Keyword Extraction in Arabic and English Using Page Rank Algorithm

Volume 9, Issue 9, September– 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24SEP454
Keyword Extraction in Arabic and

English using Page Rank Algorithm
Meran M. A. Al Hadidi
Al-Balqa Applied University (BAU)
Amman, Jordan
Abstract:- This paper shows a comparison in applying documents. Furthermore, the rise of social media has
TextRank algorithm for keyword extraction to English introduced new challenges, where rapid and accurate
and Arabic Text. TextRank algorithm is applied by extraction of information, as discussed in [6], necessitates
constructing a graph whose vertices that are formed by advanced algorithms like the Page Rank. Consequently,
candidate words extracted from the title and the abstract keyword extraction evolves as not just a technical necessity,
of the given Arabic and English after applying a tagging but as a catalyst for improving knowledge discovery across
filter to that text to decide the importance of vertices diverse applications.
within that graph.
II. THEORETICAL FRAMEWORK OF THE PAGE
Keywords:- Text Rank, Keyword Extraction, Arabic Text, RANK ALGORITHM
English Text.
Page Rank is an algorithm that was developed by Google
I. INTRODUCTION founders Larry Page and Sergey Brin in e 1998 [13].
PageRank calculates the importance of web pages based on
The quest for effective keyword extraction techniques the number and quality of links pointing to them. It is similar
has gained considerable traction, particularly in multilingual to the page rank algorithm used in Google’s search engine,
contexts, where languages such as Arabic and English present which assigns a rank to web pages based on the number and
unique computational challenges. In the digital age, the quality of inbound links to those pages.
exponential growth of online content has underscored the
necessity for sophisticated methods to distill relevant PageRank works by analyzing the link patterns between
information from vast text corpora. A significant aspect of this web pages to determine the importance of various pages. In
endeavor is the application of algorithms that can efficiently essence, it is based on the idea that if two pages have backlinks
identify and rank keywords, thereby facilitating improved from the same site, it means they are both important and
information retrieval and natural language processing tasks. closely related. Therefore, if two pages have more backlinks
Among the myriad of algorithms developed, the Page Rank than a third page, the first two pages are deemed more
algorithm, originally conceived for web page ranking, has important than the third page. PageRank works by calculating
emerged as a promising candidate for keyword extraction due the score of each page based on this link pattern.
to its ability to account for the importance of terms within the
broader textual landscape. This essay will explore the The Page Rank algorithm serves as a foundational
application of the Page Rank algorithm in both Arabic and concept in the field of information retrieval, particularly for
English contexts, illuminating its effectiveness and processing and ranking content in vast networks such as the
adaptability to the linguistic intricacies inherent in each internet. By quantifying the importance of each webpage
language. based on the links directed towards it, the algorithm efficiently
organizes information according to relevance. This theoretical
 Overview of Keyword Extraction and its Importance in framework can be extended to keyword extraction, wherein
Natural Language Processing keywords are treated analogously to web pages. The
In the realm of Natural Language Processing (NLP), significance of a keyword can thus be evaluated based on its
keyword extraction serves as a foundational technique that connections to other terms within a document or across
enhances information retrieval and text summarization. By multiple documents. This approach mirrors aspects of
identifying the most relevant terms within a text, keyword clustering and authority finding within online social networks,
extraction helps in distilling vast amounts of information into like those discussed in recent studies [3] and [4], where the
manageable and meaningful segments, thereby facilitating relationships between entities inform the identification of
effective decision-making and data analysis. Techniques such experts and topical relevance, ultimately enhancing
as text mining and semantic interpretation have been knowledge transfer. Consequently, leveraging the Page Rank
employed to refine this process and enhance its accuracy. For algorithm for keyword extraction in Arabic and English aligns
instance, in underground engineering, efficient extraction of with contemporary models of semantic analysis,
information from reports is crucial, as highlighted in [5], demonstrating its versatility and applicability across different
which discusses the integration of BERT-BiLSTM-CRF linguistic contexts.
models and visualization techniques for data mining from text
IJISRT24SEP454 www.ijisrt.com 385

