Impressive milestones have been achieved in text matching by adopting a cross-attention mechanism to capture pertinent semantic connections between two sentences. However, these cross-attention mechanisms focus on word-level links between the two inputs, neglecting the importance of contextual information. We propose a context-aware interaction network (COIN) to properly align two sequences and infer their semantic relationship. Specifically, each interaction block includes (1) a context-aware cross-attention mechanism to effectively integrate contextual information, and (2) a gate fusion layer to flexibly interpolate aligned representations. We apply multiple stacked interaction blocks to produce alignments at different levels and gradually refine the attention results. Experiments on two question matching datasets and detailed analyses confirm the effectiveness of our model.
Knowledge graphs (KG) have become increasingly important to endow modern recommender systems with the ability to generate traceable reasoning paths to explain the recommendation process. However, prior research rarely considers the faithfulness of the derived explanations to justify the decision making process. To the best of our knowledge, this is the first work that models and evaluates faithfully explainable recommendation under the framework of KG reasoning. Specifically, we propose neural logic reasoning for explainable recommendation (LOGER) by drawing on interpretable logical rules to guide the path reasoning process for explanation generation. We experiment on three large-scale datasets in the e-commerce domain, demonstrating the effectiveness of our method in delivering high-quality recommendations as well as ascertaining the faithfulness of the derived explanation.
Recommender systems are important and valuable tools for many personalized services. Collaborative Filtering (CF) algorithms -- among others -- are fundamental algorithms driving the underlying mechanism of personalized recommendation. Many of the traditional CF algorithms are designed based on the fundamental idea of mining or learning correlative patterns from data for matching, including memory-based methods such as user/item-based CF as well as learning-based methods such as matrix factorization and deep learning models. However, advancing from correlative learning to causal learning is an important problem, because causal/counterfactual modeling can help us to think outside of the observational data for user modeling and personalization. In this paper, we propose Causal Collaborative Filtering (CCF) -- a general framework for modeling causality in collaborative filtering and recommendation. We first provide a unified causal view of CF and mathematically show that many of the traditional CF algorithms are actually special cases of CCF under simplified causal graphs. We then propose a conditional intervention approach for $do$-calculus so that we can estimate the causal relations based on observational data. Finally, we further propose a general counterfactual constrained learning framework for estimating the user-item preferences. Experiments are conducted on two types of real-world datasets -- traditional and randomized trial data -- and results show that our framework can improve the recommendation performance of many CF algorithms.
BERT has achieved superior performances on Natural Language Understanding (NLU) tasks. However, BERT possesses a large number of parameters and demands certain resources to deploy. For acceleration, Dynamic Early Exiting for BERT (DeeBERT) has been proposed recently, which incorporates multiple exits and adopts a dynamic early-exit mechanism to ensure efficient inference. While obtaining an efficiency-performance tradeoff, the performances of early exits in multi-exit BERT are significantly worse than late exits. In this paper, we leverage gradient regularized self-distillation for RObust training of Multi-Exit BERT (RomeBERT), which can effectively solve the performance imbalance problem between early and late exits. Moreover, the proposed RomeBERT adopts a one-stage joint training strategy for multi-exits and the BERT backbone while DeeBERT needs two stages that require more training time. Extensive experiments on GLUE datasets are performed to demonstrate the superiority of our approach. Our code is available at https://github.com/romebert/RomeBERT.
This paper proposes a discrete knowledge graph (KG) embedding (DKGE) method, which projects KG entities and relations into the Hamming space based on a computationally tractable discrete optimization algorithm, to solve the formidable storage and computation cost challenges in traditional continuous graph embedding methods. The convergence of DKGE can be guaranteed theoretically. Extensive experiments demonstrate that DKGE achieves superior accuracy than classical hashing functions that map the effective continuous embeddings into discrete codes. Besides, DKGE reaches comparable accuracy with much lower computational complexity and storage compared to many continuous graph embedding methods.
Pre-trained language models such as BERT have achieved great success in a broad range of natural language processing tasks. However, BERT cannot well support E-commerce related tasks due to the lack of two levels of domain knowledge, i.e., phrase-level and product-level. On one hand, many E-commerce tasks require an accurate understanding of domain phrases, whereas such fine-grained phrase-level knowledge is not explicitly modeled by BERT's training objective. On the other hand, product-level knowledge like product associations can enhance the language modeling of E-commerce, but they are not factual knowledge thus using them indiscriminately may introduce noise. To tackle the problem, we propose a unified pre-training framework, namely, E-BERT. Specifically, to preserve phrase-level knowledge, we introduce Adaptive Hybrid Masking, which allows the model to adaptively switch from learning preliminary word knowledge to learning complex phrases, based on the fitting progress of two modes. To utilize product-level knowledge, we introduce Neighbor Product Reconstruction, which trains E-BERT to predict a product's associated neighbors with a denoising cross attention layer. Our investigation reveals promising results in four downstream tasks, i.e., review-based question answering, aspect extraction, aspect sentiment classification, and product classification.
In this work, we present a new dataset for conversational recommendation over knowledge graphs in e-commerce platforms called COOKIE. The dataset is constructed from an Amazon review corpus by integrating both user-agent dialogue and custom knowledge graphs for recommendation. Specifically, we first construct a unified knowledge graph and extract key entities between user--product pairs, which serve as the skeleton of a conversation. Then we simulate conversations mirroring the human coarse-to-fine process of choosing preferred items. The proposed baselines and experiments demonstrate that our dataset is able to provide innovative opportunities for conversational recommendation.
In cross-lingual text classification, one seeks to exploit labeled data from one language to train a text classification model that can then be applied to a completely different language. Recent multilingual representation models have made it much easier to achieve this. Still, there may still be subtle differences between languages that are neglected when doing so. To address this, we present a semi-supervised adversarial training process that minimizes the maximal loss for label-preserving input perturbations. The resulting model then serves as a teacher to induce labels for unlabeled target language samples that can be used during further adversarial training, allowing us to gradually adapt our model to the target language. Compared with a number of strong baselines, we observe significant gains in effectiveness on document and intent classification for a diverse set of languages.