id
stringlengths
1
4
tokens
sequence
ner_tags
sequence
2000
[ "The", "Transformer", "decoder", "generates", "the", "word", "sequences", "(transcription", "and", "translation)", "for", "ST,", "MT", "and", "ASR", "tasks." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
2001
[ "To", "explain,", "the", "Transformer", "encoder", "further", "extracts", "the", "high-level", "semantic", "hidden", "representation", "of", "two", "modalities." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
2002
[ "The", "Transformer", "encoder", "and", "decoder", "are", "using", "the", "same", "configuration", "as", "the", "original", "(Vaswani", "et", "al.,", "2017)." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
2003
[ "Both", "the", "speech", "encoder", "and", "word", "embedding", "layer", "are", "connect", "to", "Transformer", "encoder", "and", "then", "passed", "to", "the", "Transformer", "decoder." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
2004
[ "It", "is", "the", "same", "as", "word", "embedding", "for", "text", "translation." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
2005
[ "Parallel", "to", "the", "speech", "encoder", "is", "the", "word", "embeeding", "layer." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
2006
[ "Denote", "a", "=", "S-Enc(s)", "as", "the", "audio", "representation", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
2007
[ "In", "total,", "it", "shrinks", "the", "time", "dimension", "by", "a", "factor", "of", "4." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 5 ]
2008
[ "Each", "convolutional", "layer", "has", "a", "stride", "of", "4", "and", "d", "channels." ]
[ 0, 0, 0, 0, 0, 3, 0, 5, 0, 3, 3 ]
2009
[ "The", "input", "is", "raw", "waveform", "signal", "sampled", "at", "16kHz." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
2010
[ "It", "contains", "Wav2vec2.0", "(Baevski", "et", "al.,", "2020)", "and", "two", "additional", "convolutional", "layers." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
2011
[ "The", "speech", "encoder", "module", "(S-Enc)", "is", "designed", "to", "extract", "low-level", "features", "for", "speech", "signals." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
2012
[ "Such", "architecture", "enables", "a", "universal", "framework", "for", "multiple", "tasks,", "including", "ST,", "MT", "and", "ASR." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 11, 0, 0, 0 ]
2013
[ "It", "is", "designed", "to", "take", "either", "speech", "or", "a", "sentence", "as", "input,", "and", "to", "output", "either", "source", "transcript", "or", "target", "translation", "text." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
2014
[ "Our", "model", "consists", "four", "sub-modules:", "a", "speech", "encoder,", "a", "word", "embedding", "layer,", "a", "Transformer", "Encoder", "and", "a", "Transformer", "decoder", "(Figure", "2)." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
2015
[ "We", "use", "the", "same", "model", "architecture", "as", "XSTNet", "(Ye", "et", "al.,", "2021)." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
2016
[ "3.1", "Model", "Framework" ]
[ 0, 0, 0 ]
2017
[ "We", "also", "provide", "several", "feasible", "strategies", "to", "construct", "more", "positive", "and", "negative", "pairs", "to", "enhance", "the", "contrastive", "learning." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
2018
[ "In", "this", "section,", "we", "present", "the", "overall", "speech", "translation", "model", "and", "cross-modal", "contrastive", "learning." ]
[ 0, 0, 0, 0, 0, 0, 0, 11, 12, 0, 0, 1, 2, 2 ]
2019
[ "vated", "by", "the", "contrastive", "learning", "frameworks", "in", "cross-lingual", "and", "cross-modal", "topics,", "we", "introduce", "a", "similar", "idea", "in", "speech", "translation." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 11, 12 ]
2020
[ "Very", "recently,", "contrastive", "learning", "is", "also", "applied", "to", "learning", "a", "unified", "representation", "of", "image", "and", "text", "(Dong", "et", "al.,", "2019;", "Zhou", "et", "al.,", "2020;", "Li", "et", "al.,", "2021)." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
2021
[ "Shen", "et", "al.,", "2020;", "Gao", "et", "al.,", "2021;", "Wu", "et", "al.,", "2021;", "Yan", "et", "al.,", "2021;", "Fu", "et", "al.,", "2022),", "machine", "translation", "(Pan", "et", "al.,", "2021),", "and", "summarization", "(Wang", "et", "al.,", "2021b;", "Cao", "and", "Wang,", "2021)." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
