Search

This study outlines the intellectual structure of Library and Information Science in terms of the venues with RefCit2vec, an embedding method inspired by word2vec. The reference lists or cited-by lists of 62,077 articles in 35 venues (journals and proceedings) between 1928 and 2022 are converted into real number vectors by four independent models of RefCit2vec. The document similarities measured by the two models of RefCit2vec exhibit moderate correlations with bibliographical coupling metrics. In contrast, the similarities from the other two models moderately or strongly correlate with co-citation metrics. Each venue is represented by its centroid, the average vector of its constituent documents. By applying hierarchical agglomerative clustering on the venue centroids, 69% of venues robustly emerge in 6 out of 8 clusters. Four clusters consistently form the library-related branch. The bibliometrics/scientometrics branch contains only 1 cluster, whereas the information-related branch contains 3 clusters. 43% of venues are in six subgroups of consistent tree structures. An article is defined as SCIM-alike for it is closer to the SCIM centroid than half of SCIM articles are. 10% of JASIST articles are SCIM-alike upon their reference lists, and 5% of JASIST articles are SCIM-alike in terms of their cited-by lists. The percentage of SCIM-alike articles in JASIST hiked above the average between 2008 and 2018 but has dropped below the average since 2019. As we demonstrate the dynamics in LIS, citation embedding methods like RefCit2vec can incorporate citation-based, text-based, or authorship features to contribute to varied scenarios in investigating or exploring research fronts and scientific knowledge transfer. [ABSTRACT FROM AUTHOR]

Thousands of papers are published at conferences every year. Conference organizers manually assign accepted papers to the sessions according to the title and keywords given by the author. Then organizer names each session, based on his/her experience. These are complex and time-consuming processes and often fail to collect papers on a similar topic in content. This often causes the participant to exit the session after listening to the presentation of one or two papers, because the session name does not fully represent the papers in the session and the papers in the session are not close in content. As a solution to these problems, this paper proposes a method for automatically organizing conference sessions. The method first introduces a network created with a deep learning-based document similarity. Then, sessions are determined with a community discovery method specific to this network, and finally, session titles are extracted with a topic modeling approach. To the best of our knowledge, this paper is the first effort in this direction. Experiments conducted on sessions of three real conferences show that the proposed method is able to create up to 21% better similar sessions, and session names better represent the papers in that session. [ABSTRACT FROM AUTHOR]

Current machine-learning training focuses solely on accuracy. In this study, the weights of other dimensions were examined rather than measuring only the accuracy of machine learning. By comparatively analyzing the decision-making of machine learning and humans in various fields, this study examines how well organizational vision is propagated to lower levels of the organization. Also, the results evaluated by humans and machine learning models were comparatively analyzed from multiple perspectives. As numerical representation methods of words, count-based models (Bag of Words, TF-IDF), artificial neural network (ANN) models (Word2Vec, GloVe), and a vision propagation measurement (VPMS) model combining two methods were used to calculate the similarity between documents, which are comparatively analyzed with the actual results measured by an expert group. The findings of this study can be used as an evaluation metric for how effectively the vision of the upper organization is being disseminated to the lower-level organizations. Additionally, it could be utilized in developing algorithms such as customer segmentation for target marketing using text data.The study makes two key contributions - (i) providing an extensive empirical comparison of document similarity analysis by different ML techniques versus human experts, and (ii) proposing a new VPMS model that outperforms existing methods.

This study developed an automated matchmaking system using deep neural networks to enhance the efficiency of pairing researcher biosketches with funders’ requests for proposals (RFPs). In thus U.S., with over 900 federal grant programs and 86,000+ foundations, researchers often spend up to 200 hours on each application due to low success rates, forcing them to apply multiple times a year. Our approach improves on existing systems by fixing issues like unreliable keyword searches, and one-size-fits-all recommendations. We analyzed 12,991 biosketches from a research institution and 2,234 RFPs from the National Institutes of Health, spanning 2014 to 2019. Employing four advanced deep-learning models, utilizing cross and Siamese encoding strategies, we benchmarked their performance against conventional predictive models such as logistic regression and support vector machines. The most effective model integrated BERT with cross-encoding, a post-BERT BiLSTM layer, and back translation (BC2BT), achieving an F1-score of 71.15%. These results demonstrate the potential of sophisticated natural language processing techniques to automate complex matchmaking tasks in the research funding sector. This approach not only improves the precision of matching researchers to suitable funding opportunities but also sets a promising foundation for future advancements in automated funding mechanisms.

