Slashdot Mirror

I have been working with Mary TTS, an open source research TTS tool developed in Germany. It did not have any American voice at all like James Earl Jones, which is why I build a new voice, cmu-rms-hsmm. It's based on recordings from the artic CMU database. I do feel this guy has a deep resonant voice, somewhat James Earl Jones. It's a wonderful voice for high speed listening. Its included in the latest release of Mary TTS, version 4.1, which you can get at http://mary.dfki.de/Download. I use it with NVDA, through an open source project called Speech-Hub, which makes it easy to incorporate TTS engines with screen readers. You can download it at: http://www.speechhub.org/SpeechHub/install-w.

I use this voice 100% of the time with NVDA, but it has some drawbacks. Mainly, Mary TTS is a CPU hog, and is too laggy. However, with a fast CPU (you really want an i5 or better) and a couple gig or more of memory, it's quite usable. Still, it's 10X less efficient than any other TTS system I've worked with, which is why I want to spend some time over the coming year looking into a light version that is faster and needs less memory. I listen to e-books I translate to audio books with this voice, using Astro Nova Player on my Galaxy Nexus phone. It allows me to speed up to 6X speed, though only some blind people seems to be able to listen that fast.

That work was just published at the ACM Interactive Tabletops and Surfaces conference, so there is a rather more coherent explanation (and another video) at:
http://www.dfki.de/its2010/papers/sp199.html

(Not affiliated with either group, but ITS was a lot of fun)

First I was exited to see an open source translation project, especially in combination with an instant messenger. The I realized that they simply send the messages to Google for translation. So no access to a translation engine and the quality of Googles translations.

There seems to be openLogos, a commercial translation system that has recently become available as open source based on Linux and PostgreSQL. Guess I'll have to give that one a try, it looked rather complicated. Has anybody any experience with openLogos or similar (freely available) systems?

Since a few days this problem should be solved a few steps better - OpenLogos 1.0.0 has been released.
Like the german tech site heise reported last wednesday the GlobalWare AG published in cooperation with the DFKI (The German Research Center for Artificial Intelligence: "Deutsches Forschungszentrum für Künstliche Intelligenz") the free (as in speech and as in beer) translation system OpenLogos 1.0.0.
It's a PostgreSQL-based, command-line, GPL-licensed translator for Linux, which core is from 1967 but now has the according database to translate from/to english to/from french, german, italian, portuguese and spanish (where a direct tranlation from german to french or spanish is additionaly possible).

OpenLogos is meant to be a hybrid (mixture between static data and linguistic rules) translation system as a basic system for universities and other research centers to develop further going hybrid translation technologys.
Future versions of OpenLogos will include other database backends and a GUI.

Don't like babelfish? Make it better - the base system is there now.

Since a few days this problem should be solved a few steps better - OpenLogos 1.0.0 has been released.
Like the german tech site heise reported last wednesday the GlobalWare AG published in cooperation with the DFKI (The German Research Center for Artificial Intelligence: "Deutsches Forschungszentrum für Künstliche Intelligenz") the free (as in speech and as in beer) translation system OpenLogos 1.0.0.
It's a PostgreSQL-based, command-line, GPL-licensed translator for Linux, which core is from 1967 but now has the according database to translate from/to english to/from french, german, italian, portuguese and spanish (where a direct tranlation from german to french or spanish is additionaly possible).

OpenLogos is meant to be a hybrid (mixture between static data and linguistic rules) translation system as a basic system for universities and other research centers to develop further going hybrid translation technologys.
Future versions of OpenLogos will include other database backends and a GUI.

Don't like babelfish? Make it better - the base system is there now.

See: http://logos-os.dfki.de/

Hey,

Alex Waibel was one of the leading scientists in the Verbmobil project in 1995. The technology was pretty interesting (maintaining probability "graphs" from the Markov speech analysis through the syntactic and semantic analysis).

However, results were pretty poor due to the structure of the project (just too many people) and because many institutions really weren't interested in the project and went for their favourite research topic with a new name (that's how research in Germany works...). Perfectly possible that Mr. Waibel advanced with the topic, now 10 years after the first major trial...

Personally, I actually gave up AI completely after the ESSLLI (European Summer School on Logic, Language and Information) and promised not to touch the subject again until there were a "unified" formalism incorporating the old "symbolic" approach (predicate logic etc.) and the new statistical methods (Bayes, Markov, ...). Such a combination would be suitable both to deal with large amounts of data (statistical) and to deal with negation (only available in the symbolic appoach).