2022
[ "In", "the", "NLP", "area,", "the", "contrastive", "framework", "is", "used", "for", "sentence", "representation", "learning", "(Fang", "et", "al.,", "2020;" ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
2023
[ "In", "speech", "processing,", "representative", "methods", "focused", "on", "speaker", "identification", "(Ravanelli", "and", "Bengio,", "2018),", "speech", "recognition", "(Schneider", "et", "al.,", "2019),", "and", "audio", "representation", "learning", "(van", "den", "Oord", "et", "al.,", "2018;", "Baevski", "et", "al.,", "2020)." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
2024
[ "(2020)", "extended", "the", "self-supervised", "batch", "contrastive", "approach", "to", "the", "fully-supervised", "setting", "and", "proposed", "a", "supervised", "contrastive", "learning", "method." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
2025
[ "Oord", "et", "al.,", "2018;", "Chen", "et", "al.,", "2020a;", "Grill", "et", "al.,", "2020)." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
2026
[ "The", "contrastive", "learning", "method", "was", "first", "proposed", "to", "learn", "representations", "from", "unlabeled", "datasets", "(hence", "the", "term,", "self-supervised", "learning)", "by", "telling", "which", "data", "points", "are", "similar", "or", "distinct,", "especially", "in", "the", "field", "of", "computer", "vision", "(Chopra", "et", "al.,", "2005;", "Gutmann", "and", "Hyvärinen,", "2010;", "Schroff", "et", "al.,", "2015;", "Sohn,", "2016;" ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
2027
[ "Contrastive", "learning", "Our", "method", "is", "motivated", "by", "the", "recent", "success", "in", "contrastive", "representation", "learning." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
2028
[ "In", "this", "work,", "we", "consider", "the", "its", "dual", "form,", "i.e.,", "grounding", "speech", "representations", "using", "text." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
2029
[ "These", "works", "usually", "focus", "on", "enhancing", "textual", "representations", "with", "acoustic", "or", "visual", "information,", "in", "other", "words,", "grounding", "learning." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
2030
[ "We", "are", "also", "inspired", "by", "crossmodal", "representation", "learning", "in", "the", "acoustic", "word", "embedding", "(AWE)", "(Palaskar", "et", "al.,", "2019;", "Kamper", "et", "al.,", "2020;", "Hu", "et", "al.,", "2020)", "and", "the", "visuallanguage", "pre-training", "(VLP)", "(Wu", "et", "al.,", "2019;" ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
2031
[ "Cross-modal", "grounding", "learning", "This", "paper", "attempts", "to", "address", "the", "problem", "in", "speech", "translation", "from", "the", "perspective", "of", "cross-speech-text", "representation", "learning." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
2032
[ "So", "in", "this", "paper,", "we", "propose", "a", "simple", "yet", "effective", "contrastive", "learning", "method", "to", "bridge", "the", "gap", "and", "to", "improve", "ST", "performance." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 11, 0 ]
2033
[ "However,", "we", "find", "that", "this", "approach", "actually", "sacrifices", "the", "effect", "of", "MT." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
2034
[ "(Han", "et", "al.,", "2021)", "designed", "a", "fix-size", "semantic", "memory", "module", "to", "bridge", "such", "a", "gap,", "from", "the", "neuroscience", "perspective." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
2035
[ "As", "a", "cross-modal", "task,", "some", "work", "has", "noted", "the", "problem", "of", "the", "modality", "gap." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
2036
[ "2021)", "are", "widely", "applied", "to", "further", "enhance", "the", "robustness", "for", "ST." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 11 ]
2037
[ "Also,", "multi-task", "frameworks", "(Le", "et", "al.,", "2020;", "Tang", "et", "al.,", "2021b;", "Ye", "et", "al.," ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
2038
[ "Xu", "et", "al.", "(2021)", "proposed", "a", "stacked", "acoustic-and-textual", "encoder", "and", "introduced", "largescale", "out-of-domain", "data." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
