Good ASR Training System

The term "speech recognition" is a misnomer.

Why do I say that? I have explained this point in an old article "Do We Have True Open Source Dictation?, which I wrote back in 2005: To recap,  a speech recognition system consists of a Viterbi decoder, an acoustic model and a language model.  You could have a great recognizer but bad accuracy performance if the models are bad.

So how does that related to you, a developer/researcher of ASR?    The answer is ASR training tools and process usually become a core asset of your inventories.    In fact, I can tell you when I need to work on acoustic model training, I need to spend full time to work on it and it's one of the absorbing things I have done.  

Why is that?  When you look at development cycles of all tasks in making an ASR systems.   Training is the longest.  With the wrong tool, it is also the most error prone.    As an example, just take a look of Sphinx forum, you will find that majority of non-Sphinx4 questions are related to training.    Like, "I can't find the path of a certain file", "the whole thing just stuck at the middle".

Many first time users complain with frustration (and occasionally disgust) on why it is so difficult to train a model.   The frustration probably stems from the perception that "Shouldn't it be well-defined?"   The answer is again no. In fact how a model should be built (or even which model should be built) is always subjects to change.   It's also one of the two subfields in ASR, at least IMO, which is still creative and exciting in research.  (Another one: noisy speech recognition.)  What an open source software suite like Sphinx provide is a standard recipe for everyone.

Saying so, is there something we can do better for an ASR training system?   There is a lot I would say, here are some suggestions:

1. A training experiment should be created, moved and copied with ease,
2. A training experiment should be exactly repeatable given the input is exactly the same,
3. The experimenter should be able to verify the correctness of an experiment before an experiment starts. 

Ease of Creation of an Experiment

You can think of a training experiment as a recipe ...... not exactly.   When we read a recipe and implement it again, we human would make mistakes.

But hey! We are working with computers.   Why do we need to fix small things in the recipe at all? So in a computer experiment, what we are shooting for is an experiment which can be easily created and moved around.

What does that mean?  It basically means there should be no executables which are hardwired to one particular environment.   There should also be no hardware/architecture assumption in the training implementations.   If there is, they should be hidden.


Repeatability of an Experiment


Similar to the previous point, should we allow difference when running a training experiment?  The answer should be no.   So one trick you heard from experienced experimenters is that you should keep the seed of random generators.   This will avoid minute difference happens in different runs of experiments.

Here someone would ask.   Shouldn't us allow a small difference between experiments?  We are essentially running a physical experiment.

I think that's a valid approach.  But to be conscientious, you might want to run a certain experiment many times to calculate an average.    In a way, I think this is my problem with this thinking.  It is slower to repeat an experiment.    e.g.  What if you see your experiment has 1% absolute drop?  Do you let it go? Or do you just chalk it up as noise?   Once you allow yourself to not repeat an experiment exactly, there will be tons of questions you should ask.


Verifiability of an Experiment

Running an experiment sometimes takes day, how do you make sure running it is correct? I would say you should first make sure trivial issues such as missing paths, missing models, or incorrect settings was first screened out and corrected.

One of my bosses used to make a strong point and asked me to verify input paths every single time.  This is a good habit and it pays dividend.   Can we do similar things in our training systems?

Apply it on Open Source

What I mentioned above is highly influenced by my experience in the field.   I personally found that sites, which have great infrastructure to transfer experiments between developers, are the strongest and faster growing.   

To put all these ideas into open source would mean very different development paradigm.   For example, do we want to have a centralized experiment database which everyone shares?   Do we want to put common resource such as existing paramatized inputs (such as MFCC) somewhere in common for everyone?  Should we integrate the retrieval of these inputs into part of our experiment recipe? 

Those are important questions.   In a way, I think it is the most type of questions we should ask in open source. Because regardless of much volunteer's effort.  Performance of open source models is still lagging behind the commercial models.  I believe it is an issue of methodology.  

Arthur

sphinxbase 0.8 and SphinxTrain 1.08

I have done some analysis on sphinxbase0.8 and SphinxTrain 1.08 and try to understand if it is very different from sphinxbase0.7 and SphinxTrain1.0.7.  I don't see big difference but it is still a good idea to upgrade.

