The DOI matches its Crossref record; 36 of the 36 references it cites were found in a registry or an index.
Multimodal Transformer for Unaligned Multimodal Language Sequences
Open accessAbstract
Human language is often multimodal, which comprehends a mixture of natural language, facial gestures, and acoustic behaviors. However, two major challenges in modeling such multimodal human language time-series data exist: 1) inherent data non-alignment due to variable sampling rates for the sequences from each modality; and 2) long-range dependencies between elements across modalities. In this paper, we introduce the Multimodal Transformer (MulT) to generically address the above issues in an end-to-end manner without explicitly aligning the data. At the heart of our model is the directional pairwise cross-modal attention, which attends to interactions between multimodal sequences across distinct time steps and latently adapt streams from one modality to another. Comprehensive experiments on both aligned and non-aligned multimodal time-series show that our model outperforms state-of-the-art methods by a large margin. In addition, empirical analysis suggests that correlated crossmodal signals are able to be captured by the proposed crossmodal attention mechanism in MulT.
Overview
The claims and statistics of this paper haven’t been checked yet.
Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
The source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“Transformer models [158] have been shown to be useful for temporal multimodal data by automatically aligning and capturing complementary features at different time-steps [154, 174].”
From MultiBench: Multiscale Benchmarks for Multimodal Representation Learning (2021) · cites this paper as [154]
Experimental — accuracy not yet measuredThe cited source says what the claim attributes to it.
In the citing paper
“In MulT, multiple pairwise cross-modal transformers were merged for learning attention across modalities and then fused embedding from different modalities [86].”
From Challenges in AI-driven Biomedical Multimodal Data Fusion and Analysis (2025) · cites this paper as [86]
The cited source says what the claim attributes to it.
In the citing paper
“proposed many effective multimodal fusion methods to improve the emotion recognition performance (Baltruˇ saitis et al., 2018; Tsai et al., 2019; Zhao et al., 2018).”
From Missing Modality Imagination Network for Emotion Recognition with Uncertain Missing Modalities (2021) · cites this paper as Tsai et al., 2019
The cited source says what the claim attributes to it.
In the citing paper
“Previous works have shown that these different modalities are complimentary for emotion expression”
From Missing Modality Imagination Network for Emotion Recognition with Uncertain Missing Modalities (2021) · cites this paper as Tsai et al., 2019
The check wasn’t confident either way, so it gives no verdict — never disputed.
In the citing paper
“On the other hand, lightweight natural language processing (NLP) models such as TinyBERT [232] focus on reducing the size of transformer-based models via knowledge distillation, making them suitable for mobile deployment.”
From Multimodal fusion and vision–language models: A survey for robot vision (2026) · cites this paper as [232]
Experimental — accuracy not yet measured
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Citations
Cites
Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, L., & Polosukhin, I. (2017) · arXiv · doi:10.48550/arxiv.1706.03762 · Cited by 120,404 · cited 16×
“We choose the self-attention transformer (Vaswani et al., 2017).”
BackgroundMethodInfluentialBERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
Devlin, J., Chang, M.-W., Lee, K., & Toutanova, K. (2018) · arXiv · doi:10.48550/arxiv.1810.04805 · Cited by 73,173
BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
Devlin, J., Chang, M.-W., Lee, K., & Toutanova, K. (2019) · Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers) · doi:10.18653/v1/n19-1423 · Cited by 54,325 · cited 2×
“Following prior works on transformers (Vaswani et al., 2017; Chen et al., 2018; Devlin et al., 2018; Dai et al., 2018), we add a residual connection to the crossmodal attention computation.”
BackgroundMethodGlove: Global Vectors for Word Representation
Pennington, J., Socher, R., & Manning, C. (2014) · Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing (EMNLP) · doi:10.3115/v1/d14-1162 · Cited by 34,071 · cited 1×
“We convert video transcripts into pre-trained Glove word embeddings (glove.840B.300d) (Pennington et al., 2014).”
MethodLayer NormalizationPreprint
Ba, J., Kiros, J., & Hinton, G. E. (2016) · arXiv.org · doi:10.48550/arxiv.1607.06450 · Cited by 13,762 · cited 1×
“LN means layer normalization (Ba et al., 2016).”