With the progressive and availability of various search tools, interest in the evaluation of information retrieval based on user perspective has grown tremendously among researchers. The Information Retrieval System Evaluation is done through Cranfield-paradigm in which the test collections provide the foundation of the evaluation process. The test collections consist of a document corpus, topics, and a set of relevant judgments. The relevant judgments are the documents which retrieved from the test collections based on the topics. The accuracy of the evaluation process is based on the number of relevant documents in the relevance judgment set, called qrels. This paper presents a comprehensive study, which discusses the various ways to improve the number of relevant documents in the qrels to improve the quality of qrels and through that increase the accuracy of the evaluation process. Different ways in which each methodology was performed to retrieve more relevant documents were categorized, described, and analyzed, resulting in an inclusive flow of these methodologies.

Plagiarism is a significant issue, especially among students, in the education field, where the internet has made a vast amount of information easily accessible. Although numerous plagiarism detection tools are available for English text, detecting plagiarism in low-resource languages such as Hindi remains a challenge. This study proposes a novel case-based approach to evaluate the effectiveness of various plagiarism detection tools on Hindi documents. The proposed approach maps similarities between Hindi documents at the word, phrase, and sentence levels, and generates diverse cases using stopword removal and stemming pre-processing techniques. The study calculates the similarity score between the original Hindi document and its paraphrased version at the word, sentence, and document levels. The results show that the proposed approach outperforms existing plagiarism detection tools for Hindi text, indicating its potential to enhance academic integrity and prevent plagiarism. [ABSTRACT FROM AUTHOR]

Document similarity recognition is one of the most important problems in natural language processing. This paper proposes a plagiarism comparison mechanism called JCF. Initially, the TF–IDF scheme is applied to build a bag of words as the representation of the common features of all documents. Then, the plagiarism comparison is carried out in a coarse-grained manner, which speeds up the similarity comparison. Finally, the most similar documents can then be compared in detail based on a fine-grained approach. In addition, the JCF detects plagiarism at both syntax level and semantic-like level. To prevent the distortion of similarity comparison, this paper further develops a similarity restoration approach such that the proposed JCF can obtain both advantages of quickness and accuracy. Performance studies confirm that the proposed JCF outperforms existing studies in terms of precision, recall and F1 score. [ABSTRACT FROM AUTHOR]

News categorization, which is a common application area of text classification, is the task of automatic annotation of news articles with predefined categories. In parallel with the rise of deep learning techniques in the field of machine learning, neural embedding models have been widely utilized to capture hidden relationships and similarities among textual representations of news articles. In this study, we approach the Turkish news categorization problem as an ad-hoc retrieval task and investigate the effectiveness of paragraph vector models to compute and utilize document-wise similarities of Turkish news articles. We propose an ensemble categorization approach that consists of three main stages, namely, document processing, paragraph vector learning, and document similarity estimation. Extensive experiments conducted on the TTC-3600 dataset reveal that the proposed system can reach up to 93.5% classification accuracy, which is a remarkable performance when compared to the baseline and state-of-the-art methods. Moreover, it is also shown that the Distributed Bag of Words version of Paragraph Vectors performs better than the Distributed Memory Model of Paragraph Vectors in terms of both accuracy and computational performance. [ABSTRACT FROM AUTHOR]

Many journals belonging to different publishers have emerged with the advancements in research. The increase in the number of scholarly journals has made it difficult for researchers to choose the correct journal for publishing their articles. Submitting an article to the correct journal is very important in terms of academic sharing and for shortening the publication time of the article. It is time consuming to determine the most suitable journal in scope among thousands of journal choices for the user. Therefore, journal recommendation systems have been an important tool for researchers. Recommendation systems generally depend on the user's publications, relationships with other authors, etc. The fact that it is based on features makes it not useful for users who are new to the research field. In this study, an approach that recommends a journal is proposed by using the title, abstract, keyword and reference information of the article, without the need of users' information. Unlike other studies, the scope information of the journals is needed to determine the appropriate journals for the article, which is usually obtained from the articles previously published in the related journals. The publications of the journals in the last 3 years have been used to determine the scope of the journal. Unlike the publishers' journal recommendation systems developed so far, this study is a comprehensive recommendation system that includes journals from more than one publisher. In this approach, SBERT has been used to find the similarity of the scope of journals with articles. When the results are compared with the Word2vec, Glove and FastText, which are often the preferred methods in document similarity, it was observed that sentence-level similarity-based recommendations with SBERT are more successful. The experimental results show the effectiveness of our approach. [ABSTRACT FROM AUTHOR]

Companies and research institutes are acquiring monopolized technologies through patents to survive in a rapidly changing market. The applicant is guaranteed by law the exclusive right to the technology for a certain period. A prior art search is an essential task for this purpose. Therefore, the need for a successful prior art search is increasing. Previous studies have disadvantages because relevant fields are limited, requiring an in-depth understanding of specific technologies. This paper proposes a method to overcome these limitations using patent text and citation information. The proposed method searches for prior art using a citation network based on the shortest path derived from a breadth-first search. We examined based on experiments whether the proposed method could be applied to the industry. Based on the results, the proposed algorithm can efficiently search for prior art. Furthermore, it was possible to construct a counterstrategy based on the characteristics of the applicant. The code will be made available at github.com/lee-ju/k_step_PIC.

