THUIR2 at NTCIR-16 Session Search (SS) Task

TL;DR

This paper details THUIR2's participation in the NTCIR-16 Session Search Task, where they used fine-tuned language models and learning-to-rank techniques to achieve top performance in both subtasks.

Contribution

The paper introduces the use of fine-tuned pre-trained language models combined with learning-to-rank methods for session search tasks, achieving state-of-the-art results.

Findings

Achieved best performance in FOSS subtask

Achieved best performance in POSS subtask

Effective combination of fine-tuned language models and learning-to-rank

Abstract

Our team(THUIR2) participated in both FOSS and POSS subtasks of the NTCIR-161 Session Search (SS) Task. This paper describes our approaches and results. In the FOSS subtask, we submit five runs using learning-to-rank and fine-tuned pre-trained language models. We fine-tuned the pre-trained language model with ad-hoc data and session information and assembled them by a learning-to-rank method. The assembled model achieves the best performance among all participants in the preliminary evaluation. In the POSS subtask, we used an assembled model which also achieves the best performance in the preliminary evaluation.