Inferential representation of science documents

Multiple studies have investigated bibliometric features and uncategorized scholarly documents for the influential scholarly document prediction task. In this paper, we describe our work that attempts to go beyond bibliometric metadata to predict influential scholarly documents. Furthermore, this work also examines the influential scholarly document prediction task over categorized scholarly documents. We also introduce a new approach to enhance the document representation method with a domain-independent knowledge graph to find the influential scholarly document using categorized scholarly content. As the input collection, we use the WHO corpus with scholarly documents on the theme of COVID-19. This study examines different document representation methods for machine learning, including TF-IDF, BOW, and embedding-based language models (BERT). The TF-IDF document representation method works better than others. From various machine learning methods tested, logistic regression outperformed the other for scholarly document category classification, and the random forest algorithm obtained the best results for influential scholarly document prediction, with the help of a domain-independent knowledge graph, specifically DBpedia, to enhance the document representation method for predicting influential scholarly documents with categorical scholarly content. In this case, our study combines state-of-the-art machine learning methods with the BOW document representation method. We also enhance the BOW document representation with the direct type (RDF type) and unqualified relation from DBpedia. From this experiment, we did not find any impact of the enhanced document representation for the scholarly document category classification. We found an effect in the influential scholarly document prediction with categorical data.

Document representation with outlier tokens exacerbates the classification performance due to the uncertain orientation of such tokens. Most existing document representation methods in different languages including Nepali mostly ignore the strategies to filter them out from documents before learning their representations. In this article, we propose a novel document representation method based on a supervised codebook to represent the Nepali documents, where our codebook contains only semantic tokens without outliers. Our codebook is domain-specific as it is based on tokens in a given corpus that have higher similarities with the class labels in the corpus. Our method adopts a simple yet prominent representation method for each word, called probability-based word embedding. To show the efficacy of our method, we evaluate its performance in the document classification task using Support Vector Machine and validate against widely used document representation methods such as Bag of Words, Latent Dirichlet allocation, Long Short-Term Memory, Word2Vec, Bidirectional Encoder Representations from Transformers and so on, using four Nepali text datasets (we denote them shortly as A1, A2, A3 and A4). The experimental results show that our method produces state-of-the-art classification performance (77.46% accuracy on A1, 67.53% accuracy on A2, 80.54% accuracy on A3 and 89.58% accuracy on A4) compared to the widely used existing document representation methods. It yields the best classification accuracy on three datasets (A1, A2 and A3) and a comparable accuracy on the fourth dataset (A4). Furthermore, we introduce the largest Nepali document dataset (A4), called NepaliLinguistic dataset, to the linguistic community.

Evaluation is crucial in assessing the effectiveness of new information retrieval and human computer interaction techniques and systems. Relevance judgements are often performed by humans, which makes obtaining them expensive and time consuming. Consequently, relevance judgements are usually performed only on a subset of a given collection of data or experimental results with a focus on the top ranked documents. However, when assessing the performance of exploratory search systems, the diversity or subjective relevance of documents that the user was presented with over a search session are often of more importance than the relative ranking of top documents. In order to perform these types of assessment, all the documents in a given collection need to be judged for relevance. In this paper, we propose an approach based on topic modeling that can greatly accelerate document relevance judgment of an entire document collection with an expert assessor needing to mark only a small subset of documents from a given collection. Experimental results show a substantial overlap between relevance judgments compared to a human assessor.

Distributed information retrieval methods are growing rapidly because of the rising need to access and search distributed digital documents. However, the content-based information retrieval (CBIR) is concentrated to extract and retrieve the information from massive digital libraries, which require a huge amount of computing and storage resources. The grid computing provides the reliable infrastructure for effective and efficient retrieval on these large collections. In order to build an effective and efficient CBIR technique, varieties of architectures were developed based on grid technologies. The goal of such architecture is to solve interoperability and heterogeneous resource issues, and increase the efficiency and effectiveness of information retrieval (IR) techniques by harnessing the grid computing capabilities. This paper reviews and analyzes latest research carried out in the domain of large-scale dataset IR based on a grid. The evaluation is based on scalability, response time, scope, data type, search technique, middleware, and query type. The contribution is to illustrate the features, capabilities, and shortages of current solutions that can guide the researchers in this evolving area.