 Explanation of the Page Rank Algorithm and its Experiments were performed with various syntactic
Application in Text Analysis filters, including: all open class words, nouns and adjectives,
Widely recognized in the realm of information retrieval, and nouns only, and the best performance was achieved with
the PageRank algorithm evaluates the relevance and the filter that selects nouns and adjectives only [14].
importance of web pages based on their interconnections.
Initially developed for ranking web pages on Google, it Experiments were also performed with directed graphs,
operates on the principle that more important pages are likely where a direction was set following the natural flow of the
to receive more links from other pages. In text analysis, text. Their TextRank system leads to an Fmeasure higher than
PageRank can be effectively repurposed to ascertain keyword any of the previously proposed systems.
significance by treating words or phrases as nodes in a graph
connected through their co-occurrences within a given corpus.  Keyword Extraction Using The Page Rank Algorithm To
This approach allows for deeper semantic understanding and Arabic Text
authority identification, particularly in diverse contexts such The text rank algorithm was tested on a collection of
as rumor verification or question answering systems [9] and articles published in the Arabic language that were collected
[10]. By leveraging the topological structure of textual manually from the Internet from a range of disciplines: Islamic
information, the algorithm facilitates the extraction of salient law, basic and social sciences, child-rearing and IT [16]. This
keywords across different languages, including Arabic and dataset was divided into two sets: 100 documents for training,
English, thereby enhancing the accuracy and relevance of and 50 documents for a hold-out test set. Some statistics were
results in complex natural language processing tasks. calculated for this dataset which was of great benefit in the
implementation of experiment. The Arabic abstracts with their
 Keyword Extraction in English Text titles were reprinted in Notepad files.
Keyword extraction in English text has been a common
practice for many years, but it has been proven to have certain The results of this experiment show that it is possible to
limitations. One of these limitations is that it does not take into build a keyword extraction system using the page rank
account the structure of the text, such as the hierarchy of the algorithm and to apply it successfully to Arabic texts, despite
content or the relationships between the words. This has led to the difficulties of Arabic language which is morphologically
the development of alternative techniques such as Latent rich and highly ambiguous due to its complex morpho-
Dirichlet Allocation and Lat ent Semantic Indexing, which use syntactic agreement rules and the presence of a lot of irregular
probabilistic models to identify the most important words in word forms [16]. The results of several experiments on the
the text. Although these methods have their own limitations, training dataset revealed the most suitable the suitable
such as the difficulty of interpreting the model output, they techniques and tools to use to obtain the best possible results
have shown to be more accurate than traditional methods. when applying the proposed keyword extraction system to the
Page Rank, on the other hand, is a more established and test dataset.
reliable method for ranking web pages. It is based on the
assumption that a page is more valuable if it receives more III. COMPARATIVE ANALYSIS OF KEYWORD
links from other pages. This makes it well-suited for use in EXTRACTION IN ARABIC AND ENGLISH
text-based keyword extraction, as it provides a measure of the
importance of each word in the context of the entire text. The process of keyword extraction presents unique
Overall, while traditional keyword extraction algorithms may challenges and opportunities across different languages,
be less accurate than newer methods, they are still widely used particularly between Arabic and English. Recent
due to their simplicity and their ability to handle large datasets. advancements in unsupervised learning, particularly in the
Page Rank, in particular, has become a popular choice for context of authority identification and keyphrase extraction,
keyword extraction in a variety of industries and applications. shed light on the variances inherent in each language. For
instance, the comparative scarcity of annotated datasets in
Mihalcea and Tarau in [14] constructed a network graph Arabic complicates the development of robust keyword
is constructed using candidate keywords as nodes, where co- extraction methodologies similar to those available for
occurrence is used to draw edges between them, and then the English. As noted in [9], the integration of topical semantic
page rank algorithm is applied to the graph to rank the features into authority finding for Arabic Twitter demonstrates
importance of each keyword. Their text rank algorithm makes a promising approach, yet highlights the need for leveraging
use of the Hulth (2003) dataset [15]. This dataset consists of diverse linguistic characteristics unique to Arabic.
2000 English abstracts from the international Information Conversely, the English language benefits from established
Science, Physical Sciences, Engineering and Computer frameworks that utilize similarity measures and advanced
Sciences (INSPEC) database from the years 1998 to 2002 and topic modeling, as discussed in [8]. This divergence
includes articles Computers and control, and information underscores the necessity for tailored approaches that respect
technology (IT). The resulting keyword dataset was divided: each languages semantic richness while adopting effective
1000 for training, 500 for validation, and 500 for a hold out strategies like the Page Rank Algorithm to enhance overall
test. The results were evaluated using precision, recall, and F- performance in keyword extraction.
measure.