Maybe they've got it this time? It's a pitty they don't talk more about the underlying formalism.

Btw., the electrodes are probably just an enhancement of the normal speech recognition software to get a better "signal".

Bests,
Frank

http://www.project-open.com/

Open Source Artificial Intelligence requires a clunker old computer that can run Java, JavaScript, Forth and so on -- that's all.

The Stanford AI Lab (SAIL) has a slight advantage with some good beaches nearby.

The MIT AI Lab has a lot of old AI curmudgeons to confab with.

The German AI Institute -- davor schreckt man zurueck.

The Singularity Institute for Artificial Intelligence does the scutwork of informing the world population about what the Mentifex AI Lab is quietly, inexorably doing.

True AI is here.

As for the geographic location, I'm not sure it's as beautiful as you're making it out to be. Saarbruecken is possibly one of the ugliest cities in the former West Germany and much of the surrounding area is dotted with steel mills and coal mines (many closed now though). There is some nice forest land though.

I lived there for almost two years, doing my MS in graphics (although I was associated with the group at the MPI, not the one mentioned in the article). Interesting place to be for sure.

OLDMAILDIR = $MAILDIR
MAILDIR = $ARCHIVE_DIR
:0 cW: archive.lock
| /bin/gzip >>mailarchive-`date +%Y%m`.gz
MAILDIR = $OLDMAILDIR

I use grepmail to find old emails that I might need. Grepmail lets you use perl regular expressions to find messages and then outputs the entire message where a match was found. You can use grepm to open grepmail matches as a mailbox in mutt. grepine does the same for Pine, which I use.

At the end of each year I clean the spam out of my archives using a procmail recipe and spamassassin. This recipe marks messages as deleted in the mailbox. I open these in pine, sort by deleted, and double check them. Once I'm sure they're all spam, I delete them:

# vim:ft=procmail:

LINEBUF = 8192
SHELL = /bin/sh
MAILDIR = $HOME/mail

:0 fW: spamclean.lock
| spamassassin -e --prefs-file=/home/matt/.spamassassin/user_prefs-s pam_clean 2>/dev/null

# If the message was deemed to be spam, set the status to "deleted" so that
# we can delete it easily and optionally review it.
:0 e
{
:0 fhw
* ^^rom[ ]
| sed -e '1s/^/F/'

:0 f: formail.lock
| formail -I 'X-Status: D'
}

# Fix the mangled "From" line
:0 fhwE
* ^^rom[ ]
| sed -e '1s/^/F/'

# Remove the last of the SpamAssassin headers
:0 f: formail2.lock
| formail -I 'X-Spam-Checker-Version'

# File message in temporary mailbox
:0: sandbox.lock
z-cleaned_mbox

The special spamassassin config turns off bayesian filtering and sets the threshold high:

required_hits 15
clear_headers
fold_headers 0
use_bayes 0

html
html
text and ps
pdf

It doesn't seem to be more than a bit of bad psycho-babble quoting a lot of doubtful 60ies "psychology" books. He's even referring to some books on self-help psychology, that's probably the most dangerous thing you could do. Besides that, it seems to be some way of marketing his own book under
"Related projects".

However, there's not a single reference to any kind of computer linguistics or AI work at all. The German VERBMOBIL project comes to my mind. It was pretty much state of the art in computer linguistics, live translation of common language German - English - Japanese, although it was restricted to a special application domain, to keep the vocabulary small. All they did was translate, which only needs a very shallow understanding of a sentence's meaning. What this guy claims to do is actually respond in a meaningful way! Additionally, he just claims to make up some kind of huge dictionnary with obscure ethical terms and some kind of definition in a AI-wise completely outfashioned expert systems.

Well, surprise, one of the interesting findings in the VERBMOBIL project was, that humans actually make up words or at least redefine the meaning of words during an verbal interaction. How does he handle that?

As a last note, he has some funny opinions about 5th generation computing. In a revolutionary way, according to the usual unnamed experts, it will be made of logic circuits! No?! It can't be. That's incredible. Logic! I have to take a break now. I'm shocked.

The (perceived?) problem exists, and is preventing certain software from being released as GPL.

In particular, the ActiveMath system, developed in the University of Saarland and the DFKI.

I've been working in this project for one year, but I've been trying to convince them to release it as GPL for more than two years. Consistently and repeatedly, the answer has been that they wanted to do so, but "the GPL is incompatible with German law". The lawyer they consulted, some Professor at the Uni whose name I don't remember now, also wrote a lengthy analysis which I couldn't read since it was in German. But everyone here insists on that indeed the GPL is not valid in Germany.