• (sphinxbase) The bug in cmd_ln.c is a must fix.  Basically the freeing was wrong for all ARG_STRING_LIST argument.  So chances are you will get a crash when someone specify a wrong argument name and cmd_ln.c forces an exit.  This will eventually lead to a cmd_ln_val_free. 
• (sphinxbase) There were also couple of changes in fsg tools.  Mostly I feel those are rewrites.  
• (SphinxTrain) sphinxtrain, on the other hands, have new tools such as g2p framework.  Those are mostly openfst-based tool.  And it's worthwhile to put them into SphinxTrain. 

One final note here: there is a tendency of CMUSphinx, in general, starts to turn to C++.   C++ is something I love and hate. It could sometimes be nasty especially dealing with compilation.  At the same time, using C to emulate OOP features is quite painful.   So my hope is that we are using a subset of C++ which is robust across different compiler version. 

Arthur 

On sh*tty job.

I read "Why Hating Your Shitty Job Only Makes It Worse",  there is something positive about the article but I can't completely agree with the authors.

Part of the dilemma at work in a traditional office space is that inevitably some kind of a*holes and bad system will appear in your life.   The question is whether you want to ignore it or not.   You should be keenly aware of your work condition and make rational decision of staying an leaving.

Arthur

Two Views of Time-Signal : Global vs Local

As I have been working on Sphinx at work and start to chat with Nicholay more, one thing I realize is that several frequently used components of Sphinx need to rethink.  Here is one example  related to my work recently.

Speech signal or ...... in general time signal can be processed in two ways: you either process as a whole, or you process in blocks.  The former, you can call it a global view, the latter, you can call it a local view.  Of course, there are many other names: block/utterance, block/whole but essentially the terminology means the same thing.

For most of the time, global and local processing are the same.   So you can simply say: the two types of the processing are equivalent.

Of course, not when you start to an operation which use information available.   For a very simple example, look at cepstral mean normalization (CMN).  Implementing CMN in block mode is certainly an interesting problem.  For example, how do you estimate the mean if you have a running window?   When you think about it a little bit, you will realize it is not a trivial problem. That's probably why there are still papers on cepstral mean normalization.

Translate to sphinx, if you look at sphinxbase's sphinx_fe, you will realize that the implementation is based on the local mode, i.e. every once in a while, samples are consumed, processed and write onto the disc.    There is no easy way to implement CMN on sphinx_fe because it is assumed that the consumer (such as decode, bw) will do these stuffs their own.

It's all good though there are interesting consequence: what the SF's guys said about "feature" is really all the processing that can be done in the local sense.   Rather than the "feature" you see in either the decoders or bw.

This special point of view is ingrained within sphinxbase/sphinxX/sphinxtrain (Sphinx4? not sure yet.) .  This is quite different from what you will find in HTK which see feature vector as the vector used in Viterbi decoding.

That bring me to another point.  If you look deeper, HTK such as HVite/HCopy are highly abstract. So each tool was designed to take care of its own problem well. HCopy really means to provide just the feature, whereas HVite is just doing Viterbi algorithm on a bunch of features.   It's nothing complicated.  On the other hand, Sphinx are more speech-oriented.  In that world, life is more intertwined.   That's perhaps why you seldom hear people use Sphinx to do research other than speech recognition.  You can, on the other hand, do other machine learning tasks in HTK.

Which view is better?  If you ask me, I hope that both HTK and Sphinx are released in Berkeley license.  Tons of real-life work can be saved because each cover some useful functionalities.

Given that only one of them are released in a liberal license (Sphinx),  then may be what we need is to absorb some design paradigm from HTK.  For example, HTK has a sense of organizing data as pipes.   That something SphinxTrain can use.   This will enhance work of Unix users, who are usually contribute the most in the community.

I also hope that eventually there are good clones of HTK tools but made available in Berkeley/GNU license.  Not that I don't like the status quo: I am happy to read the code of HTK (unlike the time before 2.2......).   But as you work in the industry for a while, many are actually using both Sphinx and HTK to solve their speech research-related problems.   Of course, many of these guys  (, if they are honest,) need to come up with extra development time to port some HTK functions into their own production systems.  Not tough, but you will wonder whether time can be better spent ......

Arthur

Readings at Jan 8, 2012

Testing Redux  by Vivek Halder

Comment: I second Vivek,  in a large scale project, having no testing is a project-killing act.  One factor to consider: in real-life, the mythical 100X productive programmer are rarely seen.   Even then, these programmers can make a mistake or two.   Therefore, not having any automatic testing  for a group is a very bad thing. 

