International Journal on Digital Libraries最新文献

The significance of the web and the crucial role of web archives in its preservation highlight the necessity of understanding how users, both human and robot, access web archive content, and how best to satisfy this disparate needs of both types of users. To identify robots and humans in web archives and analyze their respective access patterns, we used the Internet Archive’s (IA) Wayback Machine access logs from 2012, 2015, and 2019, as well as Arquivo.pt’s (Portuguese Web Archive) access logs from 2019. We identified user sessions in the access logs and classified those sessions as human or robot based on their browsing behavior. To better understand how users navigate through the web archives, we evaluated these sessions to discover user access patterns. Based on the two archives and between the three years of IA access logs (2012 vs. 2015 vs. 2019), we present a comparison of detected robots vs. humans and their user access patterns and temporal preferences. The total number of robots detected in IA 2012 (91% of requests) and IA 2015 (88% of requests) is greater than in IA 2019 (70% of requests). Robots account for 98% of requests in Arquivo.pt (2019). We found that the robots are almost entirely limited to “Dip” and “Skim” access patterns in IA 2012 and 2015, but exhibit all the patterns and their combinations in IA 2019. Both humans and robots show a preference for web pages archived in the near past.

This study focuses on a method for differentiating between the stance of citizens and city councilors on political issues (i.e., in favor or against) and attempts to compare the arguments of both sides. We created a dataset by annotating citizen tweets and city council minutes with labels for four attributes: stance, usefulness, regional dependence, and relevance. We then fine-tuned pretrained large language model using this dataset to assign the attribute labels to a large quantity of unlabeled data automatically. We introduced multitask learning to train each attribute jointly with relevance to identify the clues by focusing on those sentences that were relevant to the political issues. Our prediction models are based on T5, a large language model suitable for multitask learning. We compared the results from our system with those that used BERT or RoBERTa. Our experimental results showed that the macro-F1-scores for stance were improved by 1.8% for citizen tweets and 1.7% for city council minutes with multitask learning. Using the fine-tuned model to analyze real opinion gaps, we found that although the vaccination regime was positively evaluated by city councilors in Fukuoka city, it was not rated very highly by citizens.

This paper examines how privacy measures, such as anonymisation and aggregation processes for email collections, can affect the perceived usefulness of email visualisations for research, especially in the humanities and social sciences. The work is intended to inform archivists and data managers who are faced with the challenge of accessioning and reviewing increasingly sizeable and complex personal digital collections. The research in this paper provides a focused user study to investigate the usefulness of data visualisation as a mediator between privacy-aware management of data and maximisation of research value of data. The research is carried out with researchers and archivists with vested interest in using, making sense of, and/or archiving the data to derive meaningful results. Participants tend to perceive email visualisations as useful, with an average rating of 4.281 (out of 7) for all the visualisations in the study, with above average ratings for mountain graphs and word trees. The study shows that while participants voice a strong desire for information identifying individuals in email data, they perceive visualisations as almost equally useful for their research and/or work when aggregation is employed in addition to anonymisation.

Web archives capture, retain, and present historical versions of web pages. Viewing web archives often amounts to a user visiting the Wayback Machine homepage, typing in a URL, then choosing a date and time significant of the capture. Other web archives also capture the web and use Memento as an interoperable point of querying their captures. Memento aggregators are web accessible software packages that allow clients to send requests for past web pages to a single endpoint source that then relays that request to a set of web archives. Though few deployed aggregator instances exist that exhibit this aggregation trait, they all, for the most part, align to a model of serving a request for a URI of an original resource (URI-R) to a client by first querying then aggregating the results of the responses from a collection of web archives. This single tier querying need not be the logical flow of an aggregator, so long as a user can still utilize the aggregator from a single URL. In this paper, we discuss theoretical aggregation models of web archives. We first describe the status quo as the conventional behavior exhibited by an aggregator. We then build on prior work to describe a multi-tiered, structured querying model that may be exhibited by an aggregator. We highlight some potential issues and high-level optimization to ensure efficient aggregation while also extending on the state-of-the-art of memento aggregation. Part of our contribution is the extension of an open-source, user-deployable Memento aggregator to exhibit the capability described in this paper. We also extend a browser extension that typically consults an aggregator to have the ability to aggregate itself rather than needing to consult an external service. A purely client-side, browser-based Memento aggregator is novel to this work.

