Moodle - Simple workflow

Finally, it's time to let others see the changes I have made although it just fixes a very simple bug.

The brief workflow should be as follows:

1) At the beginning of the day, check the upstream changes and push to my own repository at github.

2) Do the modification to the code.

3) git diff -U1 // check the changes

4) git status // check the correct status

5) git add [file/directory] // add the changed files or directories for commit

6) git commit -m "[track number] [module changed]: [changes]"

7) git push origin [working branch]

Then the changes are visible on the github.

Moodle - Initial try for bug fixing of MDL-36020

As a prerequisite, fixing existing Moodle bugs is crucial to the GSoC application for Moodle projects. Among the list of bugs suggested by Tim, it is hard to justify their difficulty for a beginner.

Anyway, browsing through the list and finally decide to try this one: https://tracker.moodle.org/browse/MDL-36020

From the description, it is some bug related to the re-grading function. The first step would thus to locate where these actions are processed. 

Directly diving into the codes makes me even confusing and cannot find even a start. Have to do the other way, i.e. trying the system. 

Creating a dummy course, registering a bunch of dummy users and design a set of questions to form a quiz. Using different accounts to attempt the quiz and finally the regrading button shows up.

The regrading lies in the Results section of Quiz module. In the source code it doesn't have a separate page but a set of functions. 

The codes are under the path moodle/mod/quiz/

Especially in the file moodle/mod/quiz/overview/report.php

Following are the functions responsible for the regrading:

process_actions($quiz, $cm, $currentgroup, $groupstudents, $allowed, $redirecturl)

Based on the action received, calling corresponding functions. Actions include "regrade", "regradeall", "regradealldry", "regradealldrydo".

regrade_attempt($attempt, $dryrun = false, $slots = null)

Regrade a particular quiz attempt. This is where the actual regrading happens. It is called by other methods.

regrade_attempts($quiz, $dryrun = false,

            $groupstudents = array(), $attemptids = array())

Regrade a set of attempts for this quiz. The set of attempts are specified by the input parameters. For each attempt, the regrading is done by the above regrade_attempt() function.

regrade_attempts_needing_it($quiz, $groupstudents)

Regrade all the attempts of the specified quiz that are marked as needing regrading.

count_question_attempts_needing_regrade($quiz, $groupstudents)

Count the number of attempts that needs regrading for the specified quiz.

has_regraded_questions($from, $where, $params)

Check whether there are any pending regrades to show.

clear_regrade_table($quiz, $groupstudents)

Remove all information about pending/complete regrades from database.

update_overall_grades($quiz)

Update the final grades for all attempts in the quiz.

These are probably the place to resolving the bug and a general problem for all the regrading action, we may need to fix it in the process_actions() function. 

-------------------------------------------------------------------------------

Next, how to fix the problem. From the description, the problem is due to the long processing time of the regrading which leads to the user session time out. 

Personally, I don't like the questions such as how good are you at xxx? how familiar are you with xxx? I have no idea what standards could be used to evaluated such as goodness or familiarity. No matter what standard you took, the answer will always be subjective and most of time  helpless.

Regarding Tim's question "what the session is". As I have never studies PHP in a systematical way, I only knows it keeps track of a user's information during his/her visiting of the website. And the only thing I have touched session in PHP is during last summer's GSoC. 

a) After a user logs in, the session_start() is called;

b) $_SESSION['variable']=value is used when I need to pass over some variables between different pages;

c) session_destroy() is called when the user logs out.

That's all I know about sessions in PHP.

Regarding the second question "why it might normally lock?", I have no idea. But with the help of Google, I learned following aspects:

a) From http://konrness.com/php5/how-to-prevent-blocking-php-requests/, the session locks due to multiple requests to access the same session file and could be solved by closing the session with session_write_close() before starting the long running process. This function only close the write session capability, the session variables are still readable. 

It seems a solution but how can we guarantee that there will be no session write requests after this?

Moodle - First attempt

After reading the Moodle community documents for several days, I finally decide to try it out and hope to fix several bugs before they give the final decision on my GSoC 2013 application.

Initially, I followed the steps provided in http://docs.moodle.org/dev/Git_for_developers on my current machine. First apply for those accounts and then fork a brunch on Github (https://github.com/troylee/moodle) and checking out a local copy with SHH-key setup. Then I'm totally lost where to start. That's a pretty big project with lots of source files. Need to find an IDE.

During the process of figuring out which is the popular IDE for Moodle development, I decided to use a completely clean system to do all the development. A Ubuntu 12.04 system is thus created in VirtualBox. 

1. Setup the LAMP system following instructions on https://help.ubuntu.com/community/ApacheMySQLPHP . It simply includes two steps:

First install tasksel...

    $ sudo apt-get install tasksel

... and then the LAMP stack:

    $ sudo tasksel install lamp-server

2. Then I followed http://wazai.net/2539/%E5%9C%A8eclipse%E4%B8%8A%E9%96%8B%E7%99%BCphp%E7%A8%8B%E5%BC%8F%E4%BB%A5moodle%E3%80%81egit%E3%80%81pdt%E7%82%BA%E4%BE%8B guide to setup the Eclipse environment for Moodle. 