When applying the Page Rank Algorithm to Arabic text, IV. CONCLUSION
several unique challenges emerge, primarily due to the
complexities inherent in the Arabic language. Arabic is In synthesizing the results of this research, it is evident
characterized by a rich morphological structure, with root- that the application of the Page Rank algorithm for keyword
based word formation that can complicate keyword extraction extraction in both Arabic and English presents a promising
processes. Additionally, the syntax of Arabic differs avenue for enhancing information retrieval systems. By
significantly from that of English, often featuring a verb- systematically evaluating the efficiency and effectiveness of
subject-object order that can affect the way terms are the algorithm in processing diverse language contexts,
prioritized. To address these challenges, adaptations to the researchers can bridge significant gaps in existing
Page Rank Algorithm have been implemented, such as methodologies. Furthermore, integrating findings from
incorporating stemming techniques that reduce words to their relevant studies demonstrates the importance of leveraging
root forms, thereby improving the algorithm's ability to advanced techniques for better data representation. For
recognize semantically similar terms. Case studies, such as instance, the authority finding method in social networks
those conducted by Alotaibi et al. (2020), have demonstrated highlights the necessity of identifying experts for knowledge
successful adaptations of the Page Rank Algorithm in Arabic sharing [11], suggesting that a similar approach can be
contexts, showing that when combined with machine learning employed to improve keyword extraction methods.
techniques, it can effectively identify keywords in news Additionally, the insights from clustering algorithms in
articles and scholarly texts, thereby enhancing information question answering systems reveal the critical role of semantic
retrieval in Arabic literature. relationships in enhancing lexical retrieval, reinforcing the
utility of Page Rank in multifaceted linguistic environments
In contrast, the application of the Page Rank Algorithm [4]. Ultimately, this exploration underscores the significance
to English text benefits from the language's relatively of continued innovation in keyword extraction techniques for
straightforward morphological structure and syntactic both Arabic and English languages.
conventions. English keywords typically exhibit clear
semantic roles, making their extraction less complicated than In conclusion, the comparative analysis of the Page Rank
in Arabic. The standard procedures utilized in the Page Rank Algorithm’s application in Arabic and English text illustrates
Algorithm for English text often include tokenization, which the significant linguistic challenges and adaptations necessary
divides text into individual words or phrases, followed by the for effective keyword extraction. While the foundational
calculation of term frequency-inverse document frequency principles of the algorithm remain consistent across
(TF-IDF) to weigh the importance of keywords. These languages, the specific characteristics of each language
procedures align well with the linear and often predictable require tailored approaches to optimize its efficacy. Arabic
structure of English sentences, facilitating efficient keyword poses unique morphological and syntactic challenges that
identification. Successful implementations of the Page Rank necessitate innovative adaptations, whereas English leverages
Algorithm for English keyword extraction can be observed in its structural simplicity for more straightforward
various domains, including academic research and digital implementations. Ultimately, understanding these differences
marketing. For example, studies by Mihalcea and Tarau not only enhances the efficacy of keyword extraction in
(2004) have illustrated the algorithm's effectiveness in diverse linguistic contexts but also contributes to the broader
summarizing scientific papers, further emphasizing its field of information retrieval, paving the way for more
adaptability and robustness in an English-language context. effective search and analysis tools in an increasingly
multilingual digital landscape.
 Challenges and Techniques in Extracting Keywords from
Arabic Texts vs. English Texts  Summary of Findings and Future Directions for Research
The extraction of keywords from Arabic texts presents in Keyword Extraction Using Page Rank Algorithm
unique challenges compared to English due to linguistic and The integration of the PageRank algorithm into keyword
structural differences inherent in each language. Arabics rich extraction has yielded promising results, particularly in
morphology, characterized by a diverse array of roots and enhancing the accuracy and relevancy of extracted terms in
affixes, complicates the identification of significant lexical both Arabic and English texts. Our findings indicate that
items, making it imperative to employ specialized techniques PageRanks ability to assess the importance of words based on
that account for this complexity. Conversely, English their contextual relationships significantly outperforms
keywords often derive from a relatively simpler traditional methods, such as frequency-based approaches.
morphological structure, allowing for more straightforward Furthermore, the comparative analysis demonstrates that the
extraction processes. Techniques such as utilizing fuzzy sets algorithm adapts well across languages, making it a versatile
and RSS-based ranking algorithms have proven advantageous tool for multilingual applications. Future research should
for refining the keyword extraction process, particularly in focus on optimizing the PageRank algorithm for domain-
managing the nuances of Arabic syntax [2]. Moreover, the specific contexts, as well as exploring hybrid models that
integration of advanced algorithms, like the PageRank combine PageRank with machine learning techniques. This
algorithm, has emerged as an effective tool for enhancing the could further refine the extraction process by incorporating
accuracy of keyword extraction across both languages by semantic understanding, thus addressing the nuances inherent
prioritizing the significance of terms based on their contextual in diverse languages. Additionally, empirical testing in
relevance, ultimately facilitating better resource extraction various digital environments will contribute to the robustness
and information retrieval [1].

