Two Knowledge-based Methods for High-Performance Sense Distribution Learning
Authors: Tommaso Pasini, Roberto Navigli
AAAI 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Intrinsic and extrinsic evaluations show that our methods outperform the current state of the art in sense distribution learning and the strongest baselines for the most frequent sense in multiple languages and on domain-specific test sets. |
| Researcher Affiliation | Academia | Tommaso Pasini, Roberto Navigli {pasini,navigli}@di.uniroma1.it |
| Pseudocode | No | The paper describes procedures using mathematical formulas and descriptive text but does not include any clearly labeled 'Pseudocode' or 'Algorithm' blocks. |
| Open Source Code | No | The paper states: 'Our sense distributions are available at http://trainomatic.org.' and 'Our data is available at http://trainomatic.org.' However, this link provides access to sense distributions and data, not the source code for the methodology described in the paper. |
| Open Datasets | Yes | We chose Wikipedia as our input corpus because it is available in hundreds of languages and it covers all domains of human knowledge. We used the October 2014 dump of Wikipedia. |
| Dataset Splits | Yes | Entropy threshold. For En Di we tried different values of the threshold θ, ranging from 0.1 to 4.0 with step 0.1, and tested the results on an in-house development set of 25 lemmas for which we computed the sense distribution in Wikipedia and then selected the value of θ based on the best results in terms of similarity to the Sem Cor distribution (as explained below). We thus set θ to 1.0. |
| Hardware Specification | No | No specific hardware details (such as GPU/CPU models, processor types, or memory amounts) used for running the experiments are provided. |
| Software Dependencies | No | The paper does not provide specific software dependencies with version numbers (e.g., library or solver names with version numbers like Python 3.8, CPLEX 12.4) needed to replicate the experiment. |
| Experiment Setup | Yes | Entropy threshold. For En Di we tried different values of the threshold θ, ranging from 0.1 to 4.0 with step 0.1, and tested the results on an in-house development set... We thus set θ to 1.0. ... We set the threshold θ for Italian and Spanish to 1.0 and 0.001 experimentally... |