Name entity recognition and part of speech tagging: case study of Kikamba

Abstract

There has been exponential multiplication of electronic information for the last two decades which has generated a large digital library for everyone to access over the internet. However, this library consists of unstructured documents where queries cannot be run as with a database so as to get preview of the content or certain details of interest. As a result, a need for language tool arises. Natural language processing has provided a channel whereby the above challenge can be resolved using Name entity recognition (NER) in which a machine learning system is developed which can identify organization, personal and location names in various documents and report them from which you can get a glimpse of the contents of the documents. In this project we present a Kikamba Name Entity Recognition using a memory based approach where supervised and bootstraps learning methods are applied to a carefully annotated corpus. To build the training set, a corpus is manually annotated. An annotated seed is also provided to facilitate bootstrap. Simultaneously, generation of Part of Speech tagging is done. The resultant classifiers are evaluated. The Aim of the project is a tool for analysis of electronic documents and at the same time find out the challenges that are peculiar to Kikamba language so as to compare with other languages which already have been tackled.