On the other, should an individual programmer always follow automatic testing?   Yes and no.  Yes in the sense you should always write a test for your program.  No in the sense that you shouldn't believe testing will make your program automatically correct.   

Bring back the 40-hour work week by Sara Robinson

Comment: very well said.  I hope this more hours = more work done crap can end soon. 

Todon't by Jeff Atwood

Comment: I like todo list alot but sometimes it takes me a while to organize them.  They also grow very quickly.  A good suggestions here is that not only you should add to your list, you should always change priorities.  One of the purposes of todo is to record your life but it has nothing to do with how you move forward with your life. 

Arthur

You got to have thick skin......

Linux Chews Up Kernel Maintainer for Introducing UserSpace Bug.

http://developers.slashdot.org/story/12/12/29/018234/linus-chews-up-kernel-maintainer-for-introducing-userspace-bug

That's just the way it is.  Remember, there are always better, stronger, more famous, more powerful programmers working with you.  So criticisms will come one day.

My take, as long as if you don't work with them in person, just treat them as an icon on the screen.   :D

Arthur

Where to start when tracing source code of a speech recognition toolkit?

Modern speech recognition software are complicated piece of software.  To understand it, you need to have some basic understanding of the principle of speech recognition, as well as some ideas on the programming language being used.

By now, you may hear a lot of people say they know about a speech recognizer.   And by now, you probably realize that most of these people have absolutely no ideas what's going on inside a recognizer.   So if you are reading this blog message, you are probably telling yourself, "I might want to trace the codebase of some recognizers' code." Be it Sphinx, HTK, Julius, Kaldi or whatever codebase you are looking at.

For the above toolkits, I will say I only know in detail about Sphinx,  probably a little bit about HTK's HVite.  But I won't say the same for others.  In fact, even in Sphinx, I only know intimately about Sphinx 3/SphinxTrain/sphinxbase triplet.   So just like you, I hope to learn more.

So here it begs the question: how would you trace a speech recognition toolkit codebase? If you think it is easy, probably because you worked in speech recognition for a while and you probably shouldn't read this post.

Let's just use sphinx as an example, there are hundreds of files in each component of Sphinx.   So where should you start?    A blunt approach would be reading each of the file one by one.   That's not a smart the way.   So here is a suggestion for you : focus on the following four things,

1. Viterbi algorithm
2. Workflow of training
3. Baum-Welch algorithm. 
4. Estimation algorithms of language models. 

When you know where the Viterbi algorithm is, you will soon figure out how the feature vector is generated.  On the same vein: if you know where the Baum-Welch algorithm, you will probably know how the statistics are generated.   If you know the workflow of the training, then you will understand the how the model is "evolved".   If you know how the language model is estimated, then you would have understanding of one of the most important heuristic of the search. 

Some of you may protest, how about the front-end? Isn't that important too?  True, but not when you try to understand a codebase.  For all practical purpose, a feature vector is just an N-dimensional vector.  The waveform is just an NxT matrix.   You can certainly do a lot of fancy things on this NxT matrix.   But when you think of Viterbi and Baum-Welch, they probably just read the frames and then calculate Gaussian distribution.  That's pretty much it's how much you want to know a front-end. 

How about adaptation algorithms?  That I think it's important.   But it should probably go after understanding of the major things in the code.   Because no matter whether you are doing adaptation online or doing this in speaker adaptive training.  It is something on top of the Baum-Welch algorithm.   Some implementation stick adaptation within the Baum-Welch executable.  There is certainly nothing wrong about it.   But it is still a kind of add-on. 

How about decoding API?  Those are useful things to know but it is more important when you just need to write an application.  For example, in Sphinx4, you just need to know how to call the Recognizer class.  In sphinx3, live_decode is what you need to know.   But only understanding those won't give you too much insights of how the decoder really works. 

How about the data structure?  Those are sort of important and should be understood when you try to understand a certain algorithm.   In the case of languages such as Java and C++, you should probably take notes on a custom-made data structure.  Or whether the designer call a specific data structure libraries.  Like Boost in C++. 

I guess this pretty much sums it all.  Now let me get back to one non-trivial item on the list, which is the workflow of training.   Many of you might think that recognition systems differ from each other because they have different decoders.  Dead wrong!  As I stressed from time to time, they differ because they have different acoustic models and language models.  So that's why in many research labs, much effort was put on preserving the parameters and procedures of how models is trained.  Much effort was also put to fine tuned this procedure.  

