Logo des Repositoriums

Auflistung nach Schlagwort "Question generation"

1 - 2 von 2
  • Konferenzbeitrag
    Educational Text Summarizer: Which sentences are worth asking for?
    (DELFI 2020 – Die 18. Fachtagung Bildungstechnologien der Gesellschaft für Informatik e.V., 2020) Rüdian, Sylvio; Heuts, Alexander; Pinkwart, Niels
    Many question generation approaches focus on the generation process itself, but they work with single sentences as input only. Although the state of the art of question generation’s results is quite good, it cannot be used practically as the selection which sentences are worth asking for in an educational setting is currently not possible in an automated way. This limits the ability to generate interactive course materials at scale. In this paper, we conduct a study where we compare teachers’ sentence selections of texts with 9 algorithms to find the most appropriate ones concerning reading comprehension. 30 teachers compared the “winner” algorithm, Edmundson with LexRank, which was found to be the optimal algorithm according to previous literature. The result shows that Edmundson outperforms LexRank.
  • Konferenzbeitrag
    Is the context-based Word2Vec representation useful to determine Question Words for Generators?
    (DELFI 2020 – Die 18. Fachtagung Bildungstechnologien der Gesellschaft für Informatik e.V., 2020) Rüdian, Sylvio; Pinkwart, Niels
    Question and answer generation approaches focus on the quality and correctness of generated questions for online courses but miss to use a good question word, which is a deficiency reported by many previous studies. In this experimental setup, we explored whether the word2vec representation, which is semantic-based, can be used to predict question words. We compare two pipelines of the prediction process and observed that splitting the problem into several subproblems performs similar to feeding a neural network with all the data. Although our approach is promising to take the context-based representation into account we can see that the success rate is still low but better than guessing.