AUTOMATIC CAUSAL RELATION EXTRACTION FOR AMHARIC LANGUAGE TEXTS USING CNN

MELKAMU, MEKURIAW

dc.contributor.author	MELKAMU, MEKURIAW
dc.date.accessioned	2020-10-07T11:27:26Z
dc.date.available	2020-10-07T11:27:26Z
dc.date.issued	2020
dc.identifier.uri	http://hdl.handle.net/123456789/11287
dc.description.abstract	The extraction of a causal relation could be a difficult task while it is very important domain area for Natural Language Processing (NLP). There are many existing approaches and techniques developed in order to tackle extraction of causality relation task. These approaches are categorized either rule-based (non-statistical) or machinelearning- based (statistical) method. In the case of statistical or rule-based approaches, it needs widespread manual works to design and construct handcrafted patterns and rules, however, the precision and recall are low because of the complexity of causal relation expressions in texts. On the other hand, the non-statistical or machine learning approaches, are the current approaches either rely on sophisticated feature engineering which is error-prone, or rely on large amount of labeled data which is impractical for the extraction of causal relation problem. In order to deal with the above issues, we have proposed a one of the deep learning approaches called a Convolutional Neural Network (CNN) for causal relation extraction in this paper. This CNN approach consists of a word embeddings and position embeddings. The word embedding and position embedding allows to represent each word with a vector form and calculate the position of the word on the given sentence from head to tail respectively for the model. Furthermore, additional semantic features that are useful for identifying causal relations that are stated in ambiguous form also allows to determine the direction of the causality are created. We have been used our own data set that we have collected from three domain areas of Amharic text, i.e., health related, agricultural related and environmental related domain areas, to assess the capability of CNN to efficiently extract the causal relation from texts, and the model that we have proposed outperforms current state-ofart models for causal relation extraction.	en_US
dc.language.iso	en	en_US
dc.subject	Information Technology	en_US
dc.title	AUTOMATIC CAUSAL RELATION EXTRACTION FOR AMHARIC LANGUAGE TEXTS USING CNN	en_US
dc.type	Thesis	en_US