Speech:Spring 2013 Kevin Annis Log


 * Home
 * Semesters
 * Spring 2013
 * Proposal
 * Report

Week Ending February 5th, 2013

 * Task:


 * Results:


 * Plan:


 * Concerns:

Week Ending February 12, 2013

 * Task:


 * Results:


 * Plan:


 * Concerns:

Week Ending February 19, 2013

 * Task: February 13th, Burned Audio transcripts from Switchboard-1 Telephone Speech Corpus. 23 disks worth of files have been uploaded. After examination of folder contents, we found that the audio files were .SPH file extensions. This is a sphinx file header, within it contained the audio and file information.


 * Results: After discovering the file extension, and purpose we researched audio converters to help making .wav file. We were able to find Sox audio converter. With this ability


 * Plan:


 * Concerns:

Week Ending February 26, 2013

 * Task:


 * Results:


 * Plan:


 * Concerns:

Week Ending March 5, 2013

 * Task:


 * Results:


 * Plan:


 * Concerns:

Week Ending March 12, 2013

 * Task:


 * Results:


 * Plan:


 * Concerns:

Week Ending March 26, 2013

 * Met with new team members. Eric offers a great deal of leadership, and direction. Exchanged emails with other team members. We discussed two week plan of modeling and experimenting. After plan was discussed, we broke our teams into individual assignments.
 * Contacted Matt for transcripts without audio files.


 * Read logs from group members. They all seem pretty current and up to date. Went over experiment logs and found them to be pretty straight forward.
 * Matt emailed me with transcripts, marked the appropriate files, and separated them core data to run scripts to find time, word count, word frequency.


 * Eric updated the procedures. Try logging into Ceasar and was denied access. Having trouble trying to get my password sorted out. I will Discuss with Mr. Jonas on resetting my profile and password.
 * Try running previous scripts for the transcripts. Seems to be an error occuring on my computer. I will discuss with Matt possible solutions on wed class.


 * Without the ability to log in. I read logs, looked at other experiments to gain a feel for what is expected

Tasks

 * Run an Experiment, and log any errors that could possible occur.


 * Run Transcript data through Scripts to have transcripts analyzed

Plan

 * Contact Mike Jonas about login problems. Mostly likely will come in early to discuss resetting login information


 * Run experiment train and decoder, log any errors occured, and email Eric with the results


 * Work with Matt on transcript files.


 * Read other experiments

Concerns

 * Being Unable to login makes me effectively useless. Reading the logs and experiment data allows me to be knowledgeable, but unable to contribute makes me an ineffective team member.

Logs

 * Discussed goals with the groups. Was able to fix my Login information with Caesar. Started running a train, and was able schedule an online meeting through Google Meet.
 * Read experiment notes, and instructions. Read logs from modeling group for the semester. Refreshed myself in Linux commands


 * Started running my experiment. Ran into compiling errors with GenTrans. Contacted Eric with the issues and was able to solve them.
 * Started Google Hangout. Eric instructed us on the decoding steps, and went over commonly occurring problems.


 * Running into problems with gentrans.pl. Try following the troubleshoot for the GenTrans.pl, caused additional problems with compiling.
 * Going to talk to Eric with possible solutions, and figuring out my next step.


 * Read Logs for group members. Talked with Matt with problems with Scripting errors.


 * Started new experiment train 0071


 * Task:
 * Run and log a successful experiment.
 * Help others if needed


 * Results:
 * Once experiment is finished, post results and make myself available for others. Maybe run an additional experiment


 * Plan:
 * Start Training, decoding and scoring. Get help if needed, and help others if needed. Maintain contact with other group members
 * Concerns:
 * The compilations errors have been a severe issue. Debating starting a new Experiment, work more with the dictionary. Look at Master list for dictionary. Maybe post common words that master dictionary may have left out.

Week Ending April 9, 2013

 * Task:
 * Set up Corpus for for 5 hour, and 30 min test.
 * Work on experiments
 * Read Logs, and Maintain contact for assistance.


 * Results:
 * Corpus Setup: Talked with Eric, and narrowed down the audio files needed.
 * Read the logs for corpus set up, discussed with Eric on filter commands, to steamline the process.
 * Split the work between me and Eric after a few hiccups occured
 * Tried to work on experiments, Corpus set up took a little longer than expected after an error occurred when the last half our turned out to be 3 hours of audio. Was able to reset the corpus, and gather the proper files needed.
 * Read Logs on experiments from other group members. Answered a few questions and concerns involving the Corpus set up.
 * Read Logs on on group members to see if there is anything I can contribute with their experiment.


 * Plan: Using the data table Matt provided, Able to calculate the audio files needed for the last 5 hours, and the first half hour for the test.Talk with Eric about the corpus setup, and find location of audio and text based transcripts. Verify the first half an hour, matches with the 5 hour transcripts


 * Corpus Set up
 * Start Experiments
 * Read logs
 * Maintain contact with other group members
 * Concerns: My only concerns are the scripting. The scripting will either make my life easier or harder. Hopefully we can streamline the corpus setup.

Week Ending April 16, 2013

 * Tasks
 * Set up experiment
 * Run Experiment
 * Score Experiment
 * Read Logs


 * Results
 * To train us in building a dictionary, actually running an experiment and scoring
 * Keep updated with current methods
 * Solve scripting problems
 * Make use of other people logs to fix errors in my own experiments


 * Plan:
 * Basically to just learn to run an experiment on my own. Then use combined knowledge from the group members to score better on future experinments


 * Concerns:
 * Scripting and training errors and/or mistakes.

Week Ending April 23, 2013
* Make P0ster * Read Logs * Get Data Abstract
 * Task:

* A poster to submit to the conference to show what we are doing in class. * Write an Abstract as part of the assignment describing our goal for the semester
 * Results:

* Using the proper formatting tools available and software. To summarize the goal for the semester in the data group. Counting the audio file times and comparing to the transcript file times.
 * Plan:


 * Concerns:
 * The only concern I have is if the information is complete

Week Ending April 30, 2013

 * Task:
 * Did more group with experiments with 0094 and 0095
 * Set up tasks directories, enabled permissions(*), generated transcripts, and setup config files for 0094 and 0095
 * Encountered an error with permissions. I enabled them as root, and messed with the groups permissions
 * Read Logs


 * Results:
 * Beside the error, the experiments were able to run correctly.
 * We were able to successfully divide work amongst the group.
 * Successful experiments


 * Plan:
 * Divide group up, and complete experiments in an orderly fashion without confusion


 * Concerns:
 * Not as much communication amongst the group this week, but it was due more to knowing what to do, rather than not talking.
 * I fear i am bugging Eric too much.

Week Ending May 7, 2013

 * Task:
 * Run experiments with the updated language model and genTrans6.pl, to improve scoring.
 * Ran Language model, decode, scoring, and posted an experiment Wiki on it.
 * Read logs to review for grading
 * Maintain contact with group members


 * Results: Score slightly improved by ten points. Experiments ran successfully. We discovered that genTrans6.pl that wasn't much of an improvement.


 * Plan:
 * Divide the work up within the group
 * Run Experiments
 * Post Results


 * Concerns:
 * We had a slight problem with Caesar over the weekend. After email exchanges amongst the group. Tyler was able to get in and fix the issue
 * Basic errors within the experiment build. (which were none to my knowledge)