Question Answering (QA) is a specialized subfield within Information Retrieval (IR) that focuses on providing precise answers to user queries. Recent advances in Large Language Models (LLMs) have greatly improved the speed and accuracy of question-answering (QA) systems, leading to significantly enhanced capabilities and performance. This study introduces a multi-PDF querying system designed to support automated question answering for Vietnamese students, leveraging the capabilities of LLMs. To enhance the retrieval accuracy from natural language queries, we integrate four Vietnamese-language embedding models. These models are evaluated using a dataset comprising questions related to the academic regulations of the University of Da Nang - University of Science and Technology (DUT), reflecting real-life scenarios students commonly encounter. Experimental results reveal that the top-performing Vietnamese embedding model achieves a Mean Average Precision (mAP) of 0.840 and an F1-score of 0.959, demonstrating its high effectiveness in information retrieval. This study makes significant contributions to the integration of Vietnamese embedding models into automated question-answering systems, aiming to effectively support students in the university environment.

The intention of this paper was to determine usability of KOHA OPAC in information retrieval with regard to usability testing by undergraduate users at Technical University of Kenya Library (TU-KL) in Nairobi County. Usability testing can be realized by way of assigning tasks to a representative of the user population in order to determine whether users can be able to achieve the desired goal of retrieving information. To a great extent, success will depend on how users perform the tasks assigned to them. This study conducted usability test on actual users of KOHA OPAC at Technical University of Kenya Library. Qualitative data was collected using the think aloud method where study participants were required to verbalize their thought process as they performed search and retrieval tasks using KOHA OPAC. The number of errors made, time taken to complete search and retrieval activities were coded, recorded and analysed during usability testing. This helped the researcher to determine whether KOHA OPAC is easy to use and whether it saves time for the user. Undergraduate students for the study were selected using a stratified random sampling technique. The population of the study (n = 382) was determined using a simplified formula developed by Yamane. The study initiated a usability test which presented search scenarios where participants were invited to participate in searching and retrieving information resources using KOHA OPAC. The results of the usability test of KOHA OPAC at TU-KL showed that users were slow at accomplishing search tasks, they took long duration to access TU-KL website that hosts KOHA OPAC and there was inefficiency in accessing library materials. This study concluded that usability challenges may have an impact on users' overall satisfaction with KOHA OPAC and information retrieval effectiveness. In order to enhance user experience and boost the effectiveness of information retrieval, the study recommends improvement of KOHA OPAC usability by revamping its interface and enhancing its features.

This paper presents a vector space model approach to representing documents and queries, which considers concepts instead of terms and uses WordNet as a light ontology. This representation reduces information redundancy with respect to conventional semantic expansion techniques. Experiments carried out on the MuchMore benchmark and on the TREC-7 and TREC-8 Ad-hoc collections demonstrate the effectiveness of the proposed approach.

Recall, the proportion of relevant documents retrieved, is an important measure of effectiveness in information retrieval, particularly in the legal, patent, and medical domains. Where document sets are too large for exhaustive relevance assessment, recall can be estimated by assessing a random sample of documents, but an indication of the reliability of this estimate is also required. In this article, we examine several methods for estimating two-tailed recall confidence intervals. We find that the normal approximation in current use provides poor coverage in many circumstances, even when adjusted to correct its inappropriate symmetry. Analytic and Bayesian methods based on the ratio of binomials are generally more accurate but are inaccurate on small populations. The method we recommend derives beta-binomial posteriors on retrieved and unretrieved yield, with fixed hyperparameters, and a Monte Carlo estimate of the posterior distribution of recall. We demonstrate that this method gives mean coverage at or near the nominal level, across several scenarios, while being balanced and stable. We offer advice on sampling design, including the allocation of assessments to the retrieved and unretrieved segments, and compare the proposed beta-binomial with the officially reported normal intervals for recent TREC Legal Track iterations.

Latent Dirichlet Allocation (LDA), is heavily cited in the machine learning literature, but its feasibility and effectiveness in information retrieval is mostly unknown. Learning to rank is useful for document retrieval, it uses feature vector to rank, but there is no feature about document topic. Our paper combines LDA and learning to rank, adds a topic feature into the feature vector of learning to rank algorithm. And the experiment results show that the topic feature has improved the rank result effectively.

Research articles in biomedicine domain have increased exponentially, which makes it more and more difficult for biologists to manually capture all the information they need. Information retrieval technologies can help to obtain the users' needed information automatically. However, it is a great challenge to apply these technologies to biomedicine domain directly because of some domain specific characteristics, such as the abundance of terminologies. To enhance the effectiveness of the biomedical information retrieval, we propose a novel framework based on the state-of-the-art information retrieval methods, called learning to rank, which has been proved effective to rank documents based on their relevance degree. In the framework, we attempt to tackle the problem of the abundance of terminologies by constructing ranking models, which focus on not only retrieving the most relevant documents but also diversifying the searching results to increase the completeness of the resulting list for a given query. In the model training, we propose two novel document labeling strategies, and combine several traditional retrieval models as learning features. Besides, we also investigate the usefulness of different learning to rank approaches in our framework. Experimental results on TREC Genomics datasets demonstrate our proposed framework is effective in improving the performance of biomedical information retrieval.

Recent studies in information retrieval have shown that gender biases have found their way into representational and algorithmic aspects of computational models. In this paper, we focus specifically on gender biases in information retrieval gold standard datasets, often referred to as relevance judgements. While not explored in the past, we submit that it is important to understand and measure the extent to which gender biases may be presented in information retrieval relevance judgements primarily because relevance judgements are not only the primary source for evaluating IR techniques but are also widely used for training end-to-end neural ranking methods. As such, the presence of bias in relevance judgements would immediately find its way into how retrieval methods operate in practice. Based on a fine-tuned BERT model, we show how queries can be labelled for gender at scale based on which we label MS MARCO queries. We then show how different psychological characteristics are exhibited within documents associated with gendered queries within the relevance judgement datasets. Our observations show that stereotypical biases are prevalent in relevance judgement documents.

The subject of study. The paper focuses on the extraction of semantics from texts in Arabic. In particular, the applicability of the Bell test to word pairs is investigated as a measure of the semantic words relatedness in a context. The study applies the quantum formalism to the task of information retrieval in Arabic texts and presents the results of this work. The authors also examine the influence of the context width on the effectiveness of information retrieval. Method. The research is based on the vector representation of the context. It uses the well-known approach based on the HAL (Hyperspace Analogue to Language) matrix and Bell test. The HAL matrix allows taking into account both the frequency of the words occurrence in the context and the distance to the target word. Quantum theory operates with probability density matrices. Quantum theory allows describing probabilities in the vector space in a more natural way i.e., words can be represented as vectors. Main results. The results demonstrate that using the Bell’s test for texts in Arabic provides a better ranking of search results compared to the results of search services. Practical significance. The research results can be used in the development of the information retrieval systems, as well as for the further development of methods based on the distributive hypothesis.

Search Engine (SE) is the most used information retrieval tool in the present scenario. In spite of the huge involvement of users in search engines, their limited capabilities to understand the user context and emotions places high load on the user to maintain the search momentum. Thus the research is being done on the web search process including the users. The aim is to reduce the contextual and emotional mismatch between the SE's and users. Personalized user query processing, current knowledge updation on users, information retrieval effectiveness and user satisfaction are the major challenges to be addressed in this field. To meet the high level requirements of the users and to improve the intellectualization level of the search, various cognitive agent based SE models are proposed in the literature. The challenges of the cognitive agents in present search engine scenario are obtaining sufficient percepts from the environment, understanding it, gathering the knowledge and retrieving the information from the huge volume of data. Recent advancements in the field of semantic search engines, knowledge gathering techniques and information retrieval methods are discussed which can better the cognitive agent based search engines by addressing the challenges. Overall the goal is to enhance the search experience by fewer repetitions of queries and making it easier for the users to retrieve the highly relevant documents of their interest.

From highly relevant to not relevant: examining different regions of relevance

The task of semantic code search is to retrieve code snippets from a source code corpus based on an information need expressed in natural language. The semantic gap between natural language and programming languages has for long been regarded as one of the most significant obstacles to the effectiveness of keyword-based information retrieval (IR) methods. It is a common assumption that “traditional” bag-of-words IR methods are poorly suited for semantic code search: our work empirically investigates this assumption. Specifically, we examine the effectiveness of two traditional IR methods, namely BM25 and RM3, on the CodeSearchNet Corpus, which consists of natural language queries paired with relevant code snippets. We find that the two keyword-based methods outperform several pre-BERT neural models. We also compare several code-specific data pre-processing strategies and find that specialized tokenization improves effectiveness.