On this part,  I got to say open source speech recognition still has long long way to go.  For starter, there is no much sharing of recipes among speech hobbyists.   What many try to do is to search for a good model.   If you don't know how to train a model, you probably don't even know how to improve it for you own project.   

Arthur

Pondering Unix Philosophy

These are two great articles by James Hague on text processing vs visual programming.

The Unix Philosophy and a Fear of Pixels
Living inside your own Black Box


His main point is visual programming is often dismissed because it is way more difficult than text processing.    It is a little bit like a lot of "stupid" things in the world such as Windows programming.   They are actually quite tough to do well.

On speech processing, I guess it is appropriate to think sound programming is tougher than text processing as well.   You may even think in speech processing, no one come up with a generic "Sound User Interface" IDE yet.

Arthur

Me and CMU Sphinx

As I update this blog more frequently, I noticed more and more people are directed to here.   Naturally,  there are many questions about some work in my past.   For example, "Are you still answering questions in CMUSphinx forum?"  and generally requests to have certain tutorial.  So I guess it is time to clarify my current position and what I plan to do in future.

Yes, I am planning to work on Sphinx again but no, I probably don't hope to be a maintainer-at-large any more.   Nick proves himself to be the most awesome maintainer in our history.   Through his stewardship, Sphinx prospered in the last couple of years.  That's what I hope and that's what we all hope.    

So for that reason, you probably won't see me much in the forum, answering questions.  Rather I will spend most of my time to implement, to experiment and to get some work done. 

There are many things ought to be done in Sphinx.  Here are my top 5 list:

1. Sphinx 4 maintenance and refactoring
2. PocketSphinx's maintenance
3. An HTKbook-like documentation : i.e. Hieroglyphs. 
4. Regression tests on all tools in SphinxTrain.
5. In general, modernization of Sphinx software, such as using WFST-based approach.

This is not a small undertaking so I am planning to spend a lot of time to relearn the software.  Yes, you hear it right.  Learning the software.  In general, I found myself very ignorant in a lot of software details of Sphinx at 2012.   There are many changes.  The parts I really catch up are probably sphinxbase, sphinx3 and SphinxTrain.   One PocketSphinx and Sphinx4, I need to learn a lot. 

That is why in this blog, you will see a lot of posts about my status of learning a certain speech recognition software.   Some could be minute details.   I share them because people can figure out a lot by going through my status.   From time to time, I will also pull these posts together and form a tutorial post. 

Before I leave, let me digress and talk about this blog a little bit: other than posts on speech recognition, I will also post a lot of things about programming, languages and other technology-related stuffs.  Part of it is that I am interested in many things.  The other part is I feel working on speech recognition actually requires one to understand a lot of programming and languages.   This might also attract a wider audience in future. 

In any case,  I hope I can keep on.  And hope you enjoy my articles!

Arthur

Some Reflections on Programming Languages

This is actually a self-criticizing piece.  Oh well, but call it reflection doesn't hurt.

When I first started out in speech recognition, I have a notion that C++ is the best language in the world.  For daily work? "Unix commands such as cut, split work well. "  To take care of most of my processing issues, I used some badly written bash shell.  Around the middle of the grad school, I started to learn that perl is lovely for string processing.   Then I thought perl is the best language in the world, except it is a bit slow.

After C++ and perl, I then learned C, Java, Python.  A little bit of objective-C and sampled many other languages.   For now, I will settle on C and Perl are probably the two languages I am most proficient.  I also tend to like them the most.   There is one difference between me and the twenty-something me though - instead of arguing which language is the best, I will simply go to learn more about any programming language in the world.

Take C as an example, many would praise it to be the procedure language which is closest to the machine.  I love to use C and write a lot of my algorithms in C.  But when you need to maintain and extend a C codebase, it is a source of a pain because, there is no inherent inheritance mechanism to work with, so a programmer needs to implement their own class-implementation.  Many function pointers.  There is also no memory-checking, so an extra step of memory checking is necessary.  Debugging is also a special skill.

Take perl.  It is very useful in text processing and has very flexible syntax.   But this flexibility also makes perl script hard to read sometimes.    For example, for a loop, do you want to implement it as a foreach-loop or by a map?   Those confuse lesser programmers.  Also, when you try to maintain large scale project with perl, many programmers remark to me OOP in perl seems to "just organize the code better".