2039
[ "(2021b)", "first", "proposed", "a", "second", "encoder", "to", "further", "extract", "semantic", "information", "of", "the", "speech", "sequence." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
2040
[ "For", "example,", "Dong", "et", "al." ]
[ 0, 0, 0, 0, 0 ]
2041
[ "Meanwhile,", "some", "work", "has", "shown", "that", "the", "encoderdecoder", "model", "with", "a", "single", "encoder", "cannot", "encode", "speech", "information", "well." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
2042
[ "Pino", "et", "al.,", "2020;", "Chen", "et", "al.,", "2021),", "pre-training", "(Weiss", "et", "al.,", "2017;", "Berard", "et", "al.,", "2018;", "Bansal", "et", "al.,", "2019;", "Wang", "et", "al.,", "2020b;", "Alinejad", "and", "Sarkar,", "2020;", "Dong", "et", "al.,", "2021a;", "Zheng", "et", "al.,", "2021),", "self-training", "(Pino", "et", "al.,", "2020;", "Wang", "et", "al.,", "2021a),", "utilizing", "selfsupervised", "pre-trained", "audio", "representation", "(Wu", "et", "al.,", "2020;", "Han", "et", "al.,", "2021;", "Ye", "et", "al.,", "2021;", "Wang", "et", "al.,", "2021a),", "are", "proved", "to", "be", "effective." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
2043
[ "However,", "training", "an", "end-to-end", "speech", "translation", "model", "is", "difficult", "because", "we", "need", "to", "design", "a", "cross-modal", "cross-language", "model,", "meanwhile,", "the", "speech-transcription-translation", "supervised", "data", "for", "speech", "translation", "is", "significantly", "less", "than", "that", "of", "MT", "and", "ASR." ]
[ 0, 0, 0, 11, 12, 12, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
2044
[ "Kano", "et", "al.", "(2017);", "Berard", "et", "al.", "(2018);", "Inaguma", "et", "al.", "(2020);", "Wang", "et", "al.", "(2020a);", "Zhao", "et", "al.", "(2021a)", "implemented", "several", "off-the-shelf", "encoderdecoder", "E2E-ST", "models,", "such", "as", "BiLSTM", "(Greff", "et", "al.,", "2016)", "and", "Speech-Transformer", "(Dong", "et", "al.,", "2018)." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 11, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0 ]
2045
[ "Weiss", "et", "al.", "(2017)", "proposed", "to", "use", "an", "end-to-end", "architecture", "to", "directly", "translate", "speech", "into", "text", "in", "another", "language,", "without", "the", "intermediate", "transcription." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
2046
[ "End-to-end", "ST", "To", "alleviate", "the", "error", "propagation", "in", "the", "cascaded", "ST", "systems", "and", "to", "make", "the", "deployment", "simpler,", "Bérard", "et", "al." ]
[ 11, 12, 0, 0, 0, 0, 0, 0, 0, 0, 11, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
2047
[ "2", "Related", "Work" ]
[ 0, 0, 0 ]
2048
[ "We", "demonstrate", "that", "ConST", "indeed", "learns", "similar", "representations", "for", "two", "modalities", "and", "better", "retrieves", "text", "with", "speech", "input." ]
[ 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
2049
[ "Our", "experiments", "on", "the", "MuST-C", "benchmark", "show", "that", "ConST", "achieves", "an", "average", "BLEU", "score", "of", "29.4,", "outperforming", "the", "best", "previous", "baselines." ]
[ 0, 0, 0, 0, 13, 0, 0, 0, 1, 0, 0, 0, 7, 0, 0, 9, 0, 0, 0, 0, 0 ]
2050
[ "We", "develop", "ConST", "for", "speech", "translation,", "a", "cross-modal", "contrastive", "learning", "method,", "on", "top", "of", "the", "multi-task", "training", "framework." ]
[ 0, 0, 1, 0, 11, 12, 0, 1, 2, 2, 0, 0, 0, 0, 0, 0, 0, 0 ]
2051
[ "Our", "contributions", "are", "as", "follows." ]
[ 0, 0, 0, 0, 0 ]
2052
[ "On", "the", "other", "hand,", "it", "reduces", "the", "gap", "between", "the", "representations", "of", "speech", "and", "its", "corresponding", "transcription." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
2053
[ "On", "the", "one", "hand,", "our", "model", "inherits", "the", "advantages", "of", "the", "previous" ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
2054
