StackOverflow

Do you know the famous resource http://stackoverflow.com by Joel Spolsky and team? Here is my own definition of stackoverflow in Java:

private static void log(String logStatement) {
   log(logStatement);
}

XML in Chrome

Tired of (manual) XML rendering in Google Chrome?
Try this:
XML Tree

NoSQL, Apache SOLR and Apache Hadoop

Check out this SlideShare Presentation -- talk for NёrdCamp April 23, 2011:

NoSQL, Apache SOLR and Apache Hadoop

View more presentations from Dmitry Kan.

NerdCamp

NerdCamp is coming and I'm going to present on Apache technologies, specifically on Apache Solr and a bit on Apache Hadoop. If you are into all this and like creative thinking, want to create a start-up and will come around Saint-Petersburg this weekend, you should definitely come to NerdCamp!

Confirmed Key Participants

Yury Lifshits, Sergey Poduzov, Alexander Shtuchkin,
Vladimir Gorovoy, Nikolay Vyahhi, Vladimir Aluferov, Yakov Sirotkin and myself

The preliminary program has juicy topics among which are:
How the Web will Transform Education by Yury Lifshits,
Geo Information Systems Around Us by Aleksander Klechikov,
Software Development for little ones by Yakov Sirotkin,
Introducing NoSQL: Apache SOLR and Hadoop by Dmitry Kan,
Introducing Cloud Services by Dmitry Petrov,
Nature of Enterpreneurship Craft, Lifestyle or Science by Alexey Baranov,
How to grow places for smart people -- workshop

and more!

Angry Birds -pelaaja: DK

Angry Birds -pelaaja: DK: "DK on kertakaikkisen kyllästynyt sikailuun! Seuraa hänen edesottamuksiaan Angry Birdsin SM-kisoissa ja tule itsekin mukaan."

concert of components?

"ZooKeeper: Because coordinating distributed systems is a Zoo"

This post is short and not useful in that it doesn't give you any code snippets or technical recommendations. I would like just to cite the SolrCloud's wiki. The SolrCloud is "the set of Solr features that take Solr's distributed search to the next level, enabling and simplifying the creation and use of Solr clusters." They use Apache ZooKeeper project (a subproject of Hadoop) as a distributed system of keeping cluster of SOLRs state updates. In a distributed system every component can potentially crash, yet the system is expected to provide it's service to its users. If a single SOLR crashes, its replica will take over, but if ZooKeeper crashes, the system will still continue serving the user requests, but no updates of the system are visible to the system (sounds interesting, I know). To improve that, this is what is possible:

"Running multiple zookeeper servers in concert (a zookeeper ensemble) allows for high availability of the zookeeper service. Every zookeeper server needs to know about every other zookeeper server in the ensemble, and a majority of servers are needed to provide service. For example, a zookeeper ensemble of 3 servers allows any one to fail with the remaining 2 constituting a majority to continue providing service. 5 zookeeper servers are needed to allow for the failure of up to 2 servers at a time."

So, if you have a big zoo with variety of animals in it, make sure you have 5 zoo keepers for at least 3 of them take care of your pigs and elephants, when 2 others got stuck somewhere else.

SOLR: speed up batch posting

If you are familiar with the Apache SOLR and deal with index worth of millions documents, it becomes quite important to be able to (re-)index fast. There are various techniques as to what to tweak for the indexing go faster. Another simple way to speed up your (batch) indexing without changing your SOLR schema is to modify logging.

Apparently, when deployed under Tomcat, SOLR logs each and every update request during POSTing process. Experience shows, that heavy http operation is done faster, when logging is minimal.

SOLR (as of 1.4 at least) has admin GUI which serves as a central information hub for the given SOLR core. Among other useful features, it has a page where one can set up logging levels of different SOLR components. In default SOLR installation you can access the page via http://localhost:8983/solr/admin/logging. By default, the logging levels amount mainly to INFO, which permits logging of all the select/update requests (imagine 1 million of such log entries for batch reindexing).