I planned to raise this issue in the round table Suggestions for encouraging the creation of Free software for researchers at the end of the Libre software and research track in the Libre Software Meeting in Metz, so if other people want to discuss it there, I'd be very interested.

Below I reproduce the content of the email I sent to activemath-dev on Tue, 30 Jul 2002 10:49:41 +0200, with what I know about this issue:

>[copyleft doesn't agree with German law]

Do you have some exact quotation? Like an email or something?
Many GPL programs are produced in Germany.

Seems like the only clash is the "no warranty" clause (points 11 and 12), that would be illegal in Germany ("Gesetz zur Regelung des Rechts der Allgemeinen Geschäftsbedingungen"). (This doesn't mean that you become a criminal by releasing your software with such a clause. It's only that said clause does not apply.)
This just means that we can't reject responsability, just like any other software publisher in Germany.
AFAIK, the rest of the GPL is perfectly applicable in Germany:
http://mailman.fsfeurope.org/pipermail/discussion/ 2000-November/000043.html

However, it seems to be the case that, when distributing something free of charge and with the source available for inspection, that responsability is greatly reduced:
http://www.ifross.de/ifross_html/art3.html

More information can be found in the other pages at the "Institut für Rechtsfragen der freien und open source Software":
http://www.ifross.de

To avoid the dangers of not having the "no warranty" clause, some people use some "Allgemeine Geschäftsbedingungen" (AGB), which put some mechanisms in place to shield the software publisher from sueing. There is a peculiarity in German law that allows anyone to ask for a fee to a publisher when warning them of some infraction, such as trademark infringement. Thus one of the points of these AGB is to allow downloading only to people that abide to them by means of a password-protected download area, and then requiring for access an agreement that the person obtaining the software renounces to use the files to look inside for infringements.
However, this restrictions cause conflict with the GPL, as they are certainly restrictions on the freedom of the users, and can easily make the software non-free:
http://mailman.fsfeurope.org/pipermail/discussion/ 2001-May/0010

This was my final year project thesis. Just remember the golden rule unstructured 2 structured == convert 2 XML I wrote a [very bad] program in C++/Perl/tcsh IPC=pipes to add XML tags to English, and then index them into a search engine which would use the lingual data stored in the XML tags to help the search.

NIST does a MASSIVE competition on this annually. I don't want to be an XML-buzzword whore <Arnold Schwarzenegger accent> (XML commando eats Green berets, C++, Java, Perl, COBOL for breakfast)</Arnold Schwarzenegger accent> but you can't beat XML for easily converting anything that you can make sense out of into computer readable format. Real h3cKoRs use SGML, but us underlings have to stick with things we can understand like XML. As for expandability, if we want to encode something else into the document, then just tag-it-and-go

It took me 200 hours to fish out all these links (before the Google days), I don't want anyone to have to waste as much time as I did feeding the search engines exotic foods. It's a year old so pardon me for the odd broken link, armed with these you could probably turn jello into XML ;-)

My favourite bookmarx
PROJect[21 links]
Beginners' Guide[13 links]
Berkeley Linguistics Dept. Course Summaries, general stuffzzzzzzzzzzzzzzCryptic IR Vocabulary defined
Explanations of weird words like hypernym zzzzzzzzzzzzzzHow do we produce and understand speech
How Inverted Files are Created - Univeristy of Berkeley zzzzzzzzzzzzzzNLP Univ. of Indiana, very good basics e.g. word sense d
Simple langauge - useful.... zzzzzzzzzzzzzzWhat is Natural Language Processing, links
What is POS tagging........ zzzzzzzzzzzzzzWord Sense Disambiguation defined
Word Sense Disambiguation in detail, scroll down far zzzzzzzzzzzzzzWord Sense Disambiguator - LOLITA (tested at MUC-7 and SENSEVAL competition as best)
XML for the absolute beginner