The only change to the steps is the one that inputs "Source Git repository". Use the previously forked Github repository and my own gihub account and password. Instead of using SSH, the HTTPS is used as there is always error when I tried SSH.

3. Till now, it is almost done. Put a symbolic link under the path /var/www to the git repository folder for Moodle (for mine, it is ~/git/moodle). And make sure the folder has been granted correct permission. I simply used chmod 777 to it as this is for development not for public testing. 

Then, open the browser and type in the http://localhost/moodle . A web-based installation will automatically start for the first time. It requires the creation of a proper permission data folder moodledata, which is better to be under the /var/www path with permission right 777. 

During the installation, it will check the system environments. For my case, I the php5-curl, php5-gd, php5-xmlrpc packages are missing. Using the Ubuntu's apt-get install to get them ready and restart the apache server by /etc/init.d/apache restart. Then click the button at the bottom of the page to redo the checking until all the items are OK. 

Next, the database configuration is required. For Ubuntu, MySql is the default one for LAMP setup, the Improved MySql (native/mysqli) is thus chosen. It at first gave me errors of PHP not configured for mysqli. This was solved by uncomment the line "mysqli.allow_local_infile = On" in the /etc/php5/aphace2/php.ini as all the other lines with mysqli configurations are by default enabled.

Meanwhile, create a MySql database and user for Moodle:

Create a mysql database

    $ mysql> CREATE DATABASE moodel;

Create a mysql user

For creating a new user with all privileges on the moodle database at mysql prompt type:

    $ mysql> GRANT ALL PRIVILEGES ON moodel.* TO 'yourusername'@'localhost' IDENTIFIED BY 'yourpassword' WITH GRANT OPTION;

Next fill out the corresponding information to Moodle.

After updating the admin account information for Moodle site, the final admin workspace will be presented to you.

4. Taking Tim's suggestions on http://dev.moodle.org/mod/forum/discuss.php?d=1803 . Following configuration is done at the admin workspace:

After installing a copy of Moodle for development, the first thing you should do is:

1. Go to Site administration -> Development -> Debugging
2. Set Debug messages to DEVELOPER, and Turn on Display debug messages. (Consider turning on some of the other options too.)
3. In the administration block, search for "Cache" then
1. Turn off Cache all language strings.
2. Set Text cache lifetime to No
3. Turn on Theme designer mode
4. Turn off Cache Javascript

5. Next as the remote repository we used to setup the Eclipse is forked from Moodle and to keep it updated with changes in Moodle, the instructions in the section "Keeping your public repository update to date" on http://docs.moodle.org/dev/Git_for_developers was followed. 

Before addressing the problem, make sure the suggested git configuration is done:

Immediately after the installation, set your name and contact e-mail. The name and e-mail will become part of your commits and they can't be changed later once your commits are accepted into the Moodle code. Therefore we ask contributors to use their real names written in capital letters, eg "John Smith" and not "john smith" or even "john5677".

   git config --global user.name "Your Name"     git config --global user.email yourmail@domain.tld  

Unless you are the repository maintainer, it is wise to set your Git to not push changes in file permissions:

   git config --global core.filemode false

Then register the upstream remote:

   cd moodle     git remote add upstream git://git.moodle.org/moodle.git  

Then use following commands to keep the standard Moodle branches at your Github repository synced with the upstream repository. You may wish to store them in a script so that you can run it every week after the upstream repository is updated.

   #!/bin/sh     git fetch upstream     for BRANCH in MOODLE_19_STABLE MOODLE_20_STABLE MOODLE_21_STABLE MOODLE_22_STABLE MOODLE_23_STABLE MOODLE_24_STABLE master; do         git push origin refs/remotes/upstream/$BRANCH:$BRANCH     done

However, as in the Eclipse configuration, the git repository is using HTTPS instead of SSH, running the update shell script will require the input of both Github username and password for each push command, which is really annoying. To avoid the trouble of tying, first the SHH-Key configuration should be done (c.f. https://help.github.com/articles/generating-ssh-keys) and added to GitHub account's SHH Keys. 

Then change the repository url by:

git remote set-url origin git@github.com:[username]/moodle.git

With this successfully done, running the above update scripts will bypass the log in phase.

6. From my understanding the changes fetched from upstreams are directly pushed to the Gibhub origin. The local working copy is not affected. It may thus require to fetch/pull changes from my own Github repository and merge with my local copy. Not sure yet. To be confirmed later. 

It seems the setup is almost done to me. It's time to start fixing some bugs!

[Paper] Increased Robustness of Noisy Speech Features Using Neural Networks

This is an early paper (1999) from Nuance working on noise robust ASR.

The assumption in this paper is that the noises are additive and can be modeled simply by a bias vector either per frame or per utterance. 

This bias vector is estimated from neural networks using different knowledge sources. They found that estimating the bias vectors for log filterbanks is easier than cepstral vectors.

The paper could be found at

Increased Robustness of Noisy Speech Features Using Neural Networks

Noisy speech

As currently working on recognizing speech under noisy conditions, I'm always wondering what's the best way of solving and evaluating this problem and how human perceive speech under noisy environments.