It would be handy to be able to automatically change the logging levels to, say, WARNING before batch POSTing and back to INFO after that. solr/admin/logging is declared as servlet in web.xml of the corresponding SOLR core:

  
    Logging
    org.apache.solr.servlet.LogLevelSelection
  

All the components which allow changing their logging levels are listed on the page http://localhost:8983/solr/admin/logging. Using curl we can send a post request to the servlet and set the desired levels. It is reasonable to implement a function, which takes logging level and url of SOLR core as parameters (choose your own favourite language, this is done in Perl):

sub setSolrLogLevel
{
 my ($url, $level) = @_;

    print "setting logging level to $level\n";
    my $res = system("curl --user user:pass -d \"submit=set&root=$level&fi=$level" .
"&fi.alphasense=$level&fi.alphasense.solr=$level&fi.alphasense.solr.query=$level" .
"&fi.alphasense.solr.query.AlphaSenseQParserPlugin=$level&httpclient=$level&httpclient.wire=$level&httpclient.wire.content=$level&httpclient.wire.header=$level&javax=$level&javax.management=$level&javax.management.mbeanserver=$level&org=$level&org.apache=$level&org.apache.catalina=$level&org.apache.catalina.core=$level&org.apache.catalina.core.ContainerBase=$level&org.apache.catalina.core.ContainerBase.%5BCatalina%5D=$level&org.apache.catalina.core.ContainerBase.%5BCatalina%5D.%5Blocalhost%5D=$level&org.apache.catalina.core.ContainerBase.%5BCatalina%5D.%5Blocalhost%5D.%5B%2Fsolrtopic%5D=$level&org.apache.catalina.core.ContainerBase.%5BCatalina%5D.%5Blocalhost%5D.%5B%2Fsolrtopic%5D.%5BLogging%5D=$level&org.apache.catalina.core.ContainerBase.%5BCatalina%5D.%5Blocalhost%5D.%5B%2Fsolrtopic%5D.%5BSolrServer%5D=$level&org.apache.catalina.core.ContainerBase.%5BCatalina%5D.%5Blocalhost%5D.%5B%2Fsolrtopic%5D.%5BSolrUpdate%5D=$level&org.apache.catalina.core.ContainerBase.%5BCatalina%5D.%5Blocalhost%5D.%5B%2Fsolrtopic%5D.%5Bdefault%5D=$level&org.apache.catalina.core.ContainerBase.%5BCatalina%5D.%5Blocalhost%5D.%5B%2Fsolrtopic%5D.%5Bjsp%5D=$level&org.apache.catalina.core.ContainerBase.%5BCatalina%5D.%5Blocalhost%5D.%5B%2Fsolrtopic%5D.%5Bping%5D=$level&org.apache.catalina.session=$level&org.apache.catalina.session.ManagerBase=$level&org.apache.commons=$level&org.apache.commons.digester=$level&org.apache.commons.digester.Digester=$level&org.apache.commons.digester.Digester.sax=$level&org.apache.commons.httpclient=$level&org.apache.commons.httpclient.ChunkedInputStream=$level&org.apache.commons.httpclient.HeaderElement=$level&org.apache.commons.httpclient.HttpClient=$level&org.apache.commons.httpclient.HttpConnection=$level&org.apache.commons.httpclient.HttpMethodBase=$level&org.apache.commons.httpclient.HttpMethodDirector=$level&org.apache.commons.httpclient.HttpParser=$level&org.apache.commons.httpclient.HttpState=$level&org.apache.commons.httpclient.MultiThreadedHttpConnectionManager=$level&org.apache.commons.httpclient.SimpleHttpConnectionManager=$level&org.apache.commons.httpclient.auth=$level&org.apache.commons.httpclient.auth.AuthChallengeProcessor=$level&org.apache.commons.httpclient.cookie=$level&org.apache.commons.httpclient.cookie.CookiePolicy=$level&org.apache.commons.httpclient.cookie.CookieSpec=$level&org.apache.commons.httpclient.methods=$level&org.apache.commons.httpclient.methods.EntityEnclosingMethod=$level&org.apache.commons.httpclient.methods.ExpectContinueMethod=$level&org.apache.commons.httpclient.methods.PostMethod=$level&org.apache.commons.httpclient.params=$level&org.apache.commons.httpclient.params.DefaultHttpParams=$level&org.apache.commons.httpclient.params.HttpMethodParams=$level&org.apache.commons.httpclient.util=$level&org.apache.commons.httpclient.util.EncodingUtil=$level&org.apache.commons.httpclient.util.ExceptionUtil=$level&org.apache.commons.httpclient.util.IdleConnectionHandler=$level&org.apache.jasper=$level&org.apache.jasper.EmbeddedServletOptions=$level&org.apache.jasper.JspCompilationContext=$level&org.apache.jasper.compiler=$level&org.apache.jasper.compiler.Compiler=$level&org.apache.jasper.compiler.JspConfig=$level&org.apache.jasper.compiler.JspRuntimeContext=$level&org.apache.jasper.compiler.TldLocationsCache=$level&org.apache.jasper.servlet=$level&org.apache.jasper.servlet.JspServlet=$level&org.apache.jasper.servlet.JspServletWrapper=$level&org.apache.solr=$level&org.apache.solr.analysis=$level&org.apache.solr.analysis.BaseTokenFilterFactory=$level&org.apache.solr.analysis.BaseTokenizerFactory=$level&org.apache.solr.client=$level&org.apache.solr.client.solrj=$level&org.apache.solr.client.solrj.impl=$level&org.apache.solr.client.solrj.impl.CommonsHttpSolrServer=$level&org.apache.solr.common=$level&org.apache.solr.common.util=$level&org.apache.solr.common.util.ConcurrentLRUCache=$level&org.apache.solr.core=$level&org.apache.solr.core.Config=$level&org.apache.solr.core.CoreContainer=$level&org.apache.solr.core.JmxMonitoredMap=$level&org.apache.solr.core.RequestHandlers=$level&org.apache.solr.core.SolrConfig=$level&org.apache.solr.core.SolrCore=$level&org.apache.solr.core.SolrResourceLoader=$level&org.apache.solr.handler=$level&org.apache.solr.handler.AnalysisRequestHandler=$level&org.apache.solr.handler.XmlUpdateRequestHandler=$level&org.apache.solr.handler.admin=$level&org.apache.solr.handler.admin.LukeRequestHandler=$level&org.apache.solr.handler.admin.SystemInfoHandler=$level&org.apache.solr.handler.component=$level&org.apache.solr.handler.component.QueryElevationComponent=$level&org.apache.solr.handler.component.SearchHandler=$level&org.apache.solr.handler.component.SpellCheckComponent=$level&org.apache.solr.highlight=$level&org.apache.solr.highlight.SolrHighlighter=$level&org.apache.solr.request=$level&org.apache.solr.request.BinaryResponseWriter=$level&org.apache.solr.request.XSLTResponseWriter=$level&org.apache.solr.schema=$level&org.apache.solr.schema.FieldType=$level&org.apache.solr.schema.IndexSchema=$level&org.apache.solr.search=$level&org.apache.solr.search.SolrIndexSearcher=$level&org.apache.solr.servlet=$level&org.apache.solr.servlet.LogLevelSelection=$level&org.apache.solr.servlet.SolrDispatchFilter=$level&org.apache.solr.servlet.SolrRequestParsers=$level&org.apache.solr.servlet.SolrServlet=$level&org.apache.solr.servlet.SolrUpdateServlet=$level&org.apache.solr.spelling=$level&org.apache.solr.spelling.AbstractLuceneSpellChecker=$level&org.apache.solr.spelling.FileBasedSpellChecker=$level&org.apache.solr.spelling.IndexBasedSpellChecker=$level&org.apache.solr.update=$level&org.apache.solr.update.SolrIndexConfig=$level&org.apache.solr.update.UpdateHandler=$level&org.apache.solr.util=$level&org.apache.solr.util.SolrPluginUtils=$level&org.apache.solr.util.plugin=$level&org.apache.solr.util.plugin.AbstractPluginLoader=$level\" $url");
    print "Result code:$res\n";
}