Research in Natural Language Processing (NLP) is increasing rapidly; as a result, a large number of research papers are being published. It is challenging to find the contributions of the research paper in any specific domain from the huge amount of unstructured data. There is a need for structuring the relevant contributions in Knowledge Graph (KG). In this paper, we describe our work to accomplish four tasks toward building the Scientific Knowledge Graph (SKG). We propose a pipelined system that performs contribution sentence identification, phrase extraction from contribution sentences, Information Units (IUs) classification, and organize phrases into triplets (subject, predicate, object) from the NLP scholarly publications. We develop a multitasking system (ContriSci) for contribution sentence identification with two supporting tasks, viz. Section Identification and Citance Classification. We use the Bidirectional Encoder Representations from Transformers (BERT)—Conditional Random Field (CRF) model for the phrase extraction and train with two additional datasets: SciERC and SciClaim. To classify the contribution sentences into IUs, we use a BERT-based model. For the triplet extraction, we categorize the triplets into five categories and classify the triplets with the BERT-based classifier. Our proposed approach yields the F1 score values of 64.21%, 77.47%, 84.52%, and 62.71% for the contribution sentence identification, phrase extraction, IUs classification, and triplet extraction, respectively, for non-end-to-end setting. The relative improvement for contribution sentence identification, IUs classification, and triplet extraction is 8.08, 2.46, and 2.31 in terms of F1 score for the NLPContributionGraph (NCG) dataset. Our system achieves the best performance (57.54% F1 score) in the end-to-end pipeline with all four sub-tasks combined. We make our codes available at: https://github.com/92Komal/pipeline_triplet_extraction.

The automatic semantic structuring of scientific text allows for more efficient reading of research articles and is an important indexing step for academic search engines. Sequential sentence classification is an essential structuring task and targets the categorisation of sentences based on their content and context. However, the potential of transfer learning for sentence classification across different scientific domains and text types, such as full papers and abstracts, has not yet been explored in prior work. In this paper, we present a systematic analysis of transfer learning for scientific sequential sentence classification. For this purpose, we derive seven research questions and present several contributions to address them: (1) We suggest a novel uniform deep learning architecture and multi-task learning for cross-domain sequential sentence classification in scientific text. (2) We tailor two transfer learning methods to deal with the given task, namely sequential transfer learning and multi-task learning. (3) We compare the results of the two best models using qualitative examples in a case study. (4) We provide an approach for the semi-automatic identification of semantically related classes across annotation schemes and analyse the results for four annotation schemes. The clusters and underlying semantic vectors are validated using k-means clustering. (5) Our comprehensive experimental results indicate that when using the proposed multi-task learning architecture, models trained on datasets from different scientific domains benefit from one another. Our approach significantly outperforms state of the art on full paper datasets while being on par for datasets consisting of abstracts.

Reading Lists Systems are widely used in tertiary education as a pedagogical tool and for tracking copyrighted material. This paper explores academics' experiences with reading lists and in particular the use of reading lists notes feature. A mixed-methods approach was employed in which we first conducted interviews with academics about their experience with reading lists. We identified the need for streamlining the workflow of the reading lists set-up, improved usability of the interfaces, and better synchronization with other teaching support systems. Next, we performed a log analysis of the use of the notes feature throughout one academic year. The results of our log analysis were that the note feature is under-utilized by academics. We recommend improving the systems’ usability by re-engineering the user workflows and to better integrate notes feature into academic teaching.

Detecting texts that contain semantic-level new information is not straightforward. The problem becomes more challenging for research articles. Over the years, many datasets and techniques have been developed to attempt automatic novelty detection. However, the majority of the existing textual novelty detection investigations are targeted toward general domains like newswire. A comprehensive dataset for scientific novelty detection is not available in the literature. In this paper, we present a new triplet-based corpus (SciND) for scientific novelty detection from research articles via knowledge graphs. The proposed dataset consists of three types of triples (i) triplet for the knowledge graph, (ii) novel triplets, and (iii) non-novel triplets. We build a scientific knowledge graph for research articles using triplets across several natural language processing (NLP) domains and extract novel triplets from the paper published in the year 2021. For the non-novel articles, we use blog post summaries of the research articles. Our knowledge graph is domain-specific. We build the knowledge graph for seven NLP domains. We further use a feature-based novelty detection scheme from the research articles as a baseline. Moreover, we show the applicability of our proposed dataset using our baseline novelty detection algorithm. Our algorithm yields a baseline F1 score of 72%. We show analysis and discuss the future scope using our proposed dataset. To the best of our knowledge, this is the very first dataset for scientific novelty detection via a knowledge graph. We make our codes and dataset publicly available at https://github.com/92Komal/Scientific_Novelty_Detection.

Abstract

Every researcher must conduct a literature review, and the document management needs of researchers working on various research topics vary. However, there are two major challenges. First, traditional methods such as the tree hierarchy of document folders and tag-based management are no longer effective with the enormous volume of publications. Second, although their bibliographic information is available to everyone, many papers can only be accessed through paid services. This study attempts to develop an interactive tool for personal literature management based solely on their bibliographic records. To make such a tool possible, we developed a principled “human-in-the-loop latent space learning” method that estimates the management criteria of each researcher based on his or her feedback to calculate the positions of documents in a two-dimensional space on the screen. As a set of bibliographic records forms a graph, our model is naturally designed as a graph-based encoder–decoder model that connects the graph and the space. In addition, we also devised an active learning framework using uncertainty sampling for it. The challenge here is to define the uncertainty in a problem setting. Experiments with ten researchers from the humanities, science, and engineering domains show that the proposed framework provides superior results to a typical graph convolutional encoder–decoder model. In addition, we found that our active learning framework was effective in selecting good samples.