Cross-stitched Multi-modal Encoders

In this paper, we propose a novel architecture for multi-modal speech and text input. We combine pretrained speech and text encoders using multi-headed cross-modal attention and jointly fine-tune on the target problem. The resultant architecture can be used for continuous token-level classification...

Full description

Saved in:
Bibliographic Details
Published in:arXiv.org 2022-04
Main Authors: Singla, Karan, Pressel, Daniel, Price, Ryan, Bhargav Srinivas Chinnari, Yeon-Jun, Kim, Bangalore, Srinivas
Format: Article
Language:English
Subjects:
Classification
Coders
Linguistics
Speech
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
cited_by
cites
container_end_page
container_issue
container_start_page
container_title arXiv.org
container_volume
creator Singla, Karan
Pressel, Daniel
Price, Ryan
Bhargav Srinivas Chinnari
Yeon-Jun, Kim
Bangalore, Srinivas
description In this paper, we propose a novel architecture for multi-modal speech and text input. We combine pretrained speech and text encoders using multi-headed cross-modal attention and jointly fine-tune on the target problem. The resultant architecture can be used for continuous token-level classification or utterance-level prediction acting on simultaneous text and speech. The resultant encoder efficiently captures both acoustic-prosodic and lexical information. We compare the benefits of multi-headed attention-based fusion for multi-modal utterance-level classification against a simple concatenation of pre-pooled, modality-specific representations. Our model architecture is compact, resource efficient, and can be trained on a single consumer GPU card.
format article
fullrecord <record><control><sourceid>proquest</sourceid><recordid>TN_cdi_proquest_journals_2653025277</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>2653025277</sourcerecordid><originalsourceid>FETCH-proquest_journals_26530252773</originalsourceid><addsrcrecordid>eNpjYuA0MjY21LUwMTLiYOAtLs4yMDAwMjM3MjU15mRQdi7KLy7WLS7JLEnOSE1R8C3NKcnUzc1PScxRcM1Lzk9JLSrmYWBNS8wpTuWF0twMym6uIc4eugVF-YWlqcUl8Vn5pUV5QKl4IzNTYwMjUyNzc2PiVAEAmdMuPQ</addsrcrecordid><sourcetype>Aggregation Database</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype><pqid>2653025277</pqid></control><display><type>article</type><title>Cross-stitched Multi-modal Encoders</title><source>Publicly Available Content (ProQuest)</source><creator>Singla, Karan ; Pressel, Daniel ; Price, Ryan ; Bhargav Srinivas Chinnari ; Yeon-Jun, Kim ; Bangalore, Srinivas</creator><creatorcontrib>Singla, Karan ; Pressel, Daniel ; Price, Ryan ; Bhargav Srinivas Chinnari ; Yeon-Jun, Kim ; Bangalore, Srinivas</creatorcontrib><description>In this paper, we propose a novel architecture for multi-modal speech and text input. We combine pretrained speech and text encoders using multi-headed cross-modal attention and jointly fine-tune on the target problem. The resultant architecture can be used for continuous token-level classification or utterance-level prediction acting on simultaneous text and speech. The resultant encoder efficiently captures both acoustic-prosodic and lexical information. We compare the benefits of multi-headed attention-based fusion for multi-modal utterance-level classification against a simple concatenation of pre-pooled, modality-specific representations. Our model architecture is compact, resource efficient, and can be trained on a single consumer GPU card.</description><identifier>EISSN: 2331-8422</identifier><language>eng</language><publisher>Ithaca: Cornell University Library, arXiv.org</publisher><subject>Classification ; Coders ; Linguistics ; Speech</subject><ispartof>arXiv.org, 2022-04</ispartof><rights>2022. This work is published under http://creativecommons.org/licenses/by-nc-sa/4.0/ (the “License”). Notwithstanding the ProQuest Terms and Conditions, you may use this content in accordance with the terms of the License.</rights><oa>free_for_read</oa><woscitedreferencessubscribed>false</woscitedreferencessubscribed></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><linktohtml>$$Uhttps://www.proquest.com/docview/2653025277?pq-origsite=primo$$EHTML$$P50$$Gproquest$$Hfree_for_read</linktohtml><link.rule.ids>777,781,25734,36993,44571</link.rule.ids></links><search><creatorcontrib>Singla, Karan</creatorcontrib><creatorcontrib>Pressel, Daniel</creatorcontrib><creatorcontrib>Price, Ryan</creatorcontrib><creatorcontrib>Bhargav Srinivas Chinnari</creatorcontrib><creatorcontrib>Yeon-Jun, Kim</creatorcontrib><creatorcontrib>Bangalore, Srinivas</creatorcontrib><title>Cross-stitched Multi-modal Encoders</title><title>arXiv.org</title><description>In this paper, we propose a novel architecture for multi-modal speech and text input. We combine pretrained speech and text encoders using multi-headed cross-modal attention and jointly fine-tune on the target problem. The resultant architecture can be used for continuous token-level classification or utterance-level prediction acting on simultaneous text and speech. The resultant encoder efficiently captures both acoustic-prosodic and lexical information. We compare the benefits of multi-headed attention-based fusion for multi-modal utterance-level classification against a simple concatenation of pre-pooled, modality-specific representations. Our model architecture is compact, resource efficient, and can be trained on a single consumer GPU card.