Wang, X., Girshick, R., Gupta, A., & He, K. (2018) · 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2018.00813 · Cited by 11,596
An argument for basic emotions
Ekman, P. (1992) · Cognition and Emotion · doi:10.1080/02699939208411068 · Cited by 9,646 · cited 1×
“We use Facet (iMotions, 2017) to indicate 35 facial action units, which records facial muscle movement (Ekman et al., 1980; Ekman, 1992) for representing per-frame basic and advanced emotions.”
MethodThe Stanford CoreNLP Natural Language Processing Toolkit
Manning, C. D., Surdeanu, M., Bauer, J., Finkel, J. R., Bethard, S. J., & McClosky, D. (2014) · Annual Meeting of the Association for Computational Linguistics: System Demonstrations (ACL Demo) · doi:10.3115/v1/p14-5010 · Cited by 7,252 · cited 1×
“This rich information provides us the benefit of understanding human behaviors and intents (Manning et al., 2014).”
BackgroundGraves, A., Fernández, S., Gomez, F., & Schmidhuber, J. (2006) · Proceedings of the 23rd international conference on Machine learning - ICML '06 · doi:10.1145/1143844.1143891 · Cited by 5,370
IEMOCAP: Interactive emotional dyadic motion capture database
Busso, C., Bulut, M., Lee, C.-C., Kazemzadeh, A., Mower, E., Kim, S., Chang, J. N., Lee, S., & Narayanan, S. S. (2008) · Language Resources and Evaluation · doi:10.1007/s10579-008-9076-6 · Cited by 3,869
(2011) · Cited by 2,752
Self-Attention with Relative Position Representations
Shaw, P., Uszkoreit, J., & Vaswani, A. (2018) · Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 2 (Short Papers) · doi:10.18653/v1/n18-2074 · Cited by 2,505 · cited 1×
“…crossmodal attention block adapts directly from the low-level feature sequence (i.e., Z [0] β in Figure 3(b)) and does not rely on self-attention, which makes it different from the NMT encoder-decoder architecture (Vaswani et al., 2017; Shaw et al., 2018) (i.e., taking intermediate-level features).”
BackgroundA Structured Self-attentive Sentence Embedding
Lin, Z., Feng, M., Santos, C. N. D., Yu, M., Xiang, B., Zhou, B., & Bengio, Y. (2017) · arXiv · doi:10.48550/arxiv.1703.03130 · Cited by 1,743 · cited 1×
“Transformer network (Vaswani et al., 2017) was first introduced for neural machine translation (NMT) tasks, where the encoder and decoder side each leverages a self-attention (Parikh et al., 2016; Lin et al., 2017; Vaswani et al., 2017) transformer.”
Multimodal Language Analysis in the Wild: CMU-MOSEI Dataset and Interpretable Dynamic Fusion Graph
Bagher Zadeh, A., Liang, P. P., Poria, S., Cambria, E., & Morency, L.-P. (2018) · Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · doi:10.18653/v1/p18-1208 · Cited by 1,433 · cited 12×
“Both tasks are frequently used to benchmark models’ ability to fuse multimodal (sentiment) information (Poria et al., 2017; Zadeh et al., 2018a; Liang et al., 2018; Tsai et al., 2019; Pham et al., 2019; Wang et al., 2019).”
BackgroundMethodInfluentialSelf-citationContext-Dependent Sentiment Analysis in User-Generated Videos
Poria, S., Cambria, E., Hazarika, D., Majumder, N., Zadeh, A., & Morency, L.-P. (2017) · Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · doi:10.18653/v1/p17-1081 · Cited by 939 · cited 3×
“Both tasks are frequently used to benchmark models’ ability to fuse multimodal (sentiment) information (Poria et al., 2017; Zadeh et al., 2018a; Liang et al., 2018; Tsai et al., 2019; Pham et al., 2019; Wang et al., 2019).”
BackgroundMethodSelf-citationFacial signs of emotional experience.
Ekman, P., Freisen, W. V., & Ancoli, S. (1980) · Journal of Personality and Social Psychology · doi:10.1037/h0077722 · Cited by 915 · cited 1×
“We use Facet (iMotions, 2017) to indicate 35 facial action units, which records facial muscle movement (Ekman et al., 1980; Ekman, 1992) for representing per-frame basic and advanced emotions.”
MethodMultimodal Sentiment Intensity Analysis in Videos: Facial Gestures and Verbal Messages
Zadeh, A., Zellers, R., Pincus, E., & Morency, L.-P. (2016) · IEEE Intelligent Systems · doi:10.1109/mis.2016.94 · Cited by 852 · cited 6×
“Meanwhile, CMU-MOSEI (Zadeh et al., 2018b) is a sentiment and emotion analysis dataset made up of 23,454 movie review video clips taken from YouTube (about 10 the size of CMU-MOSI).”