of keyword extraction methodologies, fostering advancements

in both information retrieval and natural language processing.
REFERENCES
[1]. Ahmed Al-Rawi, Carmen Celestini, Nicole K.

Stewart, Nathan Worku, "How Google Autocomplete
Algorithms about Conspiracy Theorists Mislead the
Public", 2022
[2]. Haijun Chen, Wei-Chang Yang, "Rank algorithm of
web English educational resources based on fuzzy sets
and RSS", 2021, pp. 1-11
[3]. Hend Aldahmash, "Rumor gatekeepers: Unsupervised
ranking of Arabic twitter authorities for information
verification", 2024
[4]. Rana Husni AlMahmoud, "The effect of clustering
algorithms on question answering", 2023
[5]. Ruiqi Shao, "Integrated natural language processing
method for text mining and visualization of
underground engineering text reports", 2024
[6]. Chenguang Wang, "Near-real-time earthquake-
induced fatality estimation using crowdsourced data
and large-language models", 2024
verification", 2024
[8]. Qi Liu, "AdaptiveUKE: Towards adaptive
unsupervised keyphrase extraction with gated topic
modeling", 2024
verification", 2024
verification", 2024
[13]. Brin and L. Page. 1998. The anatomy of a large-scale
hypertextual Web search engine. Computer Networks
and ISDN Systems, 30(1–7).
[14]. Mihalcea, R., and Tarau, P. 2004. Textrank: Bringing
order into texts. In Proceedings of empirical methods
on natural language processing conference.
[15]. Hulth, A. 2003 Textrank: Bringing order into texts. In
Proceedings of the 2003 conference on empirical
methods in natural language processing. Pages 216-
23. Japan.
[16]. Hadidi, M., Alzghool, M., Muaidi, H., 2019,
Keyword Extraction from Arabic Text using the Page
Rank Algorithm. Pages 3459-3504.

Keyword Extraction in Arabic and English Using Page Rank Algorithm

Uploaded by

Keyword Extraction in Arabic and English Using Page Rank Algorithm

Uploaded by

Volume 9, Issue 9, September– 2024 International Journal of Innovative Science and Research Technology

ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24SEP454

Keyword Extraction in Arabic and

IJISRT24SEP454 www.ijisrt.com 385

IJISRT24SEP454 www.ijisrt.com 386

IJISRT24SEP454 www.ijisrt.com 387

of keyword extraction methodologies, fostering advancements

[1]. Ahmed Al-Rawi, Carmen Celestini, Nicole K.

IJISRT24SEP454 www.ijisrt.com 388

You might also like