Added a reply to Bethany's email to try to get a reply out of someone but I haven't gotten one yet
    • Looked at other group's pages doesn't seem as anyone else has got to the point we're at yet, at least according to the logs
  • Attempted to mess around with the error by going into scripts and checking paths and such but no luck...
  • Met with Johnny and Brandon to help them with their experiments
    • Both are caught up to where my team is

Result 4/23

  • Attempted to mess around more with the make_feats error, really don't get why it's getting an error....
    • Made documentation on our group page

Week Ending April 30, 2012

Results

Results 4/24

  • got make feats to work
    • Documentation is in the model group log

Results 4/27

  • Read logs of modeling team
  • Aaron G sent out a couple of e-mails so I read the report and other things that he talked about
  • Attempted messing around with the executable files but so far nothing.
  • Added information to the modeling group log

Results 4/29

  • Read logs of modeling team and the group page
    • Added a comment to group page
  • Messed with paths, still nothing... This is going to be a problem..

Week Ending May 7, 2012

Results

May 1

May 7

  • Started decode
    • Logs in the modeling group page

Extra work after semester

  • Haven't been able to finish the decode. Have gotten several errors which have been documented in the modeling group page
  • Added a little bit of information to the report