Learning Hostname Preference to Enhance Search Relevance
Authors: Jingjing Wang, Changsung Kang, Yi Chang, Jiawei Han
IJCAI 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Evaluation of the learned hostname preference is performed both intrinsically on test errors, and extrinsically on the impact on search ranking relevance. Experimental results demonstrate that capturing hostname preference can significantly boost the retrieval performance. |
| Researcher Affiliation | Collaboration | 1University of Illinois at Urbana-Champaign {jwang112, hanj}@illinois.edu 2Yahoo! Labs {ckang, yichang}@yahoo-inc.com |
| Pseudocode | Yes | Algorithm 1 A Three-Party ALS Algorithm for Feature Aware Matrix Completion |
| Open Source Code | No | The paper uses Spark MLlib and provides a link to its guide (http://spark.apache.org/docs/latest/mllib-guide.html), but does not state that the code for the described methodology is open-source or provide a link to its own implementation. |
| Open Datasets | No | The paper uses 'query logs for general search (as opposed to vertical search) from Yahoo! search engine over one year' but does not provide concrete access information like a link, DOI, or a citation to a publicly available dataset. |
| Dataset Splits | No | The paper states: 'Then we randomly assign 80% of this confident dataset to the training set and the remaining 20% to the test set.' It describes a train/test split but does not specify a validation split. |
| Hardware Specification | No | The paper states: 'The algorithm is run on a commercial Hadoop cluster with normal amount of traffic. We use 20 worker nodes and 10 iterations. The maximum memory that can be used at each node is set to 4G.' This describes the computing environment and memory per node but does not provide specific CPU or GPU models. |
| Software Dependencies | No | The paper mentions building its algorithm on 'Spark MLlib' but does not provide specific version numbers for Spark MLlib or any other key software dependencies. |
| Experiment Setup | Yes | The regularization parameter λq i is set to be proportional to the size of Cq j proportional to Ch j following [Zhou et al., 2008]. With this scaling we found the model is not sensitive to the regularization. For the results reported in what follows, λq i = 0.01|Cq j = 0.01|Ch j | and λw = 0.01. |