Named Entity Recognition for Amharic Language
No Thumbnail Available
Date
2010-11
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Addis Ababa University
Abstract
Named Entity Recognition (NER) is a process of identifying and categorizing all named entities
in a document into predefined classes like person, organization, location, time, and numeral
expressions. This identification and classification of proper names in text has recently considered
as a major importance in natural language processing as it plays a significant role in various
types of NLP applications, especially in information extraction, information retrieval, machine
translation, and question-answering. This paper reports about the development of a NER system
for Amharic using Conditional Random Fields (CRFs). Though this state of the art machine
learning method has been widely applied to NER in several well-studied languages, this is the
first attempt to use this method to Amharic language.
The system makes use of different features such as word and tag context features, part of speech
tags of tokens, prefix and suffix. Since feature selection plays a crucial role in CRF framework,
experiments were carried out to find out most suitable features for Amharic NE tagging task.
During the experiment, four different scenarios were considered based on the different
combination of features. In the first scenario all the features were considered, in the second
scenario all the features except POS tags of tokens were considered. In the third and fourth
scenarios all the features except prefix and suffix respectively were considered.
The experimental results show that for different combinations of features, we have got different
results. In scenario one experiment, we have got Precision, Recall and F-measure of 72%, 75%
and 73.47% respectively. Taking this as a base line we made the remaining experiments. The
remaining experiments on scenario two, three and fourth, its F-measure of 69.70%, 74.61%, and
70.65% respectively were obtained.
From the above results, it is possible to make a conclusion that word context features, POS tags
of tokens and suffix are important features in NE recognition and classification for Amharic
text.
Keywords: Named Entity Recognition, Conditional Random fields, Named Entities, Amharic
Named Entity Recognition.
Description
Keywords
Named Entity Recognition; Conditional Random Fields; Named Entities; Amharic Named Entity Recognition