</description><subject>Classification</subject><subject>Coders</subject><subject>Linguistics</subject><subject>Speech</subject><issn>2331-8422</issn><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2022</creationdate><recordtype>article</recordtype><sourceid>PIMPY</sourceid><recordid>eNpjYuA0MjY21LUwMTLiYOAtLs4yMDAwMjM3MjU15mRQdi7KLy7WLS7JLEnOSE1R8C3NKcnUzc1PScxRcM1Lzk9JLSrmYWBNS8wpTuWF0twMym6uIc4eugVF-YWlqcUl8Vn5pUV5QKl4IzNTYwMjUyNzc2PiVAEAmdMuPQ</recordid><startdate>20220420</startdate><enddate>20220420</enddate><creator>Singla, Karan</creator><creator>Pressel, Daniel</creator><creator>Price, Ryan</creator><creator>Bhargav Srinivas Chinnari</creator><creator>Yeon-Jun, Kim</creator><creator>Bangalore, Srinivas</creator><general>Cornell University Library, arXiv.org</general><scope>8FE</scope><scope>8FG</scope><scope>ABJCF</scope><scope>ABUWG</scope><scope>AFKRA</scope><scope>AZQEC</scope><scope>BENPR</scope><scope>BGLVJ</scope><scope>CCPQU</scope><scope>DWQXO</scope><scope>HCIFZ</scope><scope>L6V</scope><scope>M7S</scope><scope>PIMPY</scope><scope>PQEST</scope><scope>PQQKQ</scope><scope>PQUKI</scope><scope>PRINS</scope><scope>PTHSS</scope></search><sort><creationdate>20220420</creationdate><title>Cross-stitched Multi-modal Encoders</title><author>Singla, Karan ; Pressel, Daniel ; Price, Ryan ; Bhargav Srinivas Chinnari ; Yeon-Jun, Kim ; Bangalore, Srinivas</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-proquest_journals_26530252773</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2022</creationdate><topic>Classification</topic><topic>Coders</topic><topic>Linguistics</topic><topic>Speech</topic><toplevel>online_resources</toplevel><creatorcontrib>Singla, Karan</creatorcontrib><creatorcontrib>Pressel, Daniel</creatorcontrib><creatorcontrib>Price, Ryan</creatorcontrib><creatorcontrib>Bhargav Srinivas Chinnari</creatorcontrib><creatorcontrib>Yeon-Jun, Kim</creatorcontrib><creatorcontrib>Bangalore, Srinivas</creatorcontrib><collection>ProQuest SciTech Collection</collection><collection>ProQuest Technology Collection</collection><collection>Materials Science &amp; Engineering Collection</collection><collection>ProQuest Central (Alumni Edition)</collection><collection>ProQuest Central</collection><collection>ProQuest Central Essentials</collection><collection>ProQuest Central</collection><collection>Technology Collection</collection><collection>ProQuest One Community College</collection><collection>ProQuest Central Korea</collection><collection>SciTech Premium Collection</collection><collection>ProQuest Engineering Collection</collection><collection>Engineering Database</collection><collection>Publicly Available Content (ProQuest)</collection><collection>ProQuest One Academic Eastern Edition (DO NOT USE)</collection><collection>ProQuest One Academic</collection><collection>ProQuest One Academic UKI Edition</collection><collection>ProQuest Central China</collection><collection>Engineering Collection</collection></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext</fulltext></delivery><addata><au>Singla, Karan</au><au>Pressel, Daniel</au><au>Price, Ryan</au><au>Bhargav Srinivas Chinnari</au><au>Yeon-Jun, Kim</au><au>Bangalore, Srinivas</au><format>book</format><genre>document</genre><ristype>GEN</ristype><atitle>Cross-stitched Multi-modal Encoders</atitle><jtitle>arXiv.org</jtitle><date>2022-04-20</date><risdate>2022</risdate><eissn>2331-8422</eissn><abstract>In this paper, we propose a novel architecture for multi-modal speech and text input. We combine pretrained speech and text encoders using multi-headed cross-modal attention and jointly fine-tune on the target problem. The resultant architecture can be used for continuous token-level classification or utterance-level prediction acting on simultaneous text and speech. The resultant encoder efficiently captures both acoustic-prosodic and lexical information. We compare the benefits of multi-headed attention-based fusion for multi-modal utterance-level classification against a simple concatenation of pre-pooled, modality-specific representations. Our model architecture is compact, resource efficient, and can be trained on a single consumer GPU card.</abstract><cop>Ithaca</cop><pub>Cornell University Library, arXiv.org</pub><oa>free_for_read</oa></addata></record>
fulltext fulltext
identifier EISSN: 2331-8422
ispartof arXiv.org, 2022-04
issn 2331-8422
language eng
recordid cdi_proquest_journals_2653025277
source Publicly Available Content (ProQuest)
subjects Classification
Coders
Linguistics
Speech
title Cross-stitched Multi-modal Encoders
url http://sfxeu10.hosted.exlibrisgroup.com/loughborough?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2025-01-18T12%3A00%3A28IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-proquest&rft_val_fmt=info:ofi/fmt:kev:mtx:book&rft.genre=document&rft.atitle=Cross-stitched%20Multi-modal%20Encoders&rft.jtitle=arXiv.org&rft.au=Singla,%20Karan&rft.date=2022-04-20&rft.eissn=2331-8422&rft_id=info:doi/&rft_dat=%3Cproquest%3E2653025277%3C/proquest%3E%3Cgrp_id%3Ecdi_FETCH-proquest_journals_26530252773%3C/grp_id%3E%3Coa%3E%3C/oa%3E%3Curl%3E%3C/url%3E&rft_id=info:oai/&rft_pqid=2653025277&rft_id=info:pmid/&rfr_iscdi=true