There you go. Call setSolrLogLevel("http://localhost:8983/solr/admin/logging", "WARNING"); before the batch POSTing and setSolrLogLevel("http://localhost:8983/solr/admin/logging", "INFO"); after the batch POSTing has finished.

"OOPness" in Java

Optimisation in my code has led me to a code refactoring, where several methods with return values became one. Since I could not use "pass by reference" feature for all my parameters (primitives and objects as collections) I had to come up with an inner class for holding all return parameters.

So by not supporting "pass by reference" feature Java makes your code even more OOP-ish. Even though the topic can be well known, this article helps to put things together and make a memory refresh.

Successive replacement in regular expressions (java)

Actually not sure, how often people out there do the successive replacement in a target text having a regular expression pattern, but Java has rather neat solution for it. I'm publishing it here, because I know, that esp. younger developers can re-invent a wheel here and have longer debugging sessions.

So the task is like this: you have a text T, like "cat-1 dog-1 cat-1 elephant-1 cat-2 dog-2 cat-3".
Suppose we want to change numerals attached to the words "cat" to their word representations: "1" to "one", "2" to "two".

One straightforward way would be to match all "cat-([0-9])+" subsequences and then run replace operation on T.

So the code would look something like this:

String T = "cat-1 dog-1 cat-1 elephant-1 cat-2 dog-2 cat-3";
Pattern catPattern = Pattern.compile("cat-([0-9]+)");
Matcher catMatcher = catPattern.matcher(T);
Map numToWord = new HashMap();
numToWord.add("1", "one");
numToWord.add("2", "two");
numToWord.add("3", "three"); // ...
while (catMatcher.find())
{
    T = T.replaceFirst(catMatcher.group(1), numToWord.get(catMatcher.group(1)));
}

This code produces:

cat-one dog-one cat-1 elephant-1 cat-two dog-2 cat-three

Which is missing one substitution. Ok, let's use replaceAll instead and make sure we touch only cats:

{
    T = T.replaceAll("cat-" + catMatcher.group(1), "cat-" + numToWord.get(catMatcher.group(1)));
}

which produces what we want:

cat-one dog-1 cat-one elephant-1 cat-two dog-2 cat-three

But now what happens inside the loop is logically out of sync with the loop condition: we iterate over matches, but call replaceAll (probably not efficient either, as replaceAll will be attempted even when not needed anymore, for duplicate matches).

Any more elegant and correct solution?

Yes! It is called Matcher.appendReplacement

  Pattern catPattern = Pattern.compile("cat-([0-9]+)");
  Matcher catMatcher = catPattern.matcher(T);
  Map numToWord = new HashMap();
  numToWord.put("1", "one");
  numToWord.put("2", "two");
  numToWord.put("3", "three"); // ...
  
  StringBuffer sb = new StringBuffer();
  
  while (catMatcher.find())
  {
   System.out.println("Match:" + catMatcher.group(1));
   catMatcher.appendReplacement(sb, "cat-" + numToWord.get(catMatcher.group(1)));
  }
  catMatcher.appendTail(sb);

now sb.toString() contains:

cat-one dog-1 cat-one elephant-1 cat-two dog-2 cat-three

If you append System.out.println(sb.toString()); inside the while loop, you will also see, that replacements happen in sync with the while loop's state, so that what is inside the loop and what while loops over are in sync.

B2B: what is WSDL (simple explanation)

Just for those wondering around what is WSDL file or technology in the area of Business 2 Business (when companies speak to each other automatically over the network): you can think of it as a declaration of the client-server conversation. In this case one B (server) provides some functionality (like taking square roots of big numbers) and another B (client) has a lot of these numbers and needs square roots of them.

So the server declares a method (which is in fact a remote method):

Double squareRoot(Double number)

in its WSDL file. The client side then takes this file and generates the client side code, which handles the protocol of communication and remote method invocation automatically. What client needs is simply to implement a business logic around this invocation, like a web page with a text field for the number or the entire hardware with touch screen display.

In general the input and return types of remote methods can be simple and complex -- in fact as complex and "proprietary" as you need, because the entire WSDL file is nothing but xml as well as the messages being sent over the network.

One of the libraries used in the industry is Axis and Axis2. Using the library you can generate the java code from wsdl for you client (and even generate wsdl from your java code for you server).

Zero-witdth negative lookahead group: example

Suppose you have to split a comma separated string in java where meaningful parts are separated by continious spaces, say from 2 to infinity. But the problem is that sometimes a meaningfull part (like a person name) contains 2 spaces. You want to exclude such cases and retain the name's parts together. Example:

name1[one_space]surname1[two_spaces]age
surname2[two_spaces]surname2[two_spaces]age

In this example it is easy to solve the problem with zero-width negative lookahead group (?!\\w), because age is always supposed to be numeric. So the final code will be something like this:

String s = current_string.split("\\s{2}(?!\\w)");

Парижские впечатления

Вернулся из второй поездки в Париж. Вот сейчас, уже по возвращению, я подумал: в Париже все отдыхают. Или это только видимость? Как бы там ни было, отдыхать там легко. Понравилось кататься на public велосипедах!