Article Similarity Identification is one of the most issue in Article Comparison. In the literature, some studies proposed the similarity comparison mechanisms based on Word2Vec, N-gram or Bert. However, a document usually contains a large number of words. Let the source of the comparison be a document. The goal of the comparison is to compare the source document with thousands of documents in the database. It was time-consuming to compare the similarity of one target document and all documents in the database, since the existing mechanisms only can compare the similarity of two documents. As a result, the plagiarism comparison is very time consuming. This paper proposes a plagiarism comparison mechanism, called BUAS, which speeds up the similarity comparison since the Bag of word scheme is initially applied to transform each document as a document vector. Then the most similar document can be found as the candidate document. As a result, the target document only needs to be compared with the candidate document. Performance studies confirm that the similarity calculation by BUAS outperforms existing studies in terms of precision, recall and F1 score. [ABSTRACT FROM AUTHOR]

Currently, as an effect of the COVID-19 pandemic, bioinformatics, genomics, and biological computations are gaining increased attention. Genomes of viruses can be represented by character strings based on their nucleobases. Document similarity metrics can be applied to these strings to measure their similarities. Clustering algorithms can be applied to the results of their document similarities to cluster them. P systems or membrane systems are computation models inspired by the flow of information in the membrane cells. These can be used for various purposes, one of them being data clustering. This paper studies a novel and versatile clustering method for genomes and the utilization of such membrane clustering models using document similarity metrics, which is not yet a well-studied use of membrane clustering models. [ABSTRACT FROM AUTHOR]

Abstract Background Clinical trial registries can be used as sources of clinical evidence for systematic review synthesis and updating. Our aim was to evaluate methods for identifying clinical trial registrations that should be screened for inclusion in updates of published systematic reviews. Methods A set of 4644 clinical trial registrations (ClinicalTrials.gov) included in 1089 systematic reviews (PubMed) were used to evaluate two methods (document similarity and hierarchical clustering) and representations (L2-normalised TF-IDF, Latent Dirichlet Allocation, and Doc2Vec) for ranking 163,501 completed clinical trials by relevance. Clinical trial registrations were ranked for each systematic review using seeding clinical trials, simulating how new relevant clinical trials could be automatically identified for an update. Performance was measured by the number of clinical trials that need to be screened to identify all relevant clinical trials. Results Using the document similarity method with TF-IDF feature representation and Euclidean distance metric, all relevant clinical trials for half of the systematic reviews were identified after screening 99 trials (IQR 19 to 491). The best-performing hierarchical clustering was using Ward agglomerative clustering (with TF-IDF representation and Euclidean distance) and needed to screen 501 clinical trials (IQR 43 to 4363) to achieve the same result. Conclusion An evaluation using a large set of mined links between published systematic reviews and clinical trial registrations showed that document similarity outperformed hierarchical clustering for identifying relevant clinical trials to include in systematic review updates.

In today's rushing world, there's an ever-increasing usage of networking equipment. These devices log their operations; however, there could be errors that result in the restart of the given device. There could be different patterns before different errors. Our main goal is to predict the upcoming error based on the log lines of the actual file. To achieve this, we use document similarity. One of the key concepts of information retrieval is document similarity which is an indicator of how analogous (or different) documents are. In this paper, we are studying the effectiveness of prediction based on cosine similarity, Jaccard similarity, and Euclidean distance of rows before restarts. We use different features like TFIDF, Doc2Vec, LSH, and others in conjunction with these distance measures. Since networking devices produce lots of log files, we use Spark for Big data computing.

