Recently, semantic search has been successfully applied to e-commerce product search and the learned semantic space(s) for query and product encoding are expected to generalize to unseen queries or products. Yet, whether generalization can conveniently emerge has not been thoroughly studied in the domain thus far. In this paper, we examine several general-domain and domain-specific pre-trained Roberta variants and discover that general-domain fine-tuning does not help generalization, which aligns with the discovery of prior art. Proper domain-specific fine-tuning with clickstream data can lead to better model generalization, based on a bucketed analysis of a publicly available manual annotated query-product pair data.
Vector quantization (VQ) based ANN indexes, such as Inverted File System (IVF) and Product Quantization (PQ), have been widely applied to embedding based document retrieval thanks to the competitive time and memory efficiency. Originally, VQ is learned to minimize the reconstruction loss, i.e., the distortions between the original dense embeddings and the reconstructed embeddings after quantization. Unfortunately, such an objective is inconsistent with the goal of selecting ground-truth documents for the input query, which may cause severe loss of retrieval quality. Recent works identify such a defect, and propose to minimize the retrieval loss through contrastive learning. However, these methods intensively rely on queries with ground-truth documents, whose performance is limited by the insufficiency of labeled data. In this paper, we propose Distill-VQ, which unifies the learning of IVF and PQ within a knowledge distillation framework. In Distill-VQ, the dense embeddings are leveraged as "teachers", which predict the query's relevance to the sampled documents. The VQ modules are treated as the "students", which are learned to reproduce the predicted relevance, such that the reconstructed embeddings may fully preserve the retrieval result of the dense embeddings. By doing so, Distill-VQ is able to derive substantial training signals from the massive unlabeled data, which significantly contributes to the retrieval quality. We perform comprehensive explorations for the optimal conduct of knowledge distillation, which may provide useful insights for the learning of VQ based ANN index. We also experimentally show that the labeled data is no longer a necessity for high-quality vector quantization, which indicates Distill-VQ's strong applicability in practice.
The lack of labeled data is a major obstacle to learning high-quality sentence embeddings. Recently, self-supervised contrastive learning (SCL) is regarded as a promising way to address this problem. However, the existing works mainly rely on hand-crafted data annotation heuristics to generate positive training samples, which not only call for domain expertise and laborious tuning, but are also prone to the following unfavorable cases: 1) trivial positives, 2) coarse-grained positives, and 3) false positives. As a result, the self-supervision's quality can be severely limited in reality. In this work, we propose a novel framework InfoCSE to address the above problems. Instead of relying on annotation heuristics defined by humans, it leverages the sentence representation model itself and realizes the following iterative self-supervision process: on one hand, the improvement of sentence representation may contribute to the quality of data annotation; on the other hand, more effective data annotation helps to generate high-quality positive samples, which will further improve the current sentence representation model. In other words, the representation learning and data annotation become mutually reinforced, where a strong self-supervision effect can be derived. Extensive experiments are performed based on three benchmark datasets, where notable improvements can be achieved against the existing SCL-based methods.
Embedding based retrieval (EBR) is a fundamental building block in many web applications. However, EBR in sponsored search is distinguished from other generic scenarios and technically challenging due to the need of serving multiple retrieval purposes: firstly, it has to retrieve high-relevance ads, which may exactly serve user's search intent; secondly, it needs to retrieve high-CTR ads so as to maximize the overall user clicks. In this paper, we present a novel representation learning framework Uni-Retriever developed for Bing Search, which unifies two different training modes knowledge distillation and contrastive learning to realize both required objectives. On one hand, the capability of making high-relevance retrieval is established by distilling knowledge from the ``relevance teacher model''. On the other hand, the capability of making high-CTR retrieval is optimized by learning to discriminate user's clicked ads from the entire corpus. The two training modes are jointly performed as a multi-objective learning process, such that the ads of high relevance and CTR can be favored by the generated embeddings. Besides the learning strategy, we also elaborate our solution for EBR serving pipeline built upon the substantially optimized DiskANN, where massive-scale EBR can be performed with competitive time and memory efficiency, and accomplished in high-quality. We make comprehensive offline and online experiments to evaluate the proposed techniques, whose findings may provide useful insights for the future development of EBR systems. Uni-Retriever has been mainstreamed as the major retrieval path in Bing's production thanks to the notable improvements on the representation and EBR serving quality.
Ad-hoc search calls for the selection of appropriate answers from a massive-scale corpus. Nowadays, the embedding-based retrieval (EBR) becomes a promising solution, where deep learning based document representation and ANN search techniques are allied to handle this task. However, a major challenge is that the ANN index can be too large to fit into memory, given the considerable size of answer corpus. In this work, we tackle this problem with Bi-Granular Document Representation, where the lightweight sparse embeddings are indexed and standby in memory for coarse-grained candidate search, and the heavyweight dense embeddings are hosted in disk for fine-grained post verification. For the best of retrieval accuracy, a Progressive Optimization framework is designed. The sparse embeddings are learned ahead for high-quality search of candidates. Conditioned on the candidate distribution induced by the sparse embeddings, the dense embeddings are continuously learned to optimize the discrimination of ground-truth from the shortlisted candidates. Besides, two techniques: the contrastive quantization and the locality-centric sampling are introduced for the learning of sparse and dense embeddings, which substantially contribute to their performances. Thanks to the above features, our method effectively handles massive-scale EBR with strong advantages in accuracy: with up to +4.3% recall gain on million-scale corpus, and up to +17.5% recall gain on billion-scale corpus. Besides, Our method is applied to a major sponsored search platform with substantial gains on revenue (+1.95%), Recall (+1.01%) and CTR (+0.49%).
