David M. Schmidt

@dmschmidt.bsky.social

116 Followers 679 Following

#NLProc PhD Student & Research Associate at Bielefeld University Working on: Question Answering over Linked Data, Semantic Web, Lexical Knowledge & Compositionality in AI https://davidmschmidt.de

International Semantic Web Research Summer School 2025 group photo

Illustration of the baseline model as well as the two adjustments added to that baseline, grammar-constrained decoding and pointer generator-like behavior. Words in boxes represent single tokens, numbers below those boxes symbolize outputs from the decoder, where higher values stand for a higher probability that this is the best next token as estimated by the model. For greedy decoding, the token with the highest value is chosen. For GCD, a filter is applied before, visualized as gray, crossed-out boxes for tokens that are filtered out. Red boxes show the selected token. (A) Greedy decoding (baseline, basic). (B) Grammar-constrained decoding (GCD). (C) Pointer generators + grammar-constrained decoding (ptr).

David M. Schmidt giving a talk on the paper "Lexicalization Is All You Need: Examining the Impact of Lexical Knowledge in a Compositional QALD System" at the 24th International Conference on Knowledge Engineering and Knowledge Management in Amsterdam.