Ontology Learning Applied In Education: A Case Of The New Testament

Abstract

The objective of this work is to show how ontology learning can be applied for the semantic analysis of text, in order to extract concepts, relations, which can be further used for automated summaries or critical comparison. Such activities are very important in education as they can allow dynamic creation of content or analyses that can be further used in the educational process. Since ontology-learning methods require large corpus of unstructured data, we have chosen the Bible as source for the text. Another reason for the choice was that the Bible is one of the most studied texts by scholars and we can test our conclusions, such as importance of some terms, with existing knowledge. In this way, the new developed methods are validated and they can be used successfully in other educational domains. The Bible is the religious text of Christians and Jews. The Bible contains a collection of scriptures that was written by many authors, at different time and locations. Computationally, the Bible contains semi-structured information due to its organized structure of scriptures and numbered chapters. We have used Text2nto as the main tool in order to obtain the most relevant concepts from the New Testament. After that we analyze extracting the most relevant concepts and the range of similarity for each domain identified in the New Testament. Those methods can be employed for automatic generation of content that can be further used in the educational process.

Keywords: Automatic Ontology GenerationTheological EducationICT in Education

Introduction

The objective of this work is to show how ontology learning can be applied for the semantic

analysis of text, in order to extract concepts, relations, which can be further used for automated

summaries or critical comparison. Such activities are very important in education as they can allow

dynamic creation of content or analyses that can be further used in the educational process.

In the field of computer science, the ontologies have gained significant importance in last years,

due to their applicability for different fields including information extraction. Information extraction in

computer science deals with the analyses of unstructured sources of text in order to extract relevant

information. The process of information extraction requires statistical analysis, natural language

processing techniques and machine learning methods.

Since ontology-learning methods require large corpus of unstructured data, we have chosen as

source for the text the Bible. Another reason for the choice was that the Bible is one of the most studied

texts by scholars and we can test our conclusions, such as importance of some terms, with existing

knowledge. In this way, the new developed methods are validated and they can be used successfully in

other educational domains.

The Bible is the religious text of Christians and Jews. The Bible contains a collection of scriptures

that was written by many authors, at different time and locations. This book is divided in two parts: The

Old Testament and The New Testament. Computationally, the Bible contains semi-structured information

due to its organized structure of scriptures and numbered chapters. We have used Text2nto (Cimiano &

Volker, 2015) as the main tool in order to obtain the most relevant concepts from the New Testament.

After that we analyze extracting the most relevant concepts and the range of similarity for each domain

identified in the New Testament. Those methods can be employed for automatic generation of content

that can be further used in the educational process - i.e. online learning (Istrate, 2010).

Paper Theoretical Foundation and Related Literature

Ontologies are defined as the theory of objects and their ties according to the articles written by

(Berners-Lee et al., 2001). In computer science, an ontology is the formal specification of a domain and is

composed of properties, types, and interrelationships of entities that should exist for a specific domain

according to the articles (Maedche et al., 2003) and (Gruber, 1993).

They provide many principles in order to distinguish different types of objects and their

dependencies. Thus, we can distinguish three different types of ontologies: descriptive, formal and

formalized ontologies.

Sharing information is an important issue for researchers and ontologies represent to them a

formal way to store and organize the information. The concepts defined in ontologies can be stored and

shared in many formats, the most common ones being OWL/RDF.

Ontologies are used today in order to obtain the most relevant terms for a given domain. We can

search for the documents containing their terms over the web. The documents containing the most

relevant terms, are the most relevant for a search. Also, documents with a larger number of terms from a

given ontology, are more relevant than documents with a small number of terms. Therefore, given an

ontology automatically generated from a bunch of text, one can get the most relevant documents from a

web search, documents that contains terms from ontologies (concepts, instances) and use them for

education purposes.

2.1Automatic Extraction of Ontologies in Religious Books

Currently we found only a few articles about searching and information extraction from religious

books using ontologies.

According to the article written by (Shamsuzzaman Sadi et al., 2016), people are often exploiting

the knowledge of the Quran. Searching mechanisms and extracting information algorithms requires an

ontological modeling of the data. In the results section of that research, verses and concepts of interest

were retrieved from the Quran using SPARQL queries. Searching for particular concepts inside the book

of Quran and retrieving verses that contains those concepts is an important issue for the followers of

Islam that are eager in gaining the Quranic knowledge. This article presents a research problem that is not

too different for the Bible’s New Testament, and other religious texts.

In the article written by (Ahmad et al., 2013), ontologies are also used to represent and encapsulate

Quran’s knowledge. In the same article a comparison based on a generic framework is done. The

comparison is focused on the role of ontologies and how these ontologies are compared to other ontology

applications in other areas.

When it comes to the Bible we found far less relevant research articles. A research for

automatically transforming Old Testament texts in Hebrew into English-based conceptual graphs was

made in an article by (Ulrik, 2007). The method utilizes ontology of the text plus syntactic analysis and

transformational rules. The results presented are interesting but not very relevant for our research.

