Contrastive Learning of User Behavior Sequence for Context-Aware Document Ranking

Abstract

Context information in search sessions has proven to be useful for capturing user search intent. Existing studies explored user behavior sequences in sessions in different ways to enhance query suggestion or document ranking. However, a user behavior sequence has often been viewed as a definite and exact signal reflecting a user's behavior. In reality, it is highly variable: user's queries for the same intent can vary, and different documents can be clicked. To learn a more robust representation of the user behavior sequence, we propose a method based on contrastive learning, which takes into account the possible variations in user's behavior sequences. Specifically, we propose three data augmentation strategies to generate similar variants of user behavior sequences and contrast them with other sequences. In so doing, the model is forced to be more robust regarding the possible variations. The optimized sequence representation is incorporated into document ranking. Experiments on two real query log datasets show that our proposed model outperforms the state-of-the-art methods significantly, which demonstrates the effectiveness of our method for context-aware document ranking. CCS CONCEPTS• Information systems → Retrieval models and ranking.

Reproducing Personalised Session Search Over the AOL Query Log

“…Since then, however, it has been an important resource to the research community (e.g., [25,30]). Even to this day, the AOL Query Log continues to enable studies in analysis of data leaks [13], search autocompletion [14], weak supervision for adhoc search [7,21], search result personalisation [9,16], and session-based search [1,2,6,28,33].…”

Target-aware Abstractive Related Work Generation with Contrastive Learning

“…Our model is also related to contrastive learning, which has boosted unsupervised representation learning in recommendation [31,40], computer vision [21,30], and natural language processing [29,43]. In text generation, Logeswaran and Lee [27] proposed to learn better sentence representations by using a classifier to distinguish context sentences from other contrastive sentences.…”

Pre-training Methods in Information Retrieval

“…Most existing personalized approaches do not involve a well-designed pre-training or self-supervised learning (SSL) process, merely utilizing the powerful learning ability of Transformer-like architectures. Recently, some researchers focused on designing pre-training objectives for personalized search (Zhou et al, 2021b) or session search (Zhu et al, 2021). Their work have shown the great potential of applying contrastive learning in encoding user search history and the content.…”