How about C++?  We love the templates, we love the structure.   In practice though, the standard changes all the time.  Most house fixes the compiler version to make sure their C++ source code compiled.

How about Java?  There is memory boundary checking.  After a year or two on a dot-com, I also learned that Tomcat servlet is a thing in web development.   It is also easy to learn and one mainstream programming language taught in school these days.  Those I dig.  What's the problem? You may say speed is an issue.  Wrong.  Many Java code can be optimized such that it is as fast as its C or C++ codebase.   The issue in practice is that the process of bytecode conversion is non-trivial to many.  That is why it raises doubts in a software team on whether the language is the cause of speed issues.  

For me, I also care about the fate of Java as an open language after Oracle bought Sun Microsystem.

How about Python?  I guess this is a language I know least about.  So far, it seems to take care of a lot of problems in perl. I found the regular expression takes some time to learn.  Though other than that, the language is quite easy to learn and quite nice to maintain.  I guess the only thing I would say it is the slight difference between different Python 2.X starts to annoy me.

I guess a more important point here:  every language has its strength and weakness.  In real life, you probably need to prepare to write the same algorithm in all languages you know.   So there is no room for you to say "Hey! Programming language A is better than programming language B. Wahahaha.  Since I am using A rather than B, I rock, you suck!"  No, rather you got to accept that writing in unfamiliar language is essential for tech person's life.

I learned this through my spiritual predecessor, Eric Thayer, who organized the source code of SphinxTrain.  He once said to me, (I rephrase here,) "Arguing about programming languages is one of the most stupidest thing in the world."

Those words enlightened me.

Perhaps that is why I have been reading "C Programming a Modern Approach", "The C++ Programming Language",  "Java in a Nutshell", "Programming Perl" and "Programming Python" from time to time because I never feel satisfy with my skills on any of them.  I hope to learn D and Go soon and make sure I am proficient in Objective-C soon.  It will take me a lifetime to learn them, but on something deep like programming, learning, other than arguing, seems to be a better strategy to go.

Arthur

Passion

For a period of time, getting up is a daunting thing to me.   You see...... computers used to be a tool to let me realize myself.  I like to work, play with one.  It was not a job.

Since when it is changed for me?  It was the time when I think of a computer to be solely a tool of making money.   That's how many people in the field think.  Programming is no longer a pursuit of skill.   It is a way to get higher salary, win programming competition and have bragging right on lunch table. Knowledge in speech recognition?  It is not to solve one of the biggest problem in human history.  It is for winning contracts from defense,  beating other sites and again bragging to your esteemed colleagues.   These sicken me.

In my view, it is fine to think of money issue.  In fact, everyone should take care of their own personal finance and have basic understanding of economics...... BUT......  It doesn't mean everything has to be driven solely money.   

Rather, everyone should have passion, which allows them to wake up everyday, not being daunted by the workload of the day, but think "Woa,  there are 10 cool things I want to do.  What should I work on today?" and feel excited about life. 

Arthur

How to Ask Questions in the Sphinx Forum?

Many go to different open source toolkits to look for a ready-to-use speech recognizer, and seldom get what they want.   Many feel disappointed and curse that developers of open source speech recognizer just couldn't catch up with commercial product.   Few know why and few decide to write about the reason.

People in the field blame Hollywood for lion share of the problem.  Indeed, many people believe ASR should work similarly to scenes of Space Odyssey 2001 or Star Trek.   We are far far away from there.   You may say SIRI is getting close.  True.   But when you look closer, SIRI doesn't always get what you say right, her strength lies on the very intelligent response system.

Unlike compilers such as GCC, speech recognition toolkit such as the CMU Sphinx project HTK are toolkits.   The mathematical models these toolkits provided were trained and fit to certain group of samples. Whereas, applications such as Google Voice or SIRI gather 100 or even 1000 times more data when they train a model.   This is the fundamental reason why you don't get the premium recognition rate you think you entitled to.

Many people (me included) saw that as a problem.  Unfortunately, to collect clean transcribed data has always been a problem.   Voxforge is the only attempt I am aware of to resolve the issue.    They are still growing up but it will be a while they can collect enough data to rival with commercial applications.

* * *
Now what does that tell you when you ask questions in CMU Sphinx or other speech recognition forum?   For users who expect out-of-the-box super performance, I would say "Sorry, we are not there yet."  In fact, speech recognition, in general, is probably not in performance shown in the original Star Trek yet (that will require accent adaptation and very good noise cancellation since the characters seem to be able to use the recognizer any time they like).

How about many users who have a little bit (or much) programming background? I would say one thing important.  As a programmer, you probably get used to look at the code, understand what it's done, do something cute and feel awesome from time-to-time.  You can't do that if you seriously want to develop a speech recognition system.

Rather, you should think like a data analyst.  For example, when you feel the recognition rate is bad, what is your evidence?  What is your data set?  What is the size of your data set? If you have a set, can you share the set?   If you don't have numerical measure, have you at least use pencil or paper to mark down at least some results and some mistakes? Report them when you ask questions, then you will get useful answers back.

If you go to look at programming forum, many ask questions with the source such that people can repeat the problem easily.    Some even go further to pinpoint location of the problem.    This is probably what you want to do if you get stuck.

* * *

Before I end this post, let's also bring up the issue of how usually ASR problem is solved?  Like...... if you see performance is bad, what should you do?

Some speech recognition problems can be solved readily.  For example, if you try to recognize digit strings but only get one digit at a time, chances are your grammar was written incorrectly.  If you see completely crappy speech recognition performance, then I will first check if the front-end of decoder match exactly as the front-end used to train the models.

For the rest,  the strength of the model is really the issue.   So most of your time should spend on learning and understanding techniques of model improvement.    For example, do you want to collect data and boost up your acoustic model?  Or if you know more about the domain, can you crawl some text on the web and help your language model?   Those are the first ideas you should think about.

There are also an exoteric group of people in the world who ask a different question, "Can we use a different estimation algorithm to make the better?"  That is the basis of MMIE, MPE and MFE.   If you found yourself mathematically proficient (perhaps need to be very proficient......), then learning those techniques and implement some of them would help boosting up the performance as well.   What I mentioned such as MMIE are just the basics,  each site has their own specialized technique and you might want to know.

Of course, you normally don't have to think so deep.   Adding more data is usually the first step of ASR improvement.    If you start to think something advance and if you can,  please try to put your implementation somewhere public such that everyone in the world can try it out.   These are something small to do, but I believe if we keep on doing something small right, there will be a day we can make open source speech recognizers as the commercial ones.

Arthur

Landscape of Open Source Speech Recognition software at the end of 2012 (I)

As I am back, I start to visit all my old friends - all open source speech recognition toolkits.  The usual suspects are still around.  There are also many new kids in town so this is a good place to take a look.

It was a good exercise for me, 5 years of not thinking about open source speech recognition is a bit long.   It feels like I am getting in touch with my body again.

I will skip CMU Sphinx in this blog post as you probably know something about it if you are reading this blog.   Sphinx is also quite a complicated projects so it is rather hard to describe  entirely in one post.   This post serves only as an overview.  Most of the toolkit listed here have rich documentation.   You will find much useful information there.

HTK

I checked out the Cambridge HTK web page.  Disappointingly, the latest version is still 3.4.1, so we are still talking about MPE and MMIE, which is still great but not as exciting as other new kids in town such as KALDI.   

HTK has always been one of my top 3 speech recognition systems since most of my graduate work are done using HTK.   There are also many tricks you can do with the tools.   

As a toolkit, I also find its software engineering practice admirable.   For example, the software command was based on common libraries written beneath.  (Earlier versions such as 1.5 or 2.1 would restrict access to the memory allocation library HMem.)   When reading the source code, you feel much regularities and there doesn't seem to be much duplicated code. 

The license disallows commercial use but that's okay.  With ATK, which is released in a freer license, you can also include the decoder code into a commercial application.

Kaldi

The new kid in town.   It is headed by Dr. Dan Povey, who researched many advanced acoustic modeling techniques.   His recognizers attract much interest as it has implemented features such as subspace GMM and FST-based speech recognizer.   Of all, this features feel like more "modern". 

I only have little exposure on the toolkit (but determined to learn more).   Unlike Sphinx and HTK, it is written in C++ instead of C.   As of this writing, Kaldi's compilation takes a long time and the binaries are *huge*.   In my setup, it took me around 5G of disc space to compile.   It probably means I haven't setup correctly ...... or more likely, the executable is not stripped.   That means working on Kaldi's source code actively would take some discretion in terms of HD.  

