ALL ACTIVITIES OF THE LAB HAVE BEEN TRANSFERED TO WWW.SLOVNIK.BG

We are trying to make popular the result of a long standing effort in the area of Bulgarian language morphology. Our understanding is that such knowledge should not be a property of a single company, laboratory or group of people.
 

The Product Availability  Coming soon The Team The Story

Important note: We are not the only or even the best laboratory working in this direction. So, we have tried to make an index of the other groups working on similar problems. They can be found here.

 

The Product

In fact, it is not a single product. We prefer to think about what we do as a lexical knowledge base, a huge amount of information about the morphology of the Bulgarian language. It can be incorporated within the appropriate software interfaces for each specific task.

So, lets first say what is being kept in the knowledge base:

• more than 65 000 base word forms or lexemes;
• all forms of each word. Totally, they are more than one million.
• syntactic characteristics of each word and each of its forms: part of speech, gender, etc.

• search of a word by any of its forms;
• analysis of word form characteristics and retrieval of the corresponding base form. If there is ambiguity, the system lists all the possibilities;
• generation of word forms on the basis of a specified base form and grammatical characteristics.

Currently we provide a relatively small number of additional services over text files:

• text analysis - part of speech tagging without disambiguation;
• word frequency - frequency analysis
• unknown words - extraction of the words which can not be found in the LKB.

Slovnik was developed with Borland C++ 5.02, making use of OWL NExt. The basic services (the morphology engine) are written in ANSI C++ in order to allow easy portability.

Availability

The Slovnik application is available via e-mail for the price of 80 USD. There are special discount for non-commercial licenses - 30 USD for academic organizations and 40 USD for personal users. Please, contact us on diogen.at.diogenes.bg for details about payment and discounts. A fully functional demo version is available without any restrictions or conditions. It works with a restricted lexicon (about 2 500 base forms) and can be downloaded directly.

Please, contact us for any special services and processing which can be developed under additional agreement.
 

Coming Soon

Our team continuously develops and improves both the lexical knowledge base and the software. Here are listed only the most significant and immediate results that can be expected.

• Cross-platform morphological server

We are very close to the release of a morphological engine that can interoperate with other programs via TCP/IP. Being written in ANSI C++ it can run on any platform. Currently it is tested on top of a number of versions of Linux and Windows.

The basic service that the engine provides is to give the list of all forms of the word, requested by any form. It is a multithread piece of software, that allows it to work in a smooth and reliable fashion when accessed from multiple users at the same time.
 

• Russian language version of the lexical knowledge base

While all the tools and the methodology are language independent, it runs with exactly the same software - you can choose what language to use, at run-time.

Now we have about 20 000 base forms in Russian. As a base we are using Zaliznjak's Grammatical dictionary of Russian Language which contains about 100 000 words. We expect to have the same coverage when we complete the development of the Russian lexical knowledge base.

Everything here became possible, due to the joined expertise and hard work of four men:

 

Dimitar Popov - lexicographer, Institute of Bulgarian Language;

• Kiril Simov - computer scientist, Linguistic Modeling Laboratory;
• Ognjan Chernokojev - professional programmer;
• Atanas Kiryakov - the always searching man Diogenes, who gathered the team.

What we present here is rooted in the work on the Bulgarian Language Pronunciation, Spelling and Punctuation Dictionary, which will be called simply "The Dictionary". Here you can learn more about this book, see it and even buy it.

The Dictionary is a special one. It was compiled with the help of a lexical knowledge base (LKB) consisting of a huge amount of formal, strict and precise declarative, logical statements about the natural language. Actually, the main section of the dictionary, the lexicon (in Bulgarian, "Slovnik"), was generated automatically from the LKB.

This approach gives an obvious twofold advantage:

 
• First, the dictionary is free from many kinds of mistakes, which come naturally in the "hand made" dictionaries;
• Second, as a "side product", we have a full, formal KB of Bulgarian language morphology.

The "real work" was done by Mr. Popov in cooperation with Mrs. Vidinska - they "encoded" and verified the Bulgarian morphology into the LKB, using the software shell developed by Kiril Simov.

The above is the history of our linguistic knowledge base: the pure knowledge, the scientific result, the abstract model of the language. In its natural form, it is as useful for real-world NLP applications, as the Relativity theory is useful for the Space Shuttle project :-)

Lets start talking about the software tools. When we started searching for a way to use the LKB for "industrial" purposes, we found ourselves facing whole new world, with its own rules and regulations. Here we list only some of the major problems we had to resolve in order to make possible the development of a set of software tools. We currently have:

 
• compact representation of the knowledge - in its full flat form, the LKB takes more than 20MB, while the alternative "compiled" representation takes about 1MB!
• efficient manipulation - the searching algorithm can find an arbitrary word form with all its characteristics without visible delay even on a 386 PC.
• usability - the tools presented are written in C++, taking all the advantages of the object-oriented programming and compiled 32-bit applications. Additionally, all the power of GUI is being used when appropriate.