Research On Text Resource Retrieval Technology For The Cyrillic Mongolian Language Distance Learning System

It is very difficult to spread education in Mongolia with a vast land and sparse population.Nowadays,web-based distance learning system is an effective way to spread education and deal with problems experienced in attaining higher education.There are various distance learning systems developed in universities and other organizations in Mongolia.However,most of them are static learning systems,characterized as passive learning.In order to improve the results of these learning systems we need to develop a dynamic distance learning system.We have developed a dynamic distance learning system based on electronic test(e-test)results.The system's main functions are using e-test results to automatically determine student's knowledge and automatically extract optimal content that matches to students' level of knowledge from learning text resources.The learning resources in Mongolia were written in the Mongolian Cyrillic script.We need to study technology for information retrieval written in the Mongolian Cyrillic script.A national search system developing work has included the action plans 2016 for the Information Technology,Post and Telecommunications Authority in Mongolia.However,there are very few studies conducted about information search technology.Only B.Khaltar,O.Chimeddorj and Atsushi Fujii have conducted related research works.For example,in B.Khaltar's study,word stem were extracted and used for searching information.O.Chimeddorj improved on B.Khaltar's method and now the new method is used for Mongolian and English statistic machine translation.In this research work,first we determined the current situation of information retrieval technology in other languages and then we studied information retrieval in dynamic distance learning systems using the Cyrillic Mongolian Script.The following items were the main novelty of this research work:1.To improve the result of text resource retrieval in distance learning system using Mongolian Cyrillic script,we studied the word structure and features of Mongolian Cyrillic script and developed a method for extracting word stem based on the orthography of Mongolian Cyrillic script.We created a database with 41000-word stems,168-word suffixes,and 935 rule of Mongolian Cyrillic script.The word stems were extracted by parsing word suffixes.We tested 560 law documents with 1.780.968 words,and 75 university learning text materials with 178.448 words.The accuracy of the word stem extraction results was 92.6 percent.Therefore we confirmed that the suggested method was effective.2.To improve the results of text resource retrieval at distance learning system in Mongolian Cyrillic script,we developed methods for defining index entry.In this method used in the orthography of Mongolian Cyrillic script and calculating for TF-IDF and expression with co-occurrence.The result of this comparative study was that we found that expression with the extracting word stem method was more effective.Also,we tested the data of 1450 law documents and 250 learning text materials of universities.The co-occurrence method had 78%accuracy while the TF-IDF method had 59%accuracy;the extracting word stem method had 88%accuracy.The co-occurrence method had 85%recall while the TF-IDF method had 67%recall,and the extracting word stem method had 87%.Therefore we confirmed that the suggested method was effective.3.To evaluate for performing text resource retrieval extracting system based on word stem and keywords,we tried using the Vector Space Model.For data of 250 learning text materials of universities,cosine average of two methods was 77%and 85%respectively.To confirm the above experiment for data of 2560 learning text materials of universities,MAP of two methods was 75%(k=100),79%(k=40)and 100%(k=1)respectively.Therefore we confirmed that extracting methods based on word stem and keyword was effective and suitable.