From Feature Fusion to FoundationModels: A Systematic Review of Multimodal Hashtag Recommendationin Social Media (2016--2025)

Abstract Hashtags enable content organisation, discoverability, and community formation on social media, yet automated recommendation of contextually appropriate hashtags from multimodal post content—text, images, audio, and video together—remains underexplored as a distinct task. To the best of our targeted search, no existing review synthesises multimodal hashtag recommendation (MHR) specifically. Following PRISMA 2020 [59], we searched IEEE Xplore, ACM Digital Library, Scopus, and SpringerLink for studies published January 2016–September 2025. After independent two-stage screening (Cohen’s κ=0.81 and 0.85 [60]), 30 studies met eligibility and were assessed with a five-criterion quality rubric (mean 5.9±1.8). We classify included studies into six architectural families—early fusion, late fusion, attention-based, graph-enhanced, transformer/LLM/RAG, and an emerging neurosymbolic-contrastive direction—and introduce the Hashtag Intelligence Maturity Model (HIMM), a five-level framework assessing MHR systems by functional capability. Image–text combinations dominate the corpus, with micro-video and retrieval-augmented generation (RAG) as emerging subfields. A reproducibility and fairness audit finds diversity, fairness, and explainability metrics essentially absent from current evaluations, accessible code in only 7% of studies, and multilingual coverage of only 13%. We identify six structural research gaps and a six-directional research agenda addressing foundation-model integration, generative hashtag synthesis, cross-platform benchmarking, multilingual coverage, explainability, and ethical deployment.

Authors

Institutions

Publication Details

Journal
Research Square
Published
2026-11-17
DOI
https://doi.org/10.21203/rs.3.rs-11232903/v1
Primary Topic
Multimodal Machine Learning Applications
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

From Feature Fusion to FoundationModels: A Systematic Review of Multimodal Hashtag Recommendationin Social Media (2016--2025)

Fatimah Khalid, Azreen Azman, Puteri Suhaiza Sulaiman, Hi Teng Heon
2 links
Research Square
Multimodal Machine Learning Applications
preprint

From Feature Fusion to FoundationModels: A Systematic Review of Multimodal Hashtag Recommendationin Social Media (2016--2025)

Fatimah Khalid, Azreen Azman, Puteri Suhaiza Sulaiman, Hi Teng Heon
preprint en

Abstract

Abstract Hashtags enable content organisation, discoverability, and community formation on social media, yet automated recommendation of contextually appropriate hashtags from multimodal post content—text, images, audio, and video together—remains underexplored as a distinct task. To the best of our targeted search, no existing review synthesises multimodal hashtag recommendation (MHR) specifically. Following PRISMA 2020 [59], we searched IEEE Xplore, ACM Digital Library, Scopus, and SpringerLink for studies published January 2016–September 2025. After independent two-stage screening (Cohen’s κ=0.81 and 0.85 [60]), 30 studies met eligibility and were assessed with a five-criterion quality rubric (mean 5.9±1.8). We classify included studies into six architectural families—early fusion, late fusion, attention-based, graph-enhanced, transformer/LLM/RAG, and an emerging neurosymbolic-contrastive direction—and introduce the Hashtag Intelligence Maturity Model (HIMM), a five-level framework assessing MHR systems by functional capability. Image–text combinations dominate the corpus, with micro-video and retrieval-augmented generation (RAG) as emerging subfields. A reproducibility and fairness audit finds diversity, fairness, and explainability metrics essentially absent from current evaluations, accessible code in only 7% of studies, and multilingual coverage of only 13%. We identify six structural research gaps and a six-directional research agenda addressing foundation-model integration, generative hashtag synthesis, cross-platform benchmarking, multilingual coverage, explainability, and ethical deployment.

Research Square
Universiti Putra Malaysia (MY)
Multimodal Machine Learning Applications
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon