Neural information retrieval: at the end of the early years

A recent “third wave” of neural network (NN) approaches now delivers state-of-the-art performance in many machine learning tasks, spanning speech recognition, computer vision, and natural language processing. Because these modern NNs often comprise multiple interconnected layers, work in this area is often referred to as deep learning. Recent years have witnessed an explosive growth of research into NN-based approaches to information retrieval (IR). A significant body of work has now been created. In this paper, we survey the current landscape of Neural IR research, paying special attention to the use of learned distributed representations of textual units. We highlight the successes of neural IR thus far, catalog obstacles to its wider adoption, and suggest potentially promising directions for future research.     

The rise of the machines: Artificial intelligence meets scholarly content

Key points

• New forms of human/machine dialogue are emerging as robots understand vast amounts of content rather than simply indexing content as strings of characters.
• Recognizing strings of characters as entities (e.g. = names = authors) allows for meaningful associations between entities and reasoning over these relationships.
• Web‐scale adoption of the Semantic Web approach has been slow because it is too complex to implement and does not scale.
• User intent, discovered through conversational models of human–computer interaction, allows for a deeper understanding of exactly what researchers are looking for.
• Personal agents hold the promise of finding information that we will find useful before we have started to look for it.
• Publishers can use Academic Knowledge APIs to interpret academic user queries and find rich information from the Microsoft Academic Graph.

     

Look who’s talking: Emergent evidence for discriminating between differences in listserv participation

This paper reports findings from a study of LookJed, the oldest and largest on-line forum for Computer Mediated Discussion among individuals interested in Jewish education. The study adopted a “cyber-ethnographic” approach, with postings to the forum seen as “acts of communication” that reveal what is important to their authors. An interest in exploring similarities between forum conversations and those in teachers’ lounges led to an investigation of Herring’s claim that most listservs do not include discussion at all, only the trading of information. This investigation found that active forum participants generally use the forum for discrete purposes, most commonly to exchange information about “subject matter” or “teaching material”, less commonly to exchange opinions and ideas, and rarely to do both. Integrating an analysis of patterns of contribution with an examination of their discursive content reveals six preeminent “types” among the population of contributors, each of whom participates in the forum in different ways and acts with different purposes. Although this typology is at best suggestive and needs to be tested against other listserv cases, its easy identification suggests that in order to better understand the cultures of virtual forums, it is important to pursue a more variegated characterization of listserv participants and their motivations than has typically been the case in CMD research where users are most frequently identified as either lurkers or fanatics, or as active or passive participants.     

Platform learning and on-demand labor: sociotechnical projections on the future of education and work

Platform learning harnesses the operating capabilities and logics of digital platforms such as Uber and Amazon to imagine synergies between on-demand labor and on-demand learning, transforming living into learning, and learning into labor. This paper seeks to make three original contributions to critical analysis of platform learning. First, as an analytical foundation, it brings together two distinct strands of scholarship on the evolving relationship between learning and late capitalism, and the digitalization of education policy and governance, synthesizing them in relation to questions concerning labor and work in the emergent on-demand economy. Second, it draws on these ideas to engage the learning and work projections of two strategic forecasting organizations, Institute for the Future and Knowledge Works, as case studies of platform learning. Third, the last section of the paper builds on the sociotechnical projections of these organizations as the basis for a critique of the political economy of platform learning, highlighting four areas requiring further inquiry: (1) value extraction; (2) exploitation of labor; (3) efficacy and inequality; (4) imagination.