Article-Journal

GCAM has annotators report which guidelines apply to each text instead of a label, so adherence becomes checkable and one annotation serves many tasks.

Over three years, 244 master students built AI players for the board game Tablut, and they report gains in AI knowledge, motivation, and teamwork.

A checklist for responsible speech datasets on mental health and neurological disorders shows that 36 dataset papers rarely report privacy and security measures.

With good machine translation, an English unfair-clause detector for Terms of Service matches detectors trained on new German, Italian, and Polish corpora.

Detecting disruptive situations on public transport by recognizing anger, fear, sadness, and disgust in speech, with small models that stay accurate in noise.

A dialogue system that keeps personal data in a language module and picks replies by argumentation over sanitized facts, so answers are defended and explainable.

An argument mining tool scores how much COVID-19 papers argue, and these scores agree fairly to moderately with Cochrane decisions and expert reviewers.

Memory networks read a knowledge base of expert legal rationales to detect unfair clauses in terms of service and to point to the reasons.

A unified model and the first taxonomy of attention in natural language processing, charting a fast-growing literature along four dimensions and its open challenges.