Dumper and sorting of keys

When you use Dumper for debugging via comfortable logging data structures by reference, you can additionally apply sorting to the output. Say, you have a hash addressed by $hash_ref reference. If you need to output the hash contents, but have its keys sorted, you can do:

use strict;
use Data::Dumper;

$hash_ref = generate your hash here

$Data::Dumper::Sortkeys = \&my_filter;
print Dumper($hash_ref), "\n";

sub my_filter {
 my ($hash) = @_;
 # return an array ref containing the hash keys to dump
 # in the order that you want them to be dumped
        return [
            sort {$a <=> $b} keys %$hash
        ];
}

Сходимость

Код программы аппроксимирует решение задачи. Сходимость кода к решению осуществляется посредством связки "изменение кода"-"компиляция/интерпретация"-"запуск". Получается что-то вроде временного ряда: код может не сходиться неделю и сойтись за один день. Если код не сходится, стоит посмотреть на саму концепцию в решении / задачу / компилятор.

Всё это к тому, что полезно преподавать программирование, рассматривающее программу как целостную математическую схему, а не только как набор алгоритмов, приёмов и дизайна.

Finnish is like C++

, because it is highly typed.

An idea for GUI designers and developers -- a standalone app / OS level change.

When I'm under VPN & putty I sometimes need to run an sftp client. Currently I have two options to use: WinSCP (I find it easier to use in some cases, like quickly watching / editing the contents of different text files) and Secure Shell Client from SSH. For some reason the first client loads in about 2-3 minutes after I log in. So: it would be great to have an option to configure launch of a program depending on a certain event, like putty execution, connected to a certain host.

This may give a lot of opportunities to quite flexibly configure your GUI world and save a lot of time spent otherwise on the monkey repetitions.

giza++ under windows: episode 2

It turned out, that in order to comfortably run GIZA++ under win32 with cygwin I had to recompile it without the flag -DBINARY_SEARCH_FOR_TTABLE (credits: http://code.google.com/p/giza-pp/issues/detail?id=9).

The full list of steps goes here (I suppose that source corpus is stored in corpus.ru and target corpus is stored in corpus.en):

1. produce vcb and dictionary files with "plain2snt.out corpus.ru corpus.en" (credits: http://vee-r.blogspot.com/2006/12/giza-guide.html)
2. produce cooc file with "snt2cooc.out corpus.ru.vcb corpus.en.vcb corpus.ru_corpus.en.snt > ru_en.cooc" (credits: myself, after analyzing the train-factored-phrase-model.perl from Moses package)
3. run GIZA++ with config file:

outputfileprefix play_giza
sourcevocabularyfile corpus.ru.vcb
targetvocabularyfile corpus.en.vcb
c corpus.ru_corpus.en.snt
CoocurrenceFile ru_en.cooc
model1dumpfrequency 1
model4smoothfactor 0.4
nsmooth 4
onlyaldumps 1
nodumps 1
p0 .999
m1 5
m2 0
m3 3
m4 3
o giza

if these options are stored in giza.config then run "GIZA++ giza.config". This produces giza.A3.final file, typical entry of which is:

# Sentence pair (1) source length 4 target length 7 alignment score : 2.25315e-10
there is a book on the table
NULL ({ }) на ({ }) столе ({ 5 6 7 }) лежит ({ 1 2 }) книга ({ 3 4 })

which means the following mapping:

"столе" --> "on the table"
"лежит" --> "there is"
"книга" --> "a book"

Running GIZA++ under win32 and under linux gives same results in terms of word mappings, except that the alignment scores may slightly differ due to possibly different float point precision models.

giza++ under windows

After an 'outrageous' attempt to compile giza++ under Visual Studio C++ Express 2008 (with more than 2000 compile time errors) I switched over to cygwin and installed:

gcc version 3.4.4 (cygming special, gdc 0.12, using dmd 0.125)
make GNU Make 3.81 (built for i686-pc-cygwin)

With

$ make

under giza-pp-v1.0.3\giza-pp I have obtained two executables: GIZA++-v2/GIZA++.exe and mkcls-v2/mkcls.exe which perfectly run under Windows XP Professional Version 2002 SP 2.

RuSSIR'2010

What is RuSSIR? It is the summer school on Information Retrieval. Where is it held? Usually in Russia, this time in Voronezh. Want to apply? Read on.