BackgroundMethodInfluentialSelf-citationTools, Language and Cognition in Human Evolution.
Latour, B., Gibson, K. R., & Ingold, T. (1994) · Man · doi:10.2307/2804508 · Cited by 823
COVAREP — A collaborative voice analysis repository for speech technologies
Degottex, G., Kane, J., Drugman, T., Raitio, T., & Scherer, S. (2014) · 2014 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) · doi:10.1109/icassp.2014.6853739 · Cited by 695
Speaker identification on the SCOTUS corpus
Yuan, J., & Liberman, M. (2008) · The Journal of the Acoustical Society of America · doi:10.1121/1.2935783 · Cited by 610 · cited 3×
“, 2019), we first use P2FA (Yuan and Liberman, 2008) to obtain the aligned timesteps (segmented w.”
MethodPham, H., Liang, P. P., Manzini, T., Morency, L.-P., & Póczos, B. (2019) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v33i01.33016892 · Cited by 512
Self-citationWords Can Shift: Dynamically Adjusting Word Representations Using Nonverbal Behaviors
Wang, Y., Shen, Y., Liu, Z., Liang, P. P., Zadeh, A., & Morency, L. (2019) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v33i01.33017216 · Cited by 501
Self-citationThe Best of Both Worlds: Combining Recent Advances in Neural Machine Translation
Chen, M. X., Fırat, O., Bapna, A., Johnson, M., Macherey, W., Foster, G. F., Jones, L., Schuster, M., Shazeer, N., Parmar, N., et al. (2018) · Proceedings of the conference - Association for Computational Linguistics. Meeting · doi:10.18653/v1/p18-1008 · Cited by 484
Linguistically-Informed Self-Attention for Semantic Role Labeling
Strubell, E., Verga, P. W., Andor, D., Weiss, D., & McCallum, A. (2018) · Conference on Empirical Methods in Natural Language Processing (EMNLP) · doi:10.18653/v1/d18-1548 · Cited by 426 · cited 1×
“…have also been successfully applied to other tasks, including language modeling (Dai et al., 2018; Baevski and Auli, 2019), semantic role labeling (Strubell et al., 2018), word sense disambiguation (Tang et al., 2018), learning sentence representations (Devlin et al., 2018), and video activity…”
MethodLearning Factorized Multimodal Representations
Tsai, Y.-H. H., Liang, P. P., Zadeh, A., Morency, L.-P., & Salakhutdinov, R. (2018) · arXiv · doi:10.48550/arxiv.1806.06176 · Cited by 323 · cited 10×
“Classical crossmodal alignment, on the other hand, can be expressed as a special (step di-Table (Liang et al., 2018) 38.3 78.4 78.0 0.922 0.681 MFM (Tsai et al., 2019) 36.2 78.1 78.1 0.951 0.662 RAVEN (Wang et al., 2019) 33.2 78.0 76.6 0.915 0.691 MCTN (Pham et al., 2019) 35 (Zadeh et al., 2018b)…”
BackgroundMethodInfluentialSelf-citationCombining Language and Vision with a Multimodal Skip-gram Model
Lazaridou, A., Pham, N. T., & Baroni, M. (2015) · Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (NAACL-HLT) · doi:10.3115/v1/n15-1016 · Cited by 296
Adaptive Input Representations for Neural Language ModelingPreprint
Baevski, A., & Auli, M. (2018) · arXiv (Cornell University) · doi:10.48550/arxiv.1809.10853 · Cited by 255 · cited 1×
“In addition to NMT, transformer networks have also been successfully applied to other tasks, including language modeling (Dai et al., 2018; Baevski and Auli, 2019), semantic role labeling (Strubell et al., 2018), word sense disambiguation (Tang et al., 2018), learning sentence representations…”
MethodMultimodal Language Analysis with Recurrent Multistage Fusion
Liang, P. P., Liu, Z., Bagher Zadeh, A., & Morency, L.-P. (2018) · Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing · doi:10.18653/v1/d18-1014 · Cited by 230 · cited 5×
“Classical crossmodal alignment, on the other hand, can be expressed as a special (step di-Table (Liang et al., 2018) 38.3 78.4 78.0 0.922 0.681 MFM (Tsai et al., 2019) 36.2 78.1 78.1 0.951 0.662 RAVEN (Wang et al., 2019) 33.2 78.0 76.6 0.915 0.691 MCTN (Pham et al., 2019) 35 (Zadeh et al., 2018b)…”
BackgroundMethodInfluentialSelf-citationMultimodal Affective Analysis Using Hierarchical Attention Strategy with Word-Level Alignment
Gu, Y., Yang, K., Fu, S., Chen, S., Li, X., & Marsic, I. (2018) · Proceedings of the conference - Association for Computational Linguistics. Meeting · doi:10.18653/v1/p18-1207 · Cited by 160 · cited 3×
“In comparison, one common way of tackling unaligned multimodal sequence is by forced word-aligning before training (Poria et al., 2017; Zadeh et al., 2018a,b; Tsai et al., 2019; Pham et al., 2019; Gu et al., 2018): manually preprocess the visual and acoustic features by aligning them to the resolution of words.”