In conclusion, this research is relevant because no other studies were made in order to search and

extract information from the Bible’s New Testament. The framework used in this research in order to

exploit the knowledge from the New Testament helps us to extract not only the most relevant concepts

but also the ranges of similarities for the concepts extracted.

Methodology

Domain ontologies are very hard to develop manually and a better approach is to generate them

automatically. The platform Text2Onto is an important part of our ontology generator (Popa et al., 2016;

Vasilăţeanu et al., 2015). In order to obtain the most relevant terms for our ontology we made some

improvements to this platform. First an algorithm was added to analyse similarity in terms of concepts

that are synonyms to another. Second, new rules for ontology generation were added to Text2Onto. The

last improvement was automatizing the process of finding similar concepts between the two types of

ontologies.

The list containing the most important concepts was obtained by drawing the information from the

New Testament using Text2Onto. In the obtained list of concepts each term is associated with a score,

called relevance. This relevance score is between zero and one. We have chosen for our research, all the

concepts with a score that is larger than a given threshold. The threshold was chosen in a dynamic way in

function of the domain and the corpus size.

Below a sample of the concepts generated by the Tet2Onto algorithm from New Testament.

See Full Size >

Results

Before presenting the relevant results, we need to discuss some limitations of the method

presented in this paper. First one should note that the best scenario was to use the original text in

Hebrew/Greek for such an analysis. However, because the tool that we use has only the grammar of the

English language, we were restricted to the New Testament text in English.

There are several translations of the text in English. Due to the online availability, we used the one

from David Robert Palmer's translation (Palmer, 2015) that has a text that is also compatible with the tool

grammar that we use.

Part of the ontology that we obtained is presented in the following figure – some of most important

concepts.

See Full Size >

Analyzing the ontology, the following conclusions can be formulated. First there are concepts

which have theological relevance such as “ man ”, “ brother ”, “ spirit ” (of God), “ disciple ”, “ heaven ”,

“ faith ”, etc. On the other hand, there are also general concepts that are not theological relevant such as

“ time ”, “ day ” (in the text with the sense of day in general), and “ word ” (common, not Word of God ). In

this respect the concepts of the ontology that are automatically generated should be also analyzed by

humans, in a post-generation process, to extract the ones that are theologically relevant.

Looking to the concepts of the ontology, we see that in the first 25 most important ones, the

followings are found that are theological relevant:

- “ man ” - similar also with “ world ” (according to the process of automatic analysis)

- “ brother ”

- “ people ”

- “ spirit ” (of God) – similar with “ life ” (according to the process of automatic analysis)

- “ heaven ”

- “ faith ”

- “ world ” – similar with “ man ”, “ earth ” (according to the process of automatic analysis)

- “ law ” (of God) – similar with “ life ” (according to the process of automatic analysis)

- “ sin ”

- “ angel ”

- “ glory ”

- “ kingdom ”

Looking to the generated concepts from the New Testament (NT) we find that:

a) on the human part : 1) they speak about this world (“ man ”, “ people ”, “ earth ”); 2) about the existing

“ sin ” and “ law ” (of God); 3) about the “ faith ” and “ kingdom ”. From a theological perspective, this can be

correlated with the message of NT that this world is under the sin , and through faith , one can reach the

kingdom of heaven – concepts present also in The Orthodox Nicene Creed (381), Athanasian Creed (293-373), etc. It is also interesting to observe the existence of the concept of “ brother ” in the first three

generated concepts, which shows that the early Christians see each others as brothers and also that among

them this concept was of importance.

b) on heaven side : 1) they speak about “ spirit ” (of God), “ kingdom ”, “ angels ”, “ heaven ” which from a

theological perspective this can be correlated with the coming kingdom of heaven and the work of angels

and spirit of God. One can note that some important concepts of NT are missing such as “ God” . “ Jesus

Christ ”. The reason for this is the fact that in the automatic generation of the concepts proper noun are

eliminated (as well as conjunctions, articles, etc.). This fact – elimination of proper nouns - which is

suitable in other domains is not suitable for theology. This is a difference between the theological domain

and other domains that should be taken into account when applying this method.

Discussions

From the analysis performed (see the discussion in the section of results) the followings can be

deducted:

- Such methods are good to generate ontology in a given domain. Looking to the generated

concepts one can get a grasp of some important terms and concepts that play a role in a given domain.

With respect to the automatic generation of content for learning, those concepts can be: 1) keywords that

can be used for content searching and 2) used for weighting of different texts that contains them for

selecting the most relevant ones as content for e-learning.

- However, the method itself has limitations. We discussed some of them. First we note that the

best was to use original Greek/Hebrew text in place of English, but we were restricted by the existence of

English grammar used by the tool. Second, the tool generated only part of the existing important

concepts. By the elimination of proper nouns, tool eliminates also important concepts such as “God”,

“Jesus Christ” that are central to the New Testament theology. Finally, such analysis cannot reveal things

such as Trinity for example.

As a conclusion, such methods are important and useful, but, as this case study reveals, some

human analysis in a post-generation phase might be needed for improvement of the ontology.