4th Russian Summer School in Information Retrieval (RuSSIR 2010)
Monday September 13 - Saturday September 18, 2010
Voronezh, Russia
http://romip.ru/russir2010/eng/

FIRST CALL FOR COURSE PROPOSALS

The 4th Russian Summer School in Information Retrieval (RuSSIR 2010) will be held on September 13-18, 2010 in Voronezh, Russia, one of the major cities in south-western Russia. The mission of the school is to teach students about modern problems and methods in Information Retrieval; to stimulate scientific research in the field of Information Retrieval; and to create an opportunity for informal contacts among scientists, students and industry professionals. The Russian Conference for Young Scientists in Information Retrieval will be co-located with the school. RuSSIR 2010 will offer 4 or 5 courses and host approximately 100 participants. The working languages of the school are English (preferable) and Russian. The target audience of RuSSIR is advanced graduate and PhD students, post-doctoral researchers, academic and industrial researchers, and developers.

The RuSSIR 2010 Organizing Committee invites proposals for courses on a wide
range of IR-related topics, including but not limited to:
- IR theory and models
- IR architectures
- Algorithms and data structures for IR
- Text IR
- Multimedia (including music, speech, image, video) IR
- Natural language techniques for IR tasks
- User interfaces for IR
- Web IR (including duplicate detection, hyperlink analysis, query logs)
- Text mining, information and fact extraction
- Mobile applications for IR
- Dynamic media IR (blogs, news, WIKIs)
- Social IR (collaborative filtering, tagging, recommender systems)
- IR evaluation.

Each course should consist of five 90-minute-long sessions (normally in five consecutive days). The course may include both lectures and practical exercises in computer labs.

RuSSIR 2010 organizers will cover travel expenses and accommodations at the school for one lecturer per course, but there is no additional honorarium. The RuSSIR organizers would highly appreciate if, whenever possible, lecturers could find alternative funding to cover travel and accommodation expenses and indicate this possibility in the proposal.

Course proposals for RuSSIR 2010 must be submitted by email to Pavel Braslavski (pb@yandex-team.ru), by February 14, 2010. A course proposal should contain a brief description of the course (up to 200 words), preferred schedule, prerequisites, equipment needs, a short description of teaching/research experience and contact information of the lecturer(s). All proposals will be evaluated by the RuSSIR 2010 program committee according to the school goals, presentation clarity, lecturer’s qualifications and experience. Topics not featured at previous RuSSIRs are preferred. All submitters will be notified by March 1, 2010. Early informal inquiries about the school or the proposal evaluation process are encouraged.

About RuSSIR: The Russian Summer School in Information Retrieval is co-organized by the Russian Information Retrieval Evaluation Seminar (ROMIP) and Voronezh State University. Previous schools took place in Ekaterinburg, Taganrog, and
Petrozavodsk. Previous RuSSIR courses included IR Models (by Djoerd Hiemstra), Modeling Web Searcher Behavior and Interactions (by Eugene Agichtein), Computational Advertising (by James Shanahan), Text Mining, Information and Fact Extraction (by Marie-Francine Moens), Natural Language Processing for Information Access (by Horacio Saggion), Music IR (by Andreas Rauber), and other. Ricardo Baeza-Yates, VP of Research for Europe and Latin America at Yahoo, has confirmed as an invited lecturer for RuSSIR 2010 with the course 'Web data mining'.

About the RuSSIR 2010 location: Voronezh is a major city in southwestern Russia, spanning both sides of the Voronezh River, with population of 850,000. Express trains from Moscow to Voronezh take about 10 hours. There are also regular flights from Moscow, Munich, Prague, Tel-Aviv, and Istanbul. The town was founded in 1586. In the 17th century, Voronezh gradually evolved into a sizeable town, especially after Tsar Peter the Great built a dockyard in Voronezh. Currently, Voronezh is an administrative, economic and cultural center of the Voronezh region. Voronezh surrounding area has many attractions including archeological museum, nature and historical reserve Divnogorie, Kostomarovo cave monastery, Orlov trotter stud farm at Khrenovoe. Voronezh has a large student population: 37 institutions of higher education and 53 colleges educating over 127,000 students today. Voronezh State University was founded in 1918 and is one of the largest universities in Russia, with a total enrollment of 22,000.

Contacts

Use the e-mail address and substitute [at] with @ and [dot] with "." school[at]romip[dot]ru.

Semantic Analysis: theory, applications and use cases

Gave a talk at 6th FRUCT seminar by Nokia Research Center and Helsinki University of Technology and Nokia Siemens Networks:

Semantic Analysis: theory, applications and use cases

View more documents from dmitrykan.