BackgroundMethodOnline Segment to Segment Neural Transduction
Yu, L., Buys, J., & Blunsom, P. (2016) · Conference on Empirical Methods in Natural Language Processing (EMNLP) · doi:10.18653/v1/d16-1138 · Cited by 84 · cited 1×
“…0.662 RAVEN (Wang et al., 2019) 33.2 78.0 76.6 0.915 0.691 MCTN (Pham et al., 2019) 35 (Zadeh et al., 2018b) 45.0 76.9 77.0 0.71 0.54 RAVEN (Wang et al., 2019) 50.0 79.1 79.5 0.614 0.662 MCTN (Pham et al., 2019) 49 agonal) crossmodal attention matrix (i.e., mono-tonic attention (Yu et al., 2016)).”
InfluentialWhy Self-Attention? A Targeted Evaluation of Neural Machine Translation ArchitecturesPreprint
Tang, G., Müller, M., Rios, A., & Sennrich, R. (2018) · arXiv (Cornell University) · doi:10.48550/arxiv.1808.08946 · Cited by 57 · cited 1×
“, 2018), word sense disambiguation (Tang et al., 2018), learning sentence representations (Devlin et al.”
BackgroundA Decomposable Attention Model for Natural Language InferencePreprint
Parikh, A. P., Täckström, O., Das, D., et al. (2016) · arXiv (Cornell University) · doi:10.48550/arxiv.1606.01933 · Cited by 57 · cited 1×
“Transformer network (Vaswani et al., 2017) was first introduced for neural machine translation (NMT) tasks, where the encoder and decoder side each leverages a self-attention (Parikh et al., 2016; Lin et al., 2017; Vaswani et al., 2017) transformer.”
Multi-attention Recurrent Network for Human Communication Comprehension
Zadeh, A., Liang, P. P., Poria, S., Vij, P., Cambria, E., & Morency, L. (2018) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v32i1.12024 · Cited by 49
Self-citationTransformer-XL: Language Modeling with Longer-Term Dependency
Dai, Z., Yang, Z., Yang, Y., Cohen, W. W., Carbonell, J., Le, Q. V., & Salakhutdinov, R. (2018) · Cited by 38
Self-citationSentiment Analysis using Imperfect Views from Spoken Language and Acoustic Modalities
Sheikh, I., Dumpala, S. H., Chakraborty, R., & Kopparapu, S. K. (2018) · doi:10.18653/v1/w18-3305 · Cited by 12 · cited 1×
“We first evaluate MulT on the word-aligned sequences — the “home turf” of prior approaches modeling human multi-modal language (Sheikh et al., 2018; Tsai et al., 2019; Pham et al., 2019; Wang et al., 2019).”
BackgroundOnline Segment to Segment Neural TransductionPreprint
Yu, L., Buys, J., & Blunsom, P. (2016) · arXiv (Cornell University) · doi:10.48550/arxiv.1609.08194 · Cited by 9 · cited 1×
“…0.662 RAVEN (Wang et al., 2019) 33.2 78.0 76.6 0.915 0.691 MCTN (Pham et al., 2019) 35 (Zadeh et al., 2018b) 45.0 76.9 77.0 0.71 0.54 RAVEN (Wang et al., 2019) 50.0 79.1 79.5 0.614 0.662 MCTN (Pham et al., 2019) 49 agonal) crossmodal attention matrix (i.e., mono-tonic attention (Yu et al., 2016)).”
