A Morphology-Aware Network for Morphological Disambiguation
Authors: Eray Yildiz, Caglar Tirkaz, H. Sahin, Mustafa Eren, Omer Sonmez
AAAI 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | In the experiments, we achieve 84.12 , 88.35 and 93.78 morphological disambiguation accuracy among the ambiguous words for Turkish, German and French respectively. |
| Researcher Affiliation | Industry | Eray Yildiz, Caglar Tirkaz, H. Bahadir Sahin, Mustafa Tolga Eren, Ozan Sonmez Huawei Turkey Research and Development Center, Umraniye, Istanbul, Turkey {eray.yildiz, mustafa.tolga.eren}@huawei.com {caglartirkaz, hbahadirsahin, osonmez}@gmail.com |
| Pseudocode | No | No pseudocode or algorithm blocks found. |
| Open Source Code | No | We make this test data publicly available 1 so that Turkish morphological disambiguation algorithms can be compared more accurately in the future. |
| Open Datasets | Yes | For Turkish, we used a semi-automatically disambiguated corpus containing 1M tokens (Y uret and T ure 2006). ... We use SPMRL 2014 dataset (Seddah and Tsarfaty 2014) for German and French. |
| Dataset Splits | Yes | It provides 90% of all sentences as training set and %10 of rest of the sentences as test set. ... The development sets for each language are randomly separated from the training data and are used to optimize the embedding lengths of morphological features. |
| Hardware Specification | No | No specific hardware details (e.g., GPU/CPU models, memory) are provided. |
| Software Dependencies | No | HFST tool (Lind en, Silfverberg, and Pirinen 2009) is used to perform morphological analysis in German and French whereas (Oflazer 1993) is used for Turkish morphological analysis. No specific version numbers for these tools or other software dependencies are provided. |
| Experiment Setup | Yes | Thus, in the experiments, we used embedding lengths 50, 20 and 5 for roots, POS tags and the other morphological features respectively. The number of filters in the first and second layers are 30 and 40 respectively. The window length, n, that determines the number of words input to the second layer is set to 5. Training is performed with stochastic gradient descent and Ada Grad (Duchi, Hazan, and Singer 2011) as the optimization algorithms. |