The massive amount of information from the internet has revolutionized the field of natural language processing. One of the challenges was estimating the similarity between texts. This has been an open research problem although various studies have proposed new methods over the years. This paper surveyed and traced the primary studies in the field of text similarity. The aim was to give a broad overview of existing issues, applications, and methods of text similarity research. This paper identified four issues and several applications of text similarity matching. It classified current studies based on intrinsic, extrinsic, and hybrid approaches. Then, we identified the methods and classified them into lexical-similarity, syntactic-similarity, semanticsimilarity, structural-similarity, and hybrid. Furthermore, this study also analyzed and discussed method improvement, current limitations, and open challenges on this topic for future research directions. As the results, this paper highlighted the importance of selecting the appropriate preprocessing algorithms to reduce data dimensionality and also combining several algorithms to enhance the overall matching and detection process. [ABSTRACT FROM AUTHOR]

Every year, thousands of papers are published in journals and conferences by researchers in many different fields. These papers are an important guide for other researchers. However, the increasing amount of digital data with the development of information technologies makes it difficult to reach the desired information. Recommendation systems play an important role in facilitating researchers' access to studies on their subjects. It provides faster and easier access to papers on the desired subject. Recommendation systems are developed according to the user profile or subject. In this paper, a novel hybrid paper recommendation system based on deep learning is proposed. The method uses a combination of document similarity, hierarchical clustering, and keyword extraction. Our aim is to group papers in different fields such as computer science, economics, medicine, or in a specific field, according to their subjects, and to present papers with high semantic similarity to the user according to the query entered. The study has been applied on real dataset containing papers from different categories such as machine learning, artificial intelligence, human–computer interaction in computer science. The success of each stage of the study has been evaluated separately. However, looking at the system as a whole, the overall performance of the proposed approach is 80%. Papers having high similarity with their queries have been recommended to users. Thus, access to the studies on the desired subject in the huge amount of papers has been made faster and easier. [ABSTRACT FROM AUTHOR]

Comparing textual content is becoming more and more problematic due to the fact that nowadays data is very dynamic. The application of sophisticated methods enables us to compare how similar the documents are to each other. In our research we apply the Cosine Similarity method to compare the similarity of several documents with each other. We also apply the TF-IDF technique which enables us to normalize the results. Normalization of these results is necessary for the fact that there are some words that are repeated several times and from this repetition is determined their importance. Finally we can see a comparison between the similarity of documents with normalized and non-normalized results. As can be seen in the research, the normalization of results has a great value in comparing documents with textual content. [ABSTRACT FROM AUTHOR]

Measuring similarity between texts is an essential task in a large variety of applications. Contemporary approaches for this task rely heavily on statistical and lexical information to represent text. They thus produce opaque and hard to interpret models that could be hard to adapt in some applications and hamper the user experience. To represent the text document more interpretable, we propose a graph-based semantic model that integrates more semantic information among keyphrases as well as the structural information of the text. The utilization of large knowledge bases (e.g. DBpedia, Wikipedia) makes available fine-grained information about concepts, entities, and their semantic relations, thus resulting in a knowledge-rich interpretation. The relevance evaluation between two documents can then be performed by calculating the semantic similarity between two keyphrase graphs that represent them. The final result comes close in performance to the specialized black-box methods particularly tuned to this task on a traditional dataset. [ABSTRACT FROM AUTHOR]

The comparison of documents—such as articles or patents search, bibliography recommendations systems, visualization of document collections, etc.—has a wide range of applications in several fields. One of the key tasks that such problems have in common is the evaluation of a similarity metric. Many such metrics have been proposed in the literature. Lately, deep learning techniques have gained a lot of popularity. However, it is difficult to analyze how those metrics perform against each other. In this paper, we present a systematic empirical evaluation of several of the most popular similarity metrics when applied to research articles. We analyze the results of those metrics in two ways, with a synthetic test that uses scientific papers and Ph.D. theses, and in a real-world scenario where we evaluate their ability to cluster papers from different areas of research. [ABSTRACT FROM AUTHOR]

The problem of detecting similar documents plays an essential role for many real-world applications, such as copyright protection and plagiarism detection. To protect data privacy, the new version of such a problem becomes more challenging, where the matched documents are distributed among two or more parties and their privacy should be preserved. In this paper, we propose new privacy-preserving document similarity detection schemes by utilizing the locality-sensitive hashing technique, which can handle the misspelled mistakes. Furthermore, the keywords' occurrences of a given document are integrated into its underlying representation to support a better ranking for the returned results. We introduced a new security definition, which hides the exact similarity scores towards the querying party. Extensive experiments on real-world data illustrate that our proposed schemes are efficient and accurate. [ABSTRACT FROM AUTHOR]

