Année : 2019
Lieu de publication de l'article :

Résumé de l'article

Analysis of student writing, both for assessment and for enablingfeedback have been of interest to the field of learning analytics.While much progress can be made through detection of local cuesin writing, structured prediction approaches offer capabilities thatare particularly well tailored to the needs of models aiming to of-fer substantive feedback on rhetorical structure. We thus cast theanalysis of rhetorical structure in academic writing as a structuredprediction task in which we employ models that leverage both localand global cues in writing. In particular, this paper presents a hier-archical neural architecture that performs this task. The evaluationdemonstrates that the architecture achieves near-human perfor-mance while significantly surpassing state-of-the-art baselines. Amultifaceted approach to model interpretation offers insights intothe inner workings of the model.CCS CONCEPTS•Computingmethodologies→Discourse, dialogue andprag-matics; • Applied computing→ Computer-assisted instruction; •Human-centered computing → Visualization techniques;KEYWORDSRhetorical structure, neural sequence model, automatic essay evalu-ation, writing feedback, neural network interpretation, hierarchical,bidirectional LSTM, conditional random fieldACM Reference Format:James Fiacco, Elena Cotos, and Carolyn Rosé. 2019. Towards Enabling Feed-back on Rhetorical Structure with Neural Sequence Models. In The 9thInternational Learning Analytics & Knowledge Conference (LAK19), March4–8, 2019, Tempe, AZ, USA. ACM, New York, NY, USA, 10 pages. https://doi.org/10.1145/3303772.3303808

Mots-clés

Baseline (configuration management),Structured prediction,Mechatronics,Human reliability,