HTML, XML stuff + parsers[19 links]
Apache plug-in that uhhh does stuff with XML zzzzzzzzzzzzzzConvert COM to XML
convert XML, HTML to Unix pipeable formats zzzzzzzzzzzzzzconverters to and from HTML
expat XML parser zzzzzzzzzzzzzzHTML Tidy - converts HTML 2 XML + source code!!
Parse DB (RDBMS, whatever) to XML zzzzzzzzzzzzzzPerl-XML Module List
PHP Manual XML parser functions - what the hell are they talking about, PHP Virtual M... zzzzzzzzzzzzzzPublic SGML-XML Software
Pyxie - XML Processor for Python, Perl, etc. zzzzzzzzzzzzzzSGML+XML tools.org
The XML Resource Centre - massive number of links zzzzzzzzzzzzzzW4F wrapper - wrapper converts XML to HTML
XFlat - convert flat file into XML zzzzzzzzzzzzzzXML Parsers and other XML stuff
XML.com - Parsers, etc. zzzzzzzzzzzzzzXML-Data Catalog System - uhhhh looks close
XTAL's general converter - convert anything 2 XML

other Background[8 links]
Is Linux ready for the Enterprise, scalable... zzzzzzzzzzzzzzLinux reliability
Linux Versus Windows NT, Mark(sysinternals bloke) zzzzzzzzzzzzzzPC reliability (pcworld)
SPEC - Standard Performance Evaluation Corp. zzzzzzzzzzzzzzSystems benchmarks
TPC - Transaction Processing Performance Council zzzzzzzzzzzzzzUnix Beats Back NT In EDA Workstation Arena
Proper TREC(-8) QA systems[2 links]

pg. 387 LIMSI-CNRS pretty deep parsing[2 links]
More links....
NLP, IR links - lots to corpii, etc.

pg. 575 U. of Ottawa and NRL (shit system, got 0%)[1 links]
LAKE Lab
pg. 607! University of Sheffield (crap system, but OPEN SOURCE!)[2 links]
GATE - FREE IE app w`source code
LaSIE - ER, coreference, template (cv)

pg. 617 Univ of Surrey (inconclusive matches)[2 links]
System Quirk - Or is this their search system..... Hmmmmmm
Univ of Surrey - pointers (hopefully this is their WILDER search system...)

SMU - Pg. 65[1 links]
Natural Language Processing Laboratory at SMU

Textract[2 links]
Cymfony - Technology
Textract - State of the Art Information Extraction

Xerox uhhhhh maybe[1 links]
Xerox Palo Alto Research Center
(OVERVIEW) 1999 TREC-8 Q&A Track Home Page
NLP bloke, Univ Sussex


Tcl-Tk[4 links] Tcl tutorial
Tcl-Tk Contributed Programs Index
Tcl-Tk Resources, sources
TclXML - manipulating XML using Tcl-Tk
Artificial Natural Language - Is this what I'm trying to parse into...
Comparison of Indexers - Prise vs. Inquery vs. MG, etc.
Eagles - Language Engineering Standards
Language Technology Group - lots of modules!
LDC - Linguistic Data Consortium, lots of corpora
Lexical Resources
Links 2 resources, indexers.....
Lots of IR stuff, University of uhhh
Managing Gigabytes Indexer
Managing Gigabytes Manuals and stuff
Htdig search system
NLP & IR (NLPIR, NIST) Group
OVERVIEW OF MUC-7-MET-2
Perl XML Indexing - XML search engine type thing
Phrasys Language Processing Software Components (money)
QA HCI bullshit
SIGIR - TREC-type thing, resources
SMART indexer system documentation
Text REtrieval Conference (TREC) Home Page
The Natural Language Software Registry
Thunderstone IE and IR products
WordNet - FREE DOWNLOADABLE lexical English database

Page created with URL+, nice utility for working with internet shortcuts

My dissertation research (see my web site) is hopefully a small stepping-stone in this direction. It involves building a software system that can acquire a basic lexicon based on visual experiences.

This, of course, does not touch on such ideas as emotion and motivation that would be crucial to the popular concepts of A.I., but in my mind, the grounding must come first. Note that some interesting work on emotion and motivation has been done by Steve Allen.

Having said that, top-down A.I. has made some incredible accomplishments considering that it has only been around for a short while (relatively speaking). However to paraphrase Steve Grand in his book Creation, Life and How to Make It, just because something exhibits intelligent behavior does not make it intelligent.

• Either you do it right (using formal logic calculus), and its extremly slow (exponential...).
• Or you do it informally, and it doesn't give you the right results.

The bottom line of this kind of technology is that although the speech recognition itself is relatively poor it is helped by the fact that most of the interesting words (names of people, places, etc.) occur very often in the same segment. So, it's all statistics. No accurate transcription needs to be made to achieve this kind of result. Therefore, applications such as automatic sub-titling are not possible with such systems. And I think they are still quite far away too.

As for the CIA claiming break throughs, well, I think other people can say wittier things about that.

Theo