Son yıllarda, farklı konular için sunulan dijital bilgi kaynaklarının sayısı aşırı miktarda artmaktadır. Bu dijital bilgi kaynaklarına erişim desteği sunan sistemlerin birçoğu tarama, arama ve bilgi geri kazanımı araçlarına odaklanmıştır. Sayısal kütüphaneler, elektronik kitaplıklar ve Web sayfaları, bilgi erişimini iyileştirmek, belge koleksiyonlarını farklı anahtar kriterlere göre hiyerarşik olarak oluşturmak ve düzenlemek için yeni birçok açılım sunmaktadır. Farklı arama araçları, bilgi erişim teknikleri kullanılarak erişilebilen belgeleri düzenlemek, endekslemek ve özetlemek için yazılım tabanlı hizmetleri kullanarak daha kapsamlı bir doküman kapsamı sunulabilmektedir. Dijital kütüphanelerdeki arama mekanizmalarına uygulanan teknolojiler, doküman koleksiyonlarını yönetmek, anlamlı veri çıkarmak ve doküman ilişkilerinin belirlenmesi için farklı yöntem ve teknolojilerin kullanımını zorunlu kılmıştır. Özellikle belgeler arasındaki ilişki ne biçimleri ne de türleri ile açıkça tanımlanamamaktadır. Bu çalışma, sayısal kütüphaneler için belgelerin içeriğinden üst-veri çıkarımı, varlık isimlerinin elde edilmesi, anahtar kelimelerin elde erilmesi ve doküman benzerliklerinin oluşturulması için kullanılan yöntem ve teknikler için kapsamlı bir çalışma sunmaktadır.

Background: Clinical trial registries can be used as sources of clinical evidence for systematic review synthesis and updating. Our aim was to evaluate methods for identifying clinical trial registrations that should be screened for inclusion in updates of published systematic reviews.Methods: A set of 4644 clinical trial registrations (ClinicalTrials.gov) included in 1089 systematic reviews (PubMed) were used to evaluate two methods (document similarity and hierarchical clustering) and representations (L2-normalised TF-IDF, Latent Dirichlet Allocation, and Doc2Vec) for ranking 163,501 completed clinical trials by relevance. Clinical trial registrations were ranked for each systematic review using seeding clinical trials, simulating how new relevant clinical trials could be automatically identified for an update. Performance was measured by the number of clinical trials that need to be screened to identify all relevant clinical trials.Results: Using the document similarity method with TF-IDF feature representation and Euclidean distance metric, all relevant clinical trials for half of the systematic reviews were identified after screening 99 trials (IQR 19 to 491). The best-performing hierarchical clustering was using Ward agglomerative clustering (with TF-IDF representation and Euclidean distance) and needed to screen 501 clinical trials (IQR 43 to 4363) to achieve the same result.Conclusion: An evaluation using a large set of mined links between published systematic reviews and clinical trial registrations showed that document similarity outperformed hierarchical clustering for identifying relevant clinical trials to include in systematic review updates. [ABSTRACT FROM AUTHOR]

Recently, similar entity searching over knowledge graph (KG) has gained much attentions by researchers. However, in rich-semantic KGs with multi-typed entities and relations, also known as heterogeneous information network, relevant entity search is considered as a challenging task due to the ambiguity as well as complexity of user's queries in realistic applications, such as QA chatbot and KG-based information retrieval. In this paper, we propose a novel approach, called W-KG2Vec which enables to automatically learn the semantic representations of entities in KG by applying the meta-path. The proposed W-KG2Vec is a meta-path-specific model which supports to evaluate both semantic relations as well as the text-based similarity between entities. The combination of text- and structure-based embedding mechanism of W-KG2Vec is promising to achieve better representations of entities in given KGs for handling complex user's queries. To effectively learn the sequential textual representations of entities' descriptions, we propose a combination of BERT pre-trained model with LTSM encoder, called BERT-Text2Vec. Then, the text-based similarity between entities is used to leverage our weighted meta-path-based random walk mechanism in W-KG2Vec model. Extensive experiences on real-world KGs (YAGO and Freebase) demonstrate the effectiveness of our proposed model against recent state-of-the-art KG embedding baselines. [ABSTRACT FROM AUTHOR]

In today's rushing world, there's an ever-increasing usage of networking equipment. These devices log their operations; however, there could be errors that result in the restart of the given device. There could be different patterns before different errors. Our main goal is to predict the upcoming error based on the log lines of the actual file. To achieve this, we use document similarity. One of the key concepts of information retrieval is document similarity which is an indicator of how analogous (or different) documents are. In this paper, we are studying the effectiveness of prediction based on cosine similarity, Jaccard similarity, and Euclidean distance of rows before restarts. We use different features like TFIDF, Doc2Vec, LSH, and others in conjunction with these distance measures. Since networking devices produce lots of log files, we use Spark for Big data computing. [ABSTRACT FROM AUTHOR]