[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-86162-en":3,"doc-seo-86162-105":30,"detail-sidebar-cat-0-en-105":91},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":20,"is_deleted":4,"is_public":21,"is_downloadable":21,"audit_status":21,"page_count":22,"language":23,"language_code":24,"site_id":25,"html_lang":24,"table_of_contents":26,"faqs":27,"seo_title":13,"seo_description":14,"update_tm":28,"read_time":29},86162,962075114101,"Seraphina","https://ap-avatar.wpscdn.com/avatar/e000253a75eb197efd?x-image-process=image/resize,m_fixed,w_180,h_180&k=1780044092746381165",8,"Research & Report","SISA-Rec: A Semantically Integrated Sequential Recommender with Contrastive Alignment","Recommendation systems assist users in selecting relevant items from large catalogs. This work targets transformer-based sequential recommendation that relies mainly on item identifiers and underuses item semantics, a gap that becomes severe in sparse and cold-start settings. SISA-Rec embeds semantic context within sequential modeling by fusing item ID embeddings with BERT-based text embeddings via gated fusion, injecting semantic similarity into self-attention, and aggregating user representations using attention. Joint learning combines BPR and contrastive alignment, evaluated on extremely sparse Amazon Beauty and Amazon Toys & Games, improving all metrics and boosting cold-start gains substantially.","arXiv :2607 . 11168v1 [ cs .CV] 13 Jul 2026  \nSISA-Rec: A Semantically Integrated Sequential Recommender with Contrastive Alignment  \nSoohan Abbasia , Shahid Munir Shahb , Rafia Shaikhc , Mahmoud Aljawarnehd  \na Department of Computer Science, DHA Suffa University, Karachi, Sindh, Pakistan b Department of Computer Science, SZABIST, Pakistan  \nc National University of Computer and Emerging Sciences (FAST), Karachi, Sindh, Pakistan d Department of Computer Science, Applied Science Private University, Amman, Jordan  \nAbstract  \nRecommendation systems help users recommend relevant items from a large collection of choices. Present work on transformer-based sequential recommendation learns user preferences from interaction logs, but it mostly focuses on item identifiers and doesn’t fully use the semantic meaning of items. This limitation becomes a major challenge in sparse and cold-start scenarios where historical interaction data is limited. To solve this problem, we introduce SISA-Rec (Semantically Integrated Sequential Recommendation), a transformer-based framework that embeds semantic context directly into sequential modeling. Our approach fuses item ID embeddings with BERT-based text embeddings via a gated fusion module, injects semantic similarity into the selfattention mechanism, and leverages an attention-based aggregation module to construct comprehensive user representations. Finally, a joint learning objective which combines Bayesian Personalized Ranking (BPR) and contrastive alignment loss, aligns the underlying behavioral and semantic spaces. Experiments were conducted on the two highly sparse Amazon Beauty and Amazon Toys & Games datasets, both having 99.93% sparsity. The results show that SISA-Rec outperforms state-of-the-art baseline models across all evaluation metrics. Compared with the BERT4Rec [1], SISA-Rec improves HR@10 by 16.6% and NDCG@10 by 10.3% on Amazon Beauty, and HR@10 by 23.1% and NDCG@10 by 17.9% on Amazon Toys & Games. Coldstart analysis further shows that the proposed model achieves the largest improvements for users with limited interaction historical records. This showcases the value of semantic information when user behavior data is scarce. Overall, the results demonstrate that integrating semantic information into the attention mechanism leads to more accurate and reliable recommendations  \n1. Introduction  \nRecommender systems have become a fundamental part of how people engage with online platforms. With the continued growth of digital content in e-commerce, streaming, and social media, the objective of these systems is to identify a manageable set of relevant items from a large pool of available options for each user. Recent advances in deep learning have significantly  \nEmail addresses: [soohan.abbasi@dsu.edu.pk](soohan.abbasi@dsu.edu.pk) (Soohan Abbasi), [dr.shahid@ghr.szabist.edu.pk](dr.shahid@ghr.szabist.edu.pk) (Shahid  \nMunir Shah), [ma_jawarneh@asu.edu.jo](ma_jawarneh@asu.edu.jo) (Mahmoud Aljawarneh)  \nimproved recommendation performance by learning richer representations from user behavior. Among these, sequential recommendation has drawn a lot of attention, since treating a user’s interaction history as an ordered sequence makes it possible to capture how preferences shift over time rather than assuming they stay fixed. Transformer-based architectures such as SASRec and BERT4Rec push this further through self-attention, which lets the model weigh relationships between items across an entire sequence and pick up the contextual dependencies that simpler models tend to miss. Alongside these developments, there has been a steady effort to bring item content, or semantics, more directly into the recommendation process. AI Models like UniSRec and MoRec have shown that textual and modality-based item representations are valuable to include, with modality-based encoders matching or even improving upon identity-based models in several settings. Knowledge graph approaches such as DiffKG and Gf","cbCaifFrIuXcqs4m","https://ap.wps.com/l/cbCaifFrIuXcqs4m","pdf",2551079,4,1,29,"English","en",105,"# Introduction\n## Problem Motivation: Sparse and Cold-Start Limitations\n## Approach Overview: Semantically Integrated Sequential Modeling\n## Experimental Evaluation and Results","[{\"question\":\"What key limitation does SISA-Rec address in transformer-based sequential recommendation?