Name
Abstract | ||
|---|---|---|
Recently, pre-training methods tailored for IR tasks have achieved great success. However, as the mechanisms behind the performance improvement remain under-investigated, the interpretability and robustness of these pre-trained models still need to be improved. Axiomatic IR aims to identify a set of desirable properties expressed mathematically as formal constraints to guide the design of ranking models. Existing studies have already shown that considering certain axioms may help improve the effectiveness and interpretability of IR models. However, there still lack efforts of incorporating these IR axioms into pre-training methodologies. To shed light on this research question, we propose a novel pre-training method with \underlineA xiomatic \underlineRe gularization for ad hoc \underlineS earch (ARES). In the ARES framework, a number of existing IR axioms are re-organized to generate training samples to be fitted in the pre-training process. These training samples then guide neural rankers to learn the desirable ranking properties. Compared to existing pre-training approaches, ARES is more intuitive and explainable. Experimental results on multiple publicly available benchmark datasets have shown the effectiveness of ARES in both full-resource and low-resource (e.g., zero-shot and few-shot) settings. An intuitive case study also indicates that ARES has learned useful knowledge that existing pre-trained models (e.g., BERT and PROP) fail to possess. This work provides insights into improving the interpretability of pre-trained models and the guidance of incorporating IR axioms or human heuristics into pre-training methods. |
Year | DOI | Venue |
|---|---|---|
2022 | 10.1145/3477495.3531943 | SIGIR '22: Proceedings of the 45th International ACM SIGIR Conference on Research and Development in Information Retrieval |
Keywords | DocType | Citations |
Pre-training Method, Pre-trained Language Model, Axiomatic IR, Ad hoc Search | Conference | 0 |
PageRank | References | Authors |
0.34 | 23 | 9 |
Name | Order | Citations | PageRank |
|---|---|---|---|
| Jia Chen | 1 | 22 | 4.16 |
| Yiqun Liu | 2 | 1592 | 136.51 |
| Fang Yan | 3 | 9 | 5.01 |
| Jiaxin Mao | 4 | 164 | 26.30 |
| Hui Fang | 5 | 918 | 63.03 |
| Shenghao Yang | 6 | 0 | 0.34 |
| Xiaohui Xie | 7 | 40 | 7.49 |
| Min Zhang | 8 | 11 | 8.01 |
| Shaoping Ma | 9 | 1544 | 126.00 |