Graph neural networks (GNNs) have been widely applied in the recommendation tasks and have obtained very appealing performance. However, most GNN-based recommendation methods suffer from the problem of data sparsity in practice. Meanwhile, pre-training techniques have achieved great success in mitigating data sparsity in various domains such as natural language processing (NLP) and computer vision (CV). Thus, graph pre-training has the great potential to alleviate data sparsity in GNN-based recommendations. However, pre-training GNNs for recommendations face unique challenges. For example, user-item interaction graphs in different recommendation tasks have distinct sets of users and items, and they often present different properties. Therefore, the successful mechanisms commonly used in NLP and CV to transfer knowledge from pre-training tasks to downstream tasks such as sharing learned embeddings or feature extractors are not directly applicable to existing GNN-based recommendations models. To tackle these challenges, we delicately design an adaptive graph pre-training framework for localized collaborative filtering (ADAPT). It does not require transferring user/item embeddings, and is able to capture both the common knowledge across different graphs and the uniqueness for each graph. Extensive experimental results have demonstrated the effectiveness and superiority of ADAPT.
Recently, Graph Neural Networks (GNNs) have proven their effectiveness for recommender systems. Existing studies have applied GNNs to capture collaborative relations in the data. However, in real-world scenarios, the relations in a recommendation graph can be of various kinds. For example, two movies may be associated either by the same genre or by the same director/actor. If we use a single graph to elaborate all these relations, the graph can be too complex to process. To address this issue, we bring the idea of pre-training to process the complex graph step by step. Based on the idea of divide-and-conquer, we separate the large graph into three sub-graphs: user graph, item graph, and user-item interaction graph. Then the user and item embeddings are pre-trained from user and item graphs, respectively. To conduct pre-training, we construct the multi-relational user graph and item graph, respectively, based on their attributes. In this paper, we propose a novel Reinforced Attentive Multi-relational Graph Neural Network (RAM-GNN) to the pre-train user and item embeddings on the user and item graph prior to the recommendation step. Specifically, we design a relation-level attention layer to learn the importance of different relations. Next, a Reinforced Neighbor Sampler (RNS) is applied to search the optimal filtering threshold for sampling top-k similar neighbors in the graph, which avoids the over-smoothing issue. We initialize the recommendation model with the pre-trained user/item embeddings. Finally, an aggregation-based GNN model is utilized to learn from the collaborative relations in the user-item interaction graph and provide recommendations. Our experiments demonstrate that RAM-GNN outperforms other state-of-the-art graph-based recommendation models and multi-relational graph neural networks.
This paper reviews and proposes concerns in adopting, fielding, and maintaining artificial intelligence (AI) systems. While the AI community has made rapid progress, there are challenges in certifying AI systems. Using procedures from design and operational test and evaluation, there are opportunities towards determining performance bounds to manage expectations of intended use. A notional use case is presented with image data fusion to support AI object recognition certifiability considering precision versus distance.
Linked text representation is critical for many intelligent web applications, such as online advertisement and recommender systems. Recent breakthroughs on pretrained language models and graph neural networks facilitate the development of corresponding techniques. However, the existing works mainly rely on cascaded model structures: the texts are independently encoded by language models at first, and the textual embeddings are further aggregated by graph neural networks. We argue that the neighbourhood information is insufficiently utilized within the above process, which restricts the representation quality. In this work, we propose GraphFormers, where graph neural networks are nested alongside each transformer layer of the language models. On top of the above architecture, the linked texts will iteratively extract neighbourhood information for the enhancement of their own semantics. Such an iterative workflow gives rise to more effective utilization of neighbourhood information, which contributes to the representation quality. We further introduce an adaptation called unidirectional GraphFormers, which is much more efficient and comparably effective; and we leverage a pretraining strategy called the neighbourhood-aware masked language modeling to enhance the training effect. We perform extensive experiment studies with three large-scale linked text datasets, whose results verify the effectiveness of our proposed methods.
Recent years, analysis dictionary learning (ADL) and its applications for classification have been well developed, due to its flexible projective ability and low classification complexity. With the learned analysis dictionary, test samples can be transformed into a sparse subspace for classification efficiently. However, the underling locality of sample data has rarely been explored in analysis dictionary to enhance the discriminative capability of the classifier. In this paper, we propose a novel locality constrained analysis dictionary learning model with a synthesis K-SVD algorithm (SK-LADL). It considers the intrinsic geometric properties by imposing graph regularization to uncover the geometric structure for the image data. Through the learned analysis dictionary, we transform the image to a new and compact space where the manifold assumption can be further guaranteed. thus, the local geometrical structure of images can be preserved in sparse representation coefficients. Moreover, the SK-LADL model is iteratively solved by the synthesis K-SVD and gradient technique. Experimental results on image classification validate the performance superiority of our SK-LADL model.