\",\"answer\":\"It addresses the tendency of prior models to focus mostly on item identifiers while not fully leveraging the semantic meaning of items, which weakens performance in sparse and cold-start scenarios.\"},{\"question\":\"How does SISA-Rec incorporate semantic information into the sequential modeling process?\",\"answer\":\"It fuses item ID embeddings with BERT-based text embeddings using a gated fusion module, injects semantic similarity into the self-attention mechanism, and uses an attention-based aggregation to form user representations.\"},{\"question\":\"What learning objective does SISA-Rec use to align behavioral and semantic spaces?\",\"answer\":\"It uses a joint objective combining Bayesian Personalized Ranking (BPR) with a contrastive alignment loss to align underlying behavioral and semantic representations.\"}]",1784209010,73,{"code":4,"msg":31,"data":32},"ok",{"site_id":25,"language":24,"slug":33,"title":13,"keywords":34,"description":14,"schema_data":35,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":28},"sisa-rec-a-semantically-integrated-sequential-recommender-with-contrastive-alignment","",{"@graph":36,"@context":85},[37,53,68],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,48,51],{"item":41,"name":42,"@type":43,"position":21},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":47},"https://docshare.wps.com/document/","Document",2,{"item":49,"name":12,"@type":43,"position":50},"https://docshare.wps.com/document/research-report/",3,{"item":52,"name":13,"@type":43,"position":20},"https://docshare.wps.com/document/sisa-rec-a-semantically-integrated-sequential-recommender-with-contrastive-alignment/86162/",{"url":52,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":24,"description":14,"dateModified":61,"datePublished":62,"encodingFormat":60,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":41,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-25","2026-07-16",true,{"@type":65,"interactionType":66,"userInteractionCount":20},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"What key limitation does SISA-Rec address in transformer-based sequential recommendation?","Question",{"text":75,"@type":76},"It addresses the tendency of prior models to focus mostly on item identifiers while not fully leveraging the semantic meaning of items, which weakens performance in sparse and cold-start scenarios.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"How does SISA-Rec incorporate semantic information into the sequential modeling process?",{"text":80,"@type":76},"It fuses item ID embeddings with BERT-based text embeddings using a gated fusion module, injects semantic similarity into the self-attention mechanism, and uses an attention-based aggregation to form user representations.",{"name":82,"@type":73,"acceptedAnswer":83},"What learning objective does SISA-Rec use to align behavioral and semantic spaces?",{"text":84,"@type":76},"It uses a joint objective combining Bayesian Personalized Ranking (BPR) with a contrastive alignment loss to align underlying behavioral and semantic representations.","https://schema.org",{"og:url":52,"og:type":87,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":89,"canonical":52},"index,follow",{"doc_id":7,"site_id":25},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,120,123,128,131,135],{"id":21,"doc_module":4,"doc_module_name":46,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":47,"doc_module":4,"doc_module_name":46,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":20,"doc_module":4,"doc_module_name":46,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":46,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":46,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":46,"category_name":117,"show_sort_weight":118,"slug":119},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":121,"slug":122},30,"research-report",{"id":124,"doc_module":4,"doc_module_name":46,"category_name":125,"show_sort_weight":126,"slug":127},9,"Religion & Spirituality",20,"religion-spirituality",{"id":126,"doc_module":4,"doc_module_name":46,"category_name":129,"show_sort_weight":126,"slug":130},"World Cup","world-cup",{"id":132,"doc_module":4,"doc_module_name":46,"category_name":133,"show_sort_weight":132,"slug":134},10,"Lifestyle","lifestyle",{"id":136,"doc_module":4,"doc_module_name":46,"category_name":137,"show_sort_weight":106,"slug":138},19,"General","general"]