For research purpose, we usually collect clean data and pure noise data and then generate noisy speech by combining them. Sometimes, we use some filters to create the channel distortion effects. Speech at different SNRs are created and used for evaluation. Like in the Aurora 2 dataset, 6 SNRs from 20dB to -5dB are used for evaluation.

To understand the difficulty in recognizing the noisy speech, the spectrograms of them at different SNRs are plotted. From these relatively high resolution spectrograms, at SNR0 the patterns are already quite confusing. At SNR-5 it is hard to extract speech patterns from the noise.

While in speech recognition, the FBank features used are rather low resolution to the spectrograms. Due to the value ranges, the patterns are relatively hard. That's also why usually we use CMVN to preprocess the features before sending to NNs.

   

The CMVN normalized FBank features are shown as follows. The dynamic parameters actually helps a lot to locate the patterns.

 

 Although from my experience, the dynamic coefficients are really helpful. I always have the question of whether is that because the high dimension. If we use higher dimensional static features, we will have much more detailed information, will that outperform the dynamic features? However, when I try to extract the same number of FBanks, there are several dimensions always giving 0 values. This may be saying the current feature extraction methods are in some aspects limited.
First the per utterance normalized static parts (40D) of the above features are displayed below:

 

Following are illustrations of 80 FBanks of static features.

 Although more FBanks are visually more favorable (at least to me), it is hard to say how the ASR can benefit form them. Maybe some automatically learnt features directly from the waveform signals would be helpful. 

GSoC 2012: Pronunciation Evaluation #Troy – Project Conclusions

This article briefly summarized the Pronunciation Evaluation Web Portal Design and Implementation for the GSoC 2012 Pronunciation Evaluation Project.

The pronunciation evaluation system mainly consists following components:

1) Database management module: Store, retrieve and update all the necessary information including both user information and various data information such as phrases, words, correct pronunciations, assessment scores and etc.

2) User management module: New user registration, information update, change/reset password and so on.

3) Audio recording and playback module: Recording the user's pronunciation for further processing.

4) Exemplar verification module: Justify whether a given recording is an exemplar or not.

5) Pronunciation assessment module: Provide numerical evaluation at the phoneme level (which could be aggregated to form higher level evaluation scores) in both acoustic and duration aspects.

6) Phrase library module: Allow users to create new phrases into the database for evaluation.

7) Human evaluation module: Support human experts to evaluate the users' pronunciations which could be compared with the automatically generated evaluations.

The website could be tested at http://talknicer.net/~li-bo/datacollection/login.php. Do let me know (troy.lee2008@gmail.com) once you encounter any problem as the site needs quite a lot testing before it works robustly. The complete setup of the website could be found athttp://cmusphinx.svn.sourceforge.net/viewvc/cmusphinx/branches/speecheval/troy/. More detailed functionality and implementations could be found in a more manual like report:

Although it is the end of this GSoC, it is just the start of our project that leveraging on open source tools to improve people's lives around the world using speech technologies. We are currently preparing using Amazon Mechanical Turk to collect more exemplar data through our web portal to build a rich database for improved pronunciation evaluation performance and further making the learning much more fun through gamification.

GSoC 2012: Pronunciation Evaluation Week 8 Status

In last week, thanks for Ronanki's help, I finally managed to integrate his exemplar outlier detection and pronunciation scoring scripts into our website, which makes it a more complete system. Following are the details of the two main parts:

1) Exemplar outlier detection

In the exemplar data collection part of the website, after the user recorded each phrase, the system will automatically decide whether the current recording is an exemplar or not. This process involves following steps:

a) Using ffmpeg to convert the FLV recording generated by the rtmplite to WAV file;

b) MFCC feature extraction using sphinx_fe;

c) Do alignment to get the phonetic transcription of the current phrase using sphinx_align;

d) Decode the current recording with edit distance grammar using sphinx_decode;

e) Compute phone error rate for the recognition results compared to the alignment;

f) Based on the phone error rate, the system will decide whether the current recording is an exemplar or not. Currently we require the phone error rate to be less than 50% for an exemplar, which could be further tuned for different cases.

g) The system then either navigate to next phrase for recording or ask the user to re-try the recording for the current phrase according to the above decision.

In the other part of the website, i.e. the student section, two new functionalities are added.

2) Exemplar recordings for student to study

When the student select a phrase for practice, maximum 5 exemplar recordings for this phrase from the database are selected and listed for the student to mimic and learn. Currently, the selection is randomly done. After we collected more exemplar recordings, it could be improved by selecting the top scored exemplars.

3) Student pronunciation scoring

Another important feature for student page is to score the recording he/she recorded. This is done with following steps:

a) Convert the FLV files from rtmplite to WAV using ffmpeg;

b) MFCC feature extraction using sphinx_fe;

c) Generate the phonetic alignment using sphnix_align;

d) Comparing the aligned phone dependent statistics with the average statistics from exemplars to evaluate the student's performance. 

e) Present both the acoustic and duration scores and navigate to next phrase.