Peculiarities of the development of the dictionary for the MT system from Azerbaijani

1Peculiarities of the Development of the Dictionary for the MT System from Azerbaijani Rauf Fatullayev1, Ali Abbasov2, Abulfat Fatullayev2 1 National E-Governance Project, Z.Aliyeva str., AZ1000 Baku, Azerbaijan 2 National Academy of Sciences, Huseyn Cavid str. 17, AZ1143 Baku, Azerbaijan fatullayev@gmail.com, ali@dcacs.ab.az, fabo@mail.ru Abstract. Azerbaijani (The Azerbaijani language) is one of the languages of Turkic group which are morphologically rich languages. While creating the machine translation (MT) system from Azerbaijani, it is not possible to create an MT dictionary consisted of all word-forms of Azerbaijani, because Azer- baijani is an agglutinative language and it is possible to generate practically “endless” number of word-forms by using the various suffix chains (compound suffixes consisted of simple suffixes). For the solution of this problem the approach which considers keeping the word stems and the suffix chains separately is offered. These databases compound together the dictionary of the MT system from Azerbaijani and yield possibility to translate any word-form. Keywords: Azerbaijani, machine translation system, MT dictionaries, automation of the translation process, Turkic languages 1 Introduction Azerbaijani is one of the languages of Turkic group and these languages form one of the largest language groups (modern Turkish, Azerbaijani, Kazakh, Uzbek, Turkmen, Tatar, Kirghiz and others). Despite some theoretical research, the most of these languages are still less investigated languages (except modern Turkish [1-2]). Turkic languages are the morphologically rich languages and these languages are characterized by highly productive morphology that may produce a very large number of word forms for a given stem [3-5]. If we take into account only verb stems (≈ 8000) in Azerbaijani [6] and minimal number of active (frequently used) suffix chains which can be joined only the verb stems (≈ 500), we would keep 4000000 verb forms in the dictionary of the MT system. If we take into account word-forms which can be formed from the word stems (≈70000 lexical units, including terms [7]) of the other parts of speech the size of the dictionary will be many times more than this number. So, modeling each word-form as a separate lexical unit leads to a number of problems while developing the formal linguistic technologies (machine translation, speech This research is carried out within the joint project “Development of the Azerbaijani-English MT system” of The Ministry of ICT of Azerbaijan and UNDP-Azerbaijan 12th EAMT conference, 22-23 September 2008, Hamburg, Germany 35 recognition, text to speech etc. systems). Including all word-forms to the dictionary of the MT system can not be considered as a rational way. In this paper we show the way to solve the problem concerning the development of the dictionary of the MT system from Azerbaijani. For this purpose, we create the dictionary of the word stems of Azerbaijani, the database of the suffix chains and the database of the translation rules of the suffix chains. Using these databases we can construct the translation of any Azerbaijani word-form. These three databases compound the dictionary of the MT system from Azerbaijani (This problem is characteristic for the MT dictionaries from …

Content truncated to 500 words