Browsing by Author "Alemu, Atelach (PhD)"
Now showing 1 - 2 of 2
Results Per Page
Sort Options
Item Development of Stemming Algorithm for Wolaytta Text(Addis Ababa University, 2003-07) Lessa, Lemma; Getachew, Mesfin (PhD); Alemu, Atelach (PhD); Eyesus, Haile (PhD)This study describes the design of a stemming algorithm for Wolaytta language. To give a solid background for the thesis, literature on conflation in general and stemming algorithms in particular were reviewed. Since it is the nature and characteristics of suffixation that guide the development of steamer, the Wolaytta language morphology was studied and described in order to model the language and develop an automatic procedure for conflation. The inflectional and derivational morphologies of the language are discussed. It is indicated that suffixation is the main word formation process in Wordplay language. It is also attempted to show that the language is morphological complex and uses extensive concatenation of suffixes The result of the study is a prototype context sensitive iterative stemmer for Wolaytta language. Error counting technique was employed to evaluate the performance of this stemmer. The stemmer was trained on 3537 words (80% of the sample text) and the improved version reveals an accuracy of 90.6% on the training set. The number of over stemmed and understeml11ed words on the training set were 8.6% (304 words) and 0.8% (28 words) respectively. When the stemmer rW1S on the unseen sample of 884 words (20% of the sample text), it performed with an accuracy of 86.9%. The percentage of endorser recorded as under stunned and over stemmed on this unseen (test set) were 9% and 4.1 %, respectively. Moreover, a dictionary reduction of 38 .92% was attained on the test set. The major sources of errors are also reported with possible recommendations to further improve the performance of the stemmer and also for further research.Item An Integra Ted Approach to Automatic Complex Sentence Parsing for Amharic Text(Addis Ababa University, 2003-06) Gochel, Daniel; Alemu, Atelach (PhD)Natural language processing is a research area which is becoming increasingly popular each day for both academic and commercial reasons. Higher NLP systems (e.g., machine translation) are materialized only when the lower ones (e.g. , partition-speech tagger, syntactic parser) are successfully built. This nonfictional dependency exists even among the lower NLP systems. A morphological analyzer can be an important component for a partoj- speech (paS) tagger particularly in dealing with unknown words. A pas tagger, which is a system that uses various sources of information to assign possibly unique pass to words, in turn, can be used as an input to a syntactic parser. Writers in the area of NLP argue that if the pas tagger is accurate, th is method is an excellent one. Th is thesis can be taken as an attempt to integrate ideas and outputs of previously attempted Amharic NLP prototypes towards solving a birther problem in the NLP of the language, i.e. automatic Amharic complex sentence parsing. Syntactic parsing underlies most of the applications in natural language processing. Parsers are already being used extensively in a number of disciplines such as in computer science (for compiler construction, database interfaces, artificial intelligence, etc), and in linguistics (for text analysis, co/pora analysis, machine translation, etc.). Although there have been some comprehensive studies of Amharic syntax from a linguistic perspective, attempts for investigating it from a computational point of view is ave/y recent story. In this thesis, Amharic word and phrase classes, sentence formalism, mo/pho logical properties peculiar to complex sentence formation in the language, and attempts to extract such features that enable implementation of automatic Amharic complex sentence parser is presented. The sample data used in this study has been taken from references that are widely used in the teaching-learning process of the language. This data has also been manually analyzed, tagged, parsed, and then used as a corpus to extract the grammar rules and to assign probabilities. Algorithms that can use the morphological, lexical and syntactic properties of the language have been customized and modified. Experiments have been conducted in this study using the training set and test set. The first experiment was conducted on the patrol-speech tagger to see the state of its performance when a morphological analysis is embedded in it. The result of this experiment showed that the tagger attained 98. 7% and 94% of ac curacies on the training set and the test set, respectively. The experiments on complex sentence parsing showed 89.6% accuracy result on the training set and 81.6% accuracy result on the test set prepared for this purpose.