Improving Conversational Recommender System by Pretraining Billion-scale Knowledge Graph

Wong, Chi-Man; Feng, Fan; Zhang, Wen; Vong, Chi‐Man; Chen, Hui; Zhang, Yichi; He, Peng; Chen, Huan; Zhao, Kun; Chen, Huajun

doi:10.1109/icde51399.2021.00291

Cited by 30 publications

(7 citation statements)

References 21 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The DIN [21], DIEN [10], TIEN [22], and MARN [8] approaches use sequential activity information, which we also rely on in our approach. CTR approaches are evaluated on different datasets, some publications and approaches rely only on closed data [13,18,[23][24][25] which are not included in Table 1. Others, as shown in Table 1, use openly available datasets to evaluate their approach.…”

Section: Approaching Click-through Rate Predictionmentioning

confidence: 99%

It’s Not Always about Wide and Deep Models: Click-Through Rate Prediction with a Customer Behavior-Embedding Representation

Alves Gomes,

Meyes,

Meisen

et al. 2024

JTAER

View full text Add to dashboard Cite

Alongside natural language processing and computer vision, large learning models have found their way into e-commerce. Especially, for recommender systems and click-through rate prediction, these models have shown great predictive power. In this work, we aim to predict the probability that a customer will click on a given recommendation, given only its current session. Therefore, we propose a two-stage approach consisting of a customer behavior-embedding representation and a recurrent neural network. In the first stage, we train a self-supervised skip-gram embedding on customer activity data. The resulting embedding representation is used in the second stage to encode the customer sequences which are then used as input to the learning model. Our proposed approach diverges from the prevailing trend of utilizing extensive end-to-end models for click-through rate prediction. The experiments, which incorporate a real-world industrial use case and a widely used as well as openly available benchmark dataset, demonstrate that our approach outperforms the current state-of-the-art models. Our approach predicts customers’ click intention with an average F1 accuracy of 94% for the industrial use case which is one percentage point higher than the state-of-the-art baseline and an average F1 accuracy of 79% for the benchmark dataset, which outperforms the best tested state-of-the-art baseline by more than seven percentage points. The results show that, contrary to current trends in that field, large end-to-end models are not always needed. The analysis of our experiments suggests that the reason for the performance of our approach is the self-supervised pre-trained embedding of customer behavior that we use as the customer representation.

show abstract

Section: Approaching Click-through Rate Predictionmentioning

confidence: 99%

It’s Not Always about Wide and Deep Models: Click-Through Rate Prediction with a Customer Behavior-Embedding Representation

Alves Gomes,

Meyes,

Meisen

et al. 2024

JTAER

View full text Add to dashboard Cite

show abstract

“…Pre-training techniques are widely used in feature-based recommendation systems to enhance user or item representations [46]. Qiu et al [24] propose review encoder pre-training to complement user representations, while Wong et al [39] utilize pre-training on a large-scale knowledge graph for conversational recommender systems. However, these methods are not applicable to medication recommendation.…”

Section: Related Workmentioning

confidence: 99%

Infectious Hospital Area, the First Affiliated Hospital of University of Science and Technology of China, Hefei, China

Zhao¹,

Hu²,

Xue³

et al. 2021

BEBM

View full text Add to dashboard Cite

With advances in deep learning, neural network based speech enhancement (SE) has developed rapidly in the last decade. Meanwhile, the self-supervised pre-trained model and vector quantization (VQ) have achieved excellent performance on many speech-related tasks, while they are less explored on SE. As it was shown in our previous work that utilizing a VQ module to discretize noisy speech representations is beneficial for speech denoising, in this work we therefore study the impact of using VQ at different layers with different number of codebooks. Different VQ modules indeed enable to extract multiplegranularity speech features. Following an attention mechanism, the contextual features extracted by a pre-trained model are fused with the local features extracted by the encoder, such that both global and local information are preserved to reconstruct the enhanced speech. Experimental results on the Valentini dataset show that the proposed model can improve the SE performance, where the impact of choosing pre-trained models is also revealed.

show abstract

“…Moreover, it can use side information to enhance the recommendation effectiveness. KG has recently been employed as auxiliary data in a variety of recommendation tasks [1], [2], as well as for news RS [3], [4]. KGs are also widely used in information retrieval [5], digital assistants [6], fraud detection [7], [8], chatbots [9], and questioning and answering(QA) systems [10].…”

Section: Introductionmentioning

confidence: 99%

Urdu Wikification and Its Application in Urdu News Recommendation System

et al. 2022

View full text Add to dashboard Cite

Wikification is the process of linking the entities found in a sample text to their individual Wikipedia or Wikidata pages. Many natural language processing applications, including question-answering systems, information retrieval, fraud detection, and recommendation systems(RS), can benefit from this information extraction technique. There has been a great deal of effort put towards entity-linking(EL) for both Asian and Western languages, with several datasets and numerous proposed methodologies. Despite millions of Urdu language users globally, the entitylinking for the Urdu language has not been worked on, to the best of our knowledge. This work proposes an Urdu EL pipeline to identify named entities in text and link them to Wikidata. Secondly, a dataset of 550 Urdu news titles relating to their respective Wiki-ids has been prepared for the examination. Third, utilizing the proposed EL pipeline, 16738 news articles from the firstever Urdu news RS dataset of 100 users are annotated. Fourthly, a sub Knowledge graph (KG) of 8439 entities and 23080 relationship tuples is retrieved from Wikidata. The Trans-E algorithm is then used to create KG embeddings so that the extracted KG may be used in an Urdu news RS. The final accuracy of Urdu news RS is 60.8%.

show abstract

Improving Conversational Recommender System by Pretraining Billion-scale Knowledge Graph

Cited by 30 publications

References 21 publications

It’s Not Always about Wide and Deep Models: Click-Through Rate Prediction with a Customer Behavior-Embedding Representation

It’s Not Always about Wide and Deep Models: Click-Through Rate Prediction with a Customer Behavior-Embedding Representation

Infectious Hospital Area, the First Affiliated Hospital of University of Science and Technology of China, Hefei, China

Urdu Wikification and Its Application in Urdu News Recommendation System

Contact Info

Product

Resources

About