DeepTileBars: Visualizing Term Distribution for Neural Information Retrieval
Authors: Zhiwen Tang, Grace Hui Yang289-296
AAAI 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Deep Tile Bars outperforms other state-of-the-art Neu-IR models on benchmark datasets including the Text REtrieval Conference (TREC) 2010-2012 Web Tracks and LETOR 4.0. |
| Researcher Affiliation | Academia | Zhiwen Tang, Grace Hui Yang Info Sense, Department of Computer Science Georgetown University zt79@georgetown.edu,huiyang@cs.georgetown.edu |
| Pseudocode | No | The paper describes the network architecture with equations and diagrams but does not provide pseudocode or algorithm blocks. |
| Open Source Code | No | The paper does not provide any links to source code or explicitly state that the code for their method is open-source or publicly available. |
| Open Datasets | Yes | We conduct experiments on the TREC 2010-2012 Web Track Ad-hoc tasks (Clarke, Craswell, and Voorhees 2012). ... Our experiment is conducted on Clue Web09 Category B, which contains more than 50 million English webpages (231 Gigabytes in size).3 ... 3http://lemurproject.org/clueweb09/. We also test our full model on the most recent MQ2008 dataset for LETOR 4.0. LETOR MQ2008 contains 784 queries and 15,211 annotated documents. |
| Dataset Splits | Yes | We implement a 10-fold cross-validation for fair evaluation. |
| Hardware Specification | No | The paper does not specify any hardware details (e.g., GPU models, CPU types, memory) used for conducting the experiments. |
| Software Dependencies | No | The paper mentions software components like 'Adam' (optimizer) and 'word2vec model' but does not provide specific version numbers for any software or libraries used. |
| Experiment Setup | Yes | For the Text Tiling algorithm, we set α to 20 and β to 6, as recommended by Hearst. For the query-document interaction matrix, the parameters are slight differences between the two datasets. In TREC Web, nq = 5 and In LETOR nq = 9. ... For both datasets, nb = 30. ... For the Deep Tile Bars algorithm, we set l = 10 for both datasets. In TREC Web Track dataset, the number of filters of CNN with same kernel size and the number of units in each LSTM are both set to 3; while in LETOR, this number is set to 9. The MLP contains two hidden layers, with 32 and 16 units for TREC Web, and 128 and 16 units for LETOR. |