Fine-Grained Length Controllable Video Captioning With Ordinal Embeddings

This paper proposes a method for video captioning that controls the length of generated captions. Previous work on length control often had few levels for expressing length. In this study, we propose two methods of length embedding for fine-grained length control. A traditional embedding method is l...

Full description

Saved in:
Bibliographic Details
Main Authors: Tomoya Nitta, Takumi Fukuzawa, Toru Tamaki
Format: Article
Language:English
Published: IEEE 2024-01-01
Series:IEEE Access
Subjects:
Online Access:https://ieeexplore.ieee.org/document/10767711/
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850052974204682240
author Tomoya Nitta
Takumi Fukuzawa
Toru Tamaki
author_facet Tomoya Nitta
Takumi Fukuzawa
Toru Tamaki
author_sort Tomoya Nitta
collection DOAJ
description This paper proposes a method for video captioning that controls the length of generated captions. Previous work on length control often had few levels for expressing length. In this study, we propose two methods of length embedding for fine-grained length control. A traditional embedding method is linear, using a one-hot vector and an embedding matrix. In this study, we propose methods that represent length in multi-hot vectors. One is bit embedding that expresses length in bit representation, and the other is ordinal embedding that uses the binary representation often used in ordinal regression. These length representations of multi-hot vectors are converted into length embedding by a nonlinear MLP. This method allows for not only the length control of caption sentences but also the control of the time when reading the caption. Experiments using ActivityNet Captions and Spoken Moments in Time show that the proposed method effectively controls the length of the generated captions. Analysis of the embedding vectors with Independent Component Analysis (ICA) shows that length and semantics were learned separately, demonstrating the effectiveness of the proposed embedding methods. Our code and online demo are available at <uri>https://huggingface.co/spaces/fztkm/length_controllable_video_captioning</uri>.
format Article
id doaj-art-6dee2fe280a34124a598531ff97df3d6
institution DOAJ
issn 2169-3536
language English
publishDate 2024-01-01
publisher IEEE
record_format Article
series IEEE Access
spelling doaj-art-6dee2fe280a34124a598531ff97df3d62025-08-20T02:52:41ZengIEEEIEEE Access2169-35362024-01-011218966718968810.1109/ACCESS.2024.350675110767711Fine-Grained Length Controllable Video Captioning With Ordinal EmbeddingsTomoya Nitta0Takumi Fukuzawa1Toru Tamaki2https://orcid.org/0000-0001-9712-7777Toshiba, Kawasaki, JapanNagoya Institute of Technology, Nagoya, JapanNagoya Institute of Technology, Nagoya, JapanThis paper proposes a method for video captioning that controls the length of generated captions. Previous work on length control often had few levels for expressing length. In this study, we propose two methods of length embedding for fine-grained length control. A traditional embedding method is linear, using a one-hot vector and an embedding matrix. In this study, we propose methods that represent length in multi-hot vectors. One is bit embedding that expresses length in bit representation, and the other is ordinal embedding that uses the binary representation often used in ordinal regression. These length representations of multi-hot vectors are converted into length embedding by a nonlinear MLP. This method allows for not only the length control of caption sentences but also the control of the time when reading the caption. Experiments using ActivityNet Captions and Spoken Moments in Time show that the proposed method effectively controls the length of the generated captions. Analysis of the embedding vectors with Independent Component Analysis (ICA) shows that length and semantics were learned separately, demonstrating the effectiveness of the proposed embedding methods. Our code and online demo are available at <uri>https://huggingface.co/spaces/fztkm/length_controllable_video_captioning</uri>.https://ieeexplore.ieee.org/document/10767711/Video captioninglength controllable generationordinal embedding
spellingShingle Tomoya Nitta
Takumi Fukuzawa
Toru Tamaki
Fine-Grained Length Controllable Video Captioning With Ordinal Embeddings
IEEE Access
Video captioning
length controllable generation
ordinal embedding
title Fine-Grained Length Controllable Video Captioning With Ordinal Embeddings
title_full Fine-Grained Length Controllable Video Captioning With Ordinal Embeddings
title_fullStr Fine-Grained Length Controllable Video Captioning With Ordinal Embeddings
title_full_unstemmed Fine-Grained Length Controllable Video Captioning With Ordinal Embeddings
title_short Fine-Grained Length Controllable Video Captioning With Ordinal Embeddings
title_sort fine grained length controllable video captioning with ordinal embeddings
topic Video captioning
length controllable generation
ordinal embedding
url https://ieeexplore.ieee.org/document/10767711/
work_keys_str_mv AT tomoyanitta finegrainedlengthcontrollablevideocaptioningwithordinalembeddings
AT takumifukuzawa finegrainedlengthcontrollablevideocaptioningwithordinalembeddings
AT torutamaki finegrainedlengthcontrollablevideocaptioningwithordinalembeddings