Influential
Research works printed but not linked: 3
- Degottex Gilles, Kane John, Drugman Thomas, Raitio Tuomo, and Scherer Stefan. 2014 Covarepa collaborative voice analysis repository for speech technologies. In ICASSP. IEEE.
- Dumpala Sri Harsha, Chakraborty Rupayan, and Kopparapu Sunil Kumar. 2019 Audio-visual fusion for sentiment classification using cross-modal autoencoder. NIPS.
- Srivastava Nitish and Salakhutdinov Ruslan R. 2012 Multimodal learning with deep boltzmann machines. In Advances in neural information processing systems, pages 2222– 2230.
Cited by
Sorted by citations
MISA: Modality-Invariant and -Specific Representations for Multimodal Sentiment Analysis
Hazarika, D., Zimmermann, R., & Poria, S. (2020) · Proceedings of the 28th ACM International Conference on Multimedia · doi:10.1145/3394171.3413678 · Cited by 1,073
Multimodal learning with transformers: A survey
Xu, P., Zhu, X., & Clifton, D. A. (2023) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2023.3275156 · Cited by 1,072
“This kind of hierarchically cross-modal interaction is also widely studied [137], [189], and alleviates the drawback of cross-attention.”
BackgroundYu, W., Xu, H., Yuan, Z., & Wu, J. (2021) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v35i12.17289 · Cited by 816
“In forward-guidance methods, researches are devoted to design interactive modules for capturing the cross-modal information (Zadeh et al. 2018a; Sun et al. 2020; Tsai et al. 2019).”
BackgroundMethodInfluentialIntegrating Multimodal Information in Large Pretrained Transformers
Rahman, W., Hasan, M. K., Lee, S., Zadeh, A. B., Mao, C., Morency, L., & Hoque, M. E. (2020) · Proceedings of the conference - Association for Computational Linguistics. Meeting · doi:10.18653/v1/2020.acl-main.214 · Cited by 778
“MulT (Multimodal Transformer for Unaligned Multimodal Language Sequence) uses three sets of Transformers and combines their output in a late fusion manner to model a multimodal sequence (Tsai et al., 2019).”
BackgroundMethodInfluentialSelf-citationTrusted Multi-View Classification With Dynamic Evidential Fusion
Han, Z., Zhang, C., Fu, H., & Zhou, J. T. (2022) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2022.3171983 · Cited by 698
“For multi-view learning, traditional methods usually assume an equal value for different views [10], [11], [12], [13] or learn a fixed weight factor for each [14], [15], [16], [17], [18].”
Background
Metadata
- Authors
- Tsai, Y.-H. H., Bai, S., Liang, P. P., Kolter, J. Z., Morency, L.-P., & Salakhutdinov, R.
- Affiliations
- Carnegie Mellon University (US)
- Robert Bosch (United States) (US)
- ORCID
- Venue
- Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics
- Published
- 2019
- Volume
- 2019
- Pages
- 6558-6569
- ISSN
- 0736-587X (ISSN-L)
- PMC
- PMC7195022
- OpenAlex
- W2947476638
- Semantic Scholar
- 949fef650da4c41afe6049a183b504b3cc91f4bd
- Cited by
- 2,259
- References
- 33
- Citations by year
- 2026: 570 · 2025: 564 · 2024: 426 · 2023: 291 · 2022: 214 · 2021: 138 · 2020: 52 · 2019: 4
- FWCI
- 32.88
- Type
- conference paper
- Language
- en
- Open access
- hybrid
- doi.org
- Licence
- cc-by
- Locations
- Proceedings of the conference - Association for Computational Linguistics. Meeting (publishedVersion, cc-by)
- PubMed (publishedVersion)
- PubMed Central (acceptedVersion)
- arXiv (Cornell University) (submittedVersion)
- arXiv (Cornell University)
- Funders
- National Science Foundation (1763562, 1750439, 1722822)
- Nvidia
- Defense Advanced Research Projects Agency (FA8750)
- Samsung
- National Institute of Mental Health (R01MH096951)
- Air Force Research Laboratory (FA8750-18-C-0014)
- Topics
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Multimodal Machine Learning Applications
- Physical Sciences › Computer Science › Signal Processing › Time Series Analysis and Forecasting
- Physical Sciences › Computer Science › Signal Processing › Speech and Audio Processing
- Keywords
- multimodal transformer, cross-modal attention, long-range dependencies, facial gestures, acoustic signals, variable sampling rate, end-to-end modeling
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
