What do You Mean? Interpreting Image Classification with Crowdsourced Concept Extraction and Analysis

Global interpretability is a vital requirement for image classification applications. Existing interpretability methods mainly explain a model behavior by identifying salient image patches, which require manual efforts from users to make sense of, and also do not typically support model validation with questions that investigate multiple...

Hybrid Annotation Systems for Music Transcription

Automated methods and human annotation are being extensively utilized to scale up modern classification systems. Processes though such as music transcription, oppose certain challenges due to the complexity of the domain and the expertise needed to read and process music scores. In this work, we examine how music transcription could benefit from...

SmartPub: A Platform for Long-Tail Entity Extraction from Scientific Publications

This demo presents SmartPub, a novel web-based platform that supports the exploration and visualization of shallow meta-data (e.g., author list, keywords) and deep meta-data--long tail named entities which are rare, and often relevant only in specific knowledge domain--from scientific publications. The platform collects documents from different...

Microtask crowdsourcing for music score Transcriptions: an experiment with error detection

Human annotation is still an essential part of modern transcription workflows for digitizing music scores, either as a standalone approach where a single expert annotator transcribes a complete score, or for supporting an automated Optical Music Recognition (OMR) system. Research on human computation has shown the effectiveness of crowdsourcing...

How can Explainability Methods be Used to Support Bug Identification in Computer Vision Models?

Deep learning models for image classification suffer from dangerous issues often discovered after deployment. The process of identifying bugs that cause these issues remains limited and understudied. Especially, explainability methods are often presented as obvious tools for bug identification. Yet, the current practice lacks an understanding...

Coner: A Collaborative Approach for Long-Tail Named Entity Recognition in Scientific Publications

Named Entity Recognition (NER) for rare long-tail entities as e.g., often found in domain-specific scientific publications is a challenging task, as typically the extensive training data and test data for fine-tuning NER algorithms is lacking. Recent approaches presented promising solutions relying on training NER algorithms in an iterative...

Training Data Augmentation for Detecting Adverse Drug Reactions in User-Generated Content

Social media provides a timely yet challenging data source for adverse drug reaction (ADR) detection. Existing dictionary-based, semi-supervised learning approaches are intrinsically limited by the coverage and maintainability of laymen health vocabularies. In this paper, we introduce a data augmentation approach that leverages variational...

Scriptoria: A Crowd-powered Music Transcription System

In this demo we present Scriptoria, an online crowdsourcing system to tackle the complex transcription process of classical orchestral scores. The system’s requirements are based on experts’ feedback from classical orchestra members. The architecture enables an end- to-end transcription process (from PDF to MEI) using a scalable microtask design...

TSE-NER: An Iterative Approach for Long-Tail Entity Extraction in Scientific Publications

Named Entity Recognition and Typing (NER/NET) is a challenging task, especially with long-tail entities such as the ones found in scientific publications. These entities (e.g. “WebKB”, “StatSnowball”) are rare, often relevant only in specific knowledge domains, yet important for retrieval and exploration purposes. State-of-the-art NER approaches...

Concept Focus: Semantic Meta-Data For Describing MOOC Content

MOOCs promised to herald a new age of open education.<br/>However, efficient access to MOOC content is still hard, thus unneces-<br/>sarily complicating many use cases like efficient re-use of material, or<br/>tailored access for life-long learning scenarios. One of the reasons for this<br/>lack of accessibility is the shortage of meaningful...