Another interesting part of Kaldi is that it is using weighted finite state transducer (WFST) as the unifying knowledge source representation.   To contrast this, you may say most of the current open source speech recognizers are using ad-hoc knowledge source.   

Are there any differences in terms of performance you ask?  In my opinion, probably not much if you are doing an apple to apple comparison.   The strength of using WFST is that when you need to introduce new knowledge,  in theory you don't have to hack the recognizer.  You just need to write your knowledge in an FST and compose it with your knowledge network, then you are all set. 

In reality, the WFST-based technology seems to still have practice problem.  As the vocabulary size goes large and knowledge source got more complicated, the composed decoding WFST would naturally outgrow the system memory.   As a result, many sites propose different technique to make decoding algorithm works.  

Those are downsides but the appeal of the technique should not be overlooked.   That's why Kaldi becomes one of my favorite toolkits recently. 

Julius

Julius is still around!  And I am absolutely jubilant about it.  Julius is a high-speed speech recognizer which can decode a 60k vocabulary. One speed-up techniques of Sphinx 3.X was context-independent phone Gaussian mixture model selection (CIGMMS) and I borrowed this idea from Julius when I first wrote.  

Julius is only the decoder and the beauty of it is that it never claims to be more than that.  Accompanied with the software, there is a new Juliusbook, which is the guide on how to use the software.  I think the documentation are in greater-depth than other similar documentations. 

Julius comes with a set of Japanese models, not English.   This might be one of the reasons why it is not as popular (more like talk about) as HTK/Sphinx/Kaldi. 

(Note at 20130320: I later learned that Julius also comes with an English model now.  In fact, some anecdotes suggest the system is more accurate than Sphinx 4 with broadcast news.  I am not surprised.  HTK was as acoustic model trainer.)

So far......

I went through three of my favorite recognition toolkits.  In the next post, I will cover several other toolkits available. 

Arthur

The Grand Janitor's Blog

For the last year or so, I have been intermittently playing with several components of CMU Sphinx.  It is an intermittent effort because I am wearing several hats in Voci.

I find myself go back to Sphinx more and more often.   Being more experienced, I start to approach the project again carefully: tracing code, taking nodes and understanding what has been going on.  It was humbling experience - speech recognition has changed, Sphinx has more improvement than I can imagine. 

The life of maintaining sphinx3 (and occasionally dip into SphinxTrain) was one of the greatest experience I had in my life.   Unfortunately, not many of my friends know.  So Sphinx and I were pretty much disconnected for several years. 

So, what I plan to do is to reconnect.    One thing I have done throughout last 5 years was blogging so my first goal is to revamp this page. 

Let's start small: I just restarted RSS feeds.   You may also see some cross links to my other two blogs, Cumulomaniac, a site on my take of life, Hong Kong affairs as well as other semi-brainy topics,  and  333 weeks, a chronicle of my thoughts on technical management as well as startup business. 

Both sites are in Chinese and I have been actively working on them and tried to update weekly. 

So why do I keep this blog then?  Obviously the reason is for speech recognition.   Though, I start to realize that doing speech recognition has much more than just writing a speech recognizer.   So from now on, I will post other topics such as natural language processing, video processing as well as many low-level programming information.   

This will mean it is a very niche blog.   Could I keep up at all?  I don't know.   As my other blogs, I will try to write around 50 messages first and see if there is any momentum. 

Arthur

Self Criticism : Hieroglyph

When I was working on CMU Sphinx, I was more an aggressive young guy and love to start many projects (still am).   So I started many projects and not many of them completed.   I wasn't completely insane: what was lacking at that point of development is that we lack of passion and momentum.  So working on many things give a sense of we are moving forward.

One of the projects, which I feel I should be responsible, is the Hieroglyph.   It was meant to be a complete set of documentation for several Sphinx components work together.   But when I finished the 3rd draft, my startup work kicked in.    That's why what you can see is only an incomplete form of the document.

Fast-forward 6 years later, it was unfortunate that the document is still the comprehensive source of sphinx if you want to understand the underlying structure/method of CMU Sphinx C-based executables.     The current CMU Sphinx encompasses way more than I decided to cover.   For example, the Java-based Sphinx4 has gained much followings.   And pocketsphinx is pretty much the de-facto speech recognizer for embedded speech recognition.

