Found in Translation: Learning Robust Joint Representations by Cyclic Translations between Modalities

Authors: Hai Pham, Paul Pu Liang, Thomas Manzini, Louis-Philippe Morency, Barnabás Póczos6892-6899

AAAI 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Additional experiments show that our model learns increasingly discriminative joint representations with more input modalities while maintaining robustness to missing or perturbed modalities. ... it achieves new state-of-the-art results on multimodal sentiment analysis datasets: CMU-MOSI, ICTMMMO, and You Tube.
Researcher Affiliation Academia School of Computer Science, Carnegie Mellon University {htpham,pliang,tmanzini,morency,bapoczos}@cs.cmu.edu
Pseudocode No No pseudocode or algorithm block was found.
Open Source Code Yes 1Our source code is released at https://github.com/hainow/ MCTN.
Open Datasets Yes We use the CMU Multimodal Opinion-level Sentiment Intensity dataset (CMU-MOSI) which contains 2199 video segments... We also run experiments on ICT-MMMO (W ollmer et al. 2013) and You Tube (Morency, Mihalcea, and Doshi 2011) which consist of online review videos annotated for sentiment.
Dataset Splits Yes To be consistent with prior work, we use 52 segments for training, 10 for validation and 31 for testing.
Hardware Specification No No specific hardware details (e.g., GPU/CPU models, memory) were mentioned for running experiments.
Software Dependencies No Mentions GloVe word embeddings, Facet, COVAREP features, and P2FA for alignment, but no specific version numbers for any software dependencies are provided.
Experiment Setup No No specific hyperparameter values (e.g., learning rate, batch size, number of epochs) or detailed training configurations are explicitly provided in the main text.