Search for: All records

Developing a universal model that can efficiently and effectively respond to a wide range of information access requests-from retrieval to recommendation to question answering--has been a long-lasting goal in the information retrieval community. This paper argues that the flexibility, efficiency, and effectiveness brought by the recent development in dense retrieval and approximate nearest neighbor search have smoothed the path towards achieving this goal. We develop a generic and extensible dense retrieval framework, called framework, that can handle a wide range of (personalized) information access requests, such as keyword search, query by example, and complementary item recommendation. Our proposed approach extends the capabilities of dense retrieval models for ad-hoc retrieval tasks by incorporating user-specific preferences through the development of a personalized attentive network. This allows for a more tailored and accurate personalized information access experience. Our experiments on real-world e-commerce data suggest the feasibility of developing universal information access models by demonstrating significant improvements even compared to competitive baselines specifically developed for each of these individual information access tasks. This work opens up a number of fundamental research directions for future exploration. 

This tutorial will provide an overview of recent advances on neuro- symbolic approaches for information retrieval. A decade ago, knowl- edge graphs and semantic annotations technology led to active research on how to best leverage symbolic knowledge. At the same time, neural methods have demonstrated to be versatile and highly effective. From a neural network perspective, the same representation approach can service document ranking or knowledge graph rea- soning. End-to-end training allows to optimize complex methods for downstream tasks. We are at the point where both the symbolic and the neural research advances are coalescing into neuro-symbolic approaches. The underlying research questions are how to best combine sym- bolic and neural approaches, what kind of symbolic/neural ap- proaches are most suitable for which use case, and how to best integrate both ideas to advance the state of the art in information retrieval. Materials are available online: https://github.com/laura-dietz/ neurosymbolic-representations-for-IR 

We introduce a new design method to tailor the physical structure of a powered ankle-foot orthosis to the wearer’s leg morphology and improve fit. We present a digital modeling and fabrication workflow that combines scan-based design, parametric configurable modeling, and additive manufacturing (AM) to enable the efficient creation of personalized ankle-foot orthoses with minimal lead-time and explicit inputs. The workflow consists of an initial one-time generic modeling step to generate a parameterized design that can be rapidly configured to customizable shapes and sizes using a design table. This step is then followed by a wearer-specific personalization step that consists of performing a 3D scan of the wearer’s leg, extracting key parameters of the wearer’s leg morphology, generating a personalized design using the configurable parametric design, and digital fabrication of the individualized ankle-foot orthosis using additive manufacturing. The paper builds upon the design of the Stevens Ankle-Foot Electromechanical (SAFE) orthosis presented in prior work and introduces a new, individualized structural design (SAFE II orthosis) that is modeled and fabricated using the presented digital workflow. The workflow is demonstrated by designing a personalized ankle-foot orthosis for an individual based on 3D scan data and printing a personalized design to perform preliminary fit testing. Implications of the presented methodology for the design and fabrication of future personalized powered orthoses are discussed, along with avenues for future work.