Conclusions

The objective of this work is to show how ontology learning can be applied for the semantic

analysis of text, in order to extract concepts, relations, which can be further used for automated

summaries or critical comparison. Such activities are very important in education as they can allow

dynamic creation of content or analyses that can be further used in the educational process.

Since ontology-learning methods require large corpus of unstructured data, we have chosen as

source for the text the Bible. Another reason for the choice was that the Bible is one of the most studied

texts by scholars and we can test our conclusions, such as importance of some terms, with existing

knowledge. In this way, the new developed methods are validated and they can be used successfully in

other educational domains. The Bible is the religious text of Christians and Jews. The Bible contains a collection of scriptures

that was written by many authors, at different time and locations. Computationally, the Bible contains

semi-structured information due to its organized structure of scriptures and numbered chapters. We have

used Text2nto as the main tool in order to obtain the most relevant concepts from the New Testament.

After that we analyze extracting the most relevant concepts and the range of similarity for each domain

identified in the New Testament.

From the analysis performed, it results that such methods are good to generate ontology in a given

domain. Looking to the generated concepts one can get a grasp of some important terms and concepts that

play a role in a given domain. From automatic generation of content for learning, they can be keywords

that can be used for content searching and also automatic weighting of different text that contains them.

However, the method itself has limitations. We discussed some of them. First we note that the best

was to use original Greek/Hebrew text in place of English, but we were restricted by the existence of

grammar used by the tool. Second, the tool generated only part of existing important concepts. By the

elimination of proper nouns, tool eliminates also important concepts such as “God”, “Jesus Christ” that

are central to the New Testament theology. Finally, such analysis can not reveal things such are Trinity

for example.

As a conclusion, such methods are important and useful, but, as this case study reveals, some

human analysis in a post-generation phase might be needed for improvement of the ontology.

References

1. Ahmad, O., Hyder, I., Iqbal, R., Azmi Murad, M. A., Mustapha, A., Sharef, N. M., Mansoor, M. (2013).
2. A survey of searching and information extraction on a classical text using ontology based semantics modeling: A case of Quran, Life Science Journal. Retrieved from http://www.lifesciencesite.com/lsj/life1004/181_19656life1004_1370_1377.pdf Athanasian Creed (293-373) Retrieved from https://www.urcna.org/urcna/Creeds/Athanasian%20Creed.pdf Berners-Lee, T., Hendler, J., Lassila, O. (2001). The semantic web. Scientific american, vol. 284, no. 5, pp. 28-37.
3. Cimiano, P., V¨olker, J. (2015). Text2Onto A framework for ontology learning and data-driven change discovery. Retrieved from http://www.dfki.de/~horacek/text2onto.pdf Gruber, T. R. (1993). A translation approach to portable ontology specifications, in knowledge acquisition, vol. 5 no. 2, pp: 199-220. Retrieved from http://tomgruber.org/writing/ontolingua-kaj-1993.pdf în educație, Retrieved fromIstrate, O.(2010). Efecte și rezultate ale utilizării TIC https://www.academia.edu/3356979/Efecte_si_rezultate_ale_utiliz%C4%83rii_TIC_%C3%AEn_e ducatieMaedche, A., Neumann, G. and Staab, S. (2002). Bootstrapping an ontology-based information extraction system, Intelligent exploration of the web. Physica-Verlag HD, pp. 345-359. Retrieved from ftp://ftp.dfki.de/papers/local/bootstrapping.pdf Palmer, D. R. translation. (2015). Holly Bible - New Testament.
4. Popa, R. C., Vasilateanu, A., Goga, N. (2016). Ontology based multi-system for SME knowledge workers, IEEE ISSE, Edinburg. Retrieved from https://www.researchgate.net/publication/310808122_Ontology_based_multisystem_for_SME_knowledge_workers Shamsuzzaman Sadi, A. B. M., Towfique, A., Mohamed, A., Abdillahi, H. A., Sazid, Z. K., Mohamed, M. R., Ghassan, S. (2016). Applying ontological modeling on quranic nature domain published at the International Conference on Information and Communication Systems, 7th International Conference on Information and Communication Systems (ICICS). Retieved from https://arxiv.org/ftp/arxiv/papers/1604/1604.03318.pdf The Orthodox The Nicene Creed (381).
5. Ulrik, S. P. (2007). Genesis 1:1-3 in Graphs: Extracting conceptual structures from biblical hebrew.
6. from https://www.researchgate.net/publication/228734851_Genesis_1_1-Retrieved 3_in_Graphs_Extracting_Conceptual_Structures_from_Biblical_Hebrew Vasilăţeanu, A., Goga, N., Tanase, E. A., Marin, I. (2015). Enterprise domain ontology learning from web-based corpus, 6th ICCCNT 2015 Conference Proceedings, USA, pg. 112-117. Retrieved from https://www.researchgate.net/publication/281649559_Enterprise_Domain_Ontology_Learning_Fr om_Web-Based_Corpus