FACES: Diversity-Aware Entity Summarization Using Incremental Hierarchical Conceptual Clustering
Authors: Kalpa Gunaratna, Krishnaparasad Thirunarayan, Amit Sheth
AAAI 2015 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We evaluate our approach against the state-of-the-art techniques and show that our work improves both the quality and the efficiency of entity summarization. We evaluate FACES against RELIN that outperformed earlier entity summarization tools and SUMMARUM which is DBpedia specific. We created a gold standard for the evaluation due to unavailability of the evaluation data of RELIN (as confirmed by the authors of RELIN). We randomly selected 50 entities from DBpedia (English version 3.9) that have at least 17 distinct properties per entity. Our evaluation results and statistics are presented in Table 1. |
| Researcher Affiliation | Academia | Kalpa Gunaratna, Krishnaprasad Thirunarayan and Amit Sheth Kno.e.sis, Wright State University, Dayton OH, USA {kalpa, tkprasad, amit}@knoesis.org |
| Pseudocode | No | The paper describes algorithms in text but does not include structured pseudocode or algorithm blocks. |
| Open Source Code | No | The paper states: 'More details about the approach and gold standard dataset can be found at our web page4.' with footnote 4 being 'http://wiki.knoesis.org/index.php/FACES'. This is a general web page and does not explicitly state that the source code for the methodology is provided or linked. |
| Open Datasets | Yes | We selected the DBpedia dataset for our evaluation as it was the benchmark dataset selected in (Cheng, Tran, and Qu 2011) and contains entities that belong to different domains. |
| Dataset Splits | No | The paper describes the creation of a 'gold standard' for evaluation purposes, which involved human judges selecting summaries, but it does not specify traditional training, validation, and test splits for model development as the approach is algorithmic and not a learned model in the typical sense. |
| Hardware Specification | Yes | All experiments were performed using a Core i7 3.4 GHz Desktop machine with 12 GB of RAM. |
| Software Dependencies | No | The paper mentions external services like 'Sindice API' but does not specify the programming languages, libraries, or frameworks with their version numbers used for the implementation of their approach. |
| Experiment Setup | Yes | We empirically determined to cut FACES cluster hierarchies at level 3 which gave good results. We also set the cut-off threshold of Cobweb to 5, which gave the optimal results. For RELIN, we set the jump probability and number of iterations to 0.85 and 10, respectively. |