Alexandervanderleek/MorphosyntacticApproachNounClassification

Evaluating a Morphosyntactic Approach to Noun Classification for Sepedi and isiZulu. # A Morphosyntactic Approach to Noun Classification ## Summary Evaluating a Morphosyntactic Approach to Noun Classification for Sepedi and isiZulu. Code from our cs honours project, investigating the use of an approximation of morphosyntax through the use of charachter n-grams in various variations to classify nouns into noun classes. Models are to be trained on the noun data and associated class labels i.e abantu: 2, to then classify unseen nouns. The datasets used in our paper have been redacted since it is not publically available. Models are trained on the fly, to allow for the mass testing and continued testing of model variations. Models created and tested included the following. - K-nearest neighbors: KNNClassifcationModel.py, KNNCombinedModel.py - Support vector machine: SVMClassificationModel.py, SVMCombinedModel.py - Decision Tree: DTClassifcationModel.py, DTCombinedModel.py Where the normal classifcation model variant is trained on a single class noun dataset i.e abantu: 2, and the combined model varaint is trained on the singular and plural from together i.e abantu: 1/2 which then feeds said combined prediction to a model trained on a single class classifcation model to deteremine if the classification should be 1 or 2. ## How To Use > [!IMPORTANT] > Before using these models, you need to add your own traning and testing .xlxs sets in the format of column 1 -> noun | column 2 -> class. > You can do this by adding your own noun data to the empty xlsx files located in the /data/[datatype]/data files. > You can use a different language but please note that this is setup for the testing of Sepedi and isiZulu and may cause naming issues. Running **DTClassifcationModel.py** will result in the following prompts. |Prompt |Input |Description | |--------|---------------|-------------| | Select Class format (1: Dual Class, 2: Single Class): | 2 | Train on "1" or "1/2". | | Select language (1: northern sotho, 2: isizulu): | 2 | Language selection (can add ow …