If you were following me (unlikely but possible), I have personally changed substantially.   For example, my job experience taught me that Java is a very important language and having a recognizer in Java would significantly boost the project.    I also feel embedded speech recognition is probably the real future of our life.

Back to Hieroglyph, suffice to say it is not yet a sufficient document.   I hope that I can go back to it and ask what I can do to make it better.

Arthur

What should be our focus in Speech Recognition?

If you worked in a business long enough, you start to understand better what type of work are important.   As many things in life, sometimes the answer is not trivial.   For example, in speech recognition, what are the important ingredients to work on?

Many people will instinctively say the decoder.  For many, the decoder, the speech recognizer, oorr the "computer thing" which does all the magic of recognizing speech, is the core of the works.

Indeed, working on a decoding is loads of fun.  If you a fresh new programmer, it is also one of those experiences, which will teach you a lot of things.   Unlike thousands of small, "cool" algorithms, writing a speech recognizer requires you to work out a lot of file format issues, system issues.   You will also touch a fairly advanced dynamic programming problem : writing a Viterbi search.   For many, it means several years of studying source code bases from the greats such as HTK, Sphinx and perhaps in house recognizers.

Writing a speech recognizer is also very important when you need to deal with speed issues.  You might want to fit a recognizer into your mobile phone or even just a chip.   For example, in Voci, an FPGA-based speech recognizer was built to cater ultra-high speed speech recognition (faster than 100xRT).   All these system-related issues required understanding of the decoder itself.

This makes speech recognition an exciting field similar to chess programming.  Indeed the two fields are very similar in terms of code development.   Both require deep understanding of search as a process. Both have eccentric figures popped up and popped out.   There are more stories untold than told in both field.  Both are fascinating fields.

There is one thing which speech recognition and chess programming are very different.   This is also a subtle point which even many savvy and resourceful programmers don't understand.   That is how each of these machines derived their knowledge sources.   In speech, you need to have a good model to do decent jobs for your task.   In chess though, most programmers can proceed to write a chess player with the standard piece values.   As a result, there is a process before anyone can use a speech recognizer.  That is to first train an acoustic model and a language model.  

The same decoder, having different acoustic models and language models, can give users perceptions ranging from a total trainwreck to the a modern wonder, borderline to magic.   Those are the true ingredients of our magic.   Unlike magicians though, we are never shy to talk about these secret ingredients.   They are just too subtle to discuss.   For example, you won't go to a party and tell your friends that "Using an ML estimate is not as good as using an MPFE estimate in speech recognition.  It usually results in absolutely 10% performance gap."  Those are not party talks.  Those are talks when you want to have no friends. :)

In both type of tasks, one require learning different from a programming training.   10 years ago, those skill are generally carried by "Mathematician, Statistician or People who specialized in Machine Learning".   Now there is new name : "Big Data Analyst".

Before I stopped, let me mention another type of work, which are important in real life.  What I want to say is transcription and dictionary work.   If you asked some high-minded researchers in the field, they will almost think those are not interesting work.   Yet, in real-life, you can almost always learn something new and improve your systems based on them.  May be I will talk about this more next time.

The Grand Janitor

Being a programmer in your 20s and 30s

It's funny how a person changes.  I always thought 20s was my best time.  It sort of was.   Generally, that was the moment you are energetic, can burn as much as you can, naively think that life, relationship can last forever. Also, you unconditionally trust other people.  

Things will turn when you are 30s, you start to realize your skill, your prowess to growth has a limit.  In exchange, you grow wiser.  In my case, I found my reads on people are much better, I start to go behind other people's word and try to understand people's intention.  I start to treasure genuine friendship, protest contrived politeness and faked honesty.

I also start to know when is the best time to be quiet and when is the best time to give a come back.  The former is important because if you are the only one who shine, your team will perpetually has the capability of one person, who is you.  

The latter is also important because if you are always quiet, there are just people who will step on your toes harder and harder.   They will think you are weak and can be bullied.   In real life, as in the time when you are in high-school, the bullies love to bully the weak.   Making sure they have a hard time to do so, is a very important life skill.

I will never go back to the time I can hack a program for 20 hours, sleep, and then hack it again for another 20 hours.  Will I feel regret about it? Probably not, in exchange, I learn that sometimes you can solve a 20-hour problem with 2 hour, you can still sleep and make a living.   For all that matter, it seems to be a better deal. :)

The Grand Janitor