Generating Character Descriptions for Automatic Summarization of Fiction
Authors: Weiwei Zhang, Jackie Chi Kit Cheung, Joel Oren7476-7483
AAAI 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We propose two approaches to generate character descriptions, one based on ranking attributes found in the story text, the other based on classifying into a list of pre-defined attributes. We find that the classification-based approach performs the best in predicting character descriptions. |
| Researcher Affiliation | Collaboration | Weiwei Zhang,1 Jackie Chi Kit Cheung,1,2 Joel Oren3 1Mc Gill University, Montreal, Canada 2Mila, Montreal, Canada 3Yahoo! Research, Haifa, Israel |
| Pseudocode | No | The paper does not contain any structured pseudocode or algorithm blocks. |
| Open Source Code | No | The paper does not provide an explicit statement about the release of its source code or a link to a code repository for the methodology described. |
| Open Datasets | No | We obtained a dataset2 of stories and author-written summaries from Wattpad, a popular online story sharing community. This dataset contains 1,036,965 stories and 942,218 summaries provided by authors. Wattpad offers the dataset under a non-commercial academic licence. (This indicates it's not generally publicly available for all users.) |
| Dataset Splits | Yes | For the abstractive models... we picked the model (5 epochs) with the best performance on the validation data (10% of the automatically extracted classification set). |
| Hardware Specification | No | The paper does not provide specific hardware details (e.g., GPU/CPU models, memory amounts) used for running its experiments. |
| Software Dependencies | No | The paper mentions software like Book NLP, GBDT, word2vec, SVM, LSTM, and FNN but does not provide specific version numbers for any of them. |
| Experiment Setup | Yes | For the extractive models, the hyperparameters (the learning rate α, the number of boosting stages n estimators and the maximum depth of a tree max depth) of GBDT were tuned on a grid of values ( α [0.01, 0.05, 0.1, 0.15, 0.2, 0.3, 0.4, 0.5], n estimators [40, 60, 80, 100, 120, 140, 160 , 180, 200] and max depth [3, 5, 7, 9, 11])... GBDT was trained on the entire automatically extracted ranking set with α = 0.1, n estimators = 100 and max depth = 5. For the abstractive models... trained for up to 100 epochs... The output of LSTM has 64 dimensions and FNN contains two hidden layers with 64 hidden nodes. For both models, the word embedding size is set to 64. |