[ "Inspired", "by", "the", "recent", "progress", "of", "contrastive", "learning", "approaches", "in", "cross-lingual", "(Lample", "and", "Conneau,", "2019;", "Pan", "et", "al.,", "2021)", "and", "cross-modal", "vision-and-language", "domains", "(Li", "et", "al.,", "2021;", "Zhou", "et", "al.,", "2020;", "Dong", "et", "al.,", "2019),", "we", "designed", "a", "simple", "contrastive", "learning", "method", "for", "ST", "(ConST)", "to", "learn", "the", "representations", "that", "meet", "the", "aforementioned", "conditions", "explicitly." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 2, 0, 0, 11, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
2055
[ "Nevertheless,", "how", "to", "learn", "unified", "and", "aligned", "speech-text", "representations?" ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
2056
[ "An", "ideal", "representation", "should", "satisfy:", "if", "the", "content", "of", "the", "speech", "and", "transcription", "are", "similar,", "their", "encoded", "representations", "should", "likewise", "be", "close", "to", "each", "other", "(as", "in", "Figure", "1b)." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
2057
[ "Existing", "approaches", "for", "ST", "focus", "on", "using", "additional", "data", "from", "MT", "and", "automatic", "speech", "recognition", "(ASR)." ]
[ 0, 0, 0, 11, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
2058
[ "We", "analyze", "Transformer", "models", "for", "speech", "translation", "and", "observe", "a", "noticeable", "modality", "gap", "between", "encoder", "representations", "of", "speech", "and", "text", "from", "existing", "ST", "models", "(as", "in", "Figure", "1a." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 11, 0, 0, 0, 0, 0 ]
2059
[ "We", "believe", "that", "when", "the", "representation", "of", "audio", "input", "is", "similar", "to", "its", "corresponding", "textual", "representation,", "it", "is", "easier", "for", "information", "to", "transfer", "from", "MT", "to", "ST,", "thus", "improving", "speech", "translation", "performance." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 11, 0, 0, 11, 12, 0 ]
2060
[ "Different", "from", "the", "data", "perspective,", "this", "paper", "investigates", "the", "bottleneck", "of", "E2E", "ST", "from", "the", "neural", "representation", "perspective." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 11, 12, 0, 0, 0, 0, 0 ]
2061
[ "Ye", "et", "al.,", "2021;", "Han", "et", "al.,", "2021)." ]
[ 0, 0, 0, 0, 0, 0, 0, 0 ]
2062
[ "This", "can", "be", "realized", "through", "pretraining", "approaches", "(Zheng", "et", "al.,", "2021;", "Dong", "et", "al.,", "2021b,a)", "or", "multi-task", "training", "frameworks", "(Tang", "et", "al.,", "2021b;" ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
2063
[ "The", "performance", "of", "an", "E2E", "ST", "model", "is", "still", "restricted", "for", "languages", "with", "relatively", "small", "parallel", "data,", "compared", "to", "text", "machine", "translation", "(MT)." ]
[ 0, 0, 0, 0, 11, 12, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
2064
[ "Compared", "with", "the", "conventional", "cascade", "ST", "models,", "E2E", "ST", "models", "have", "achieved", "almost", "comparable", "(Bentivogli", "et", "al.,", "2021)", "or", "even", "superior", "(Ansari", "et", "al.,", "2020;", "Potapczyk", "and", "Przybysz,", "2020;", "Xu", "et", "al.,", "2021)", "performance." ]
[ 0, 0, 0, 0, 0, 0, 0, 11, 12, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
2065
[ "An", "E2E", "ST", "system", "accepts", "audio", "signals", "as", "the", "input", "and", "generates", "the", "target", "translation", "using", "a", "single", "model." ]
[ 0, 11, 12, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
2066
[ "End-to-end", "speech-to-text", "translation", "(E2E", "ST)", "becomes", "important", "in", "many", "internet", "products", "and", "real", "applications." ]
[ 11, 12, 12, 11, 12, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
2067
[ "Introduction" ]
[ 0 ]
2068
[ "com/ReneeYe/ConST." ]
[ 0 ]
2069
[ "Code", "and", "models", "are", "available", "at", "https://github." ]
[ 0, 0, 0, 0, 0, 0, 0 ]
2070
[ "The", "analysis", "further", "verifies", "that", "ConST", "indeed", "closes", "the", "representation", "gap", "of", "different", "modalities", "—", "its", "learned", "representation", "improves", "the", "accuracy", "of", "cross-modal", "speechtext", "retrieval", "from", "4%", "to", "88%." ]
[ 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 7, 0, 11, 12, 12, 0, 9, 0, 9 ]
2071
[ "We", "evaluate", "ConST", "and", "a", "variety", "of", "previous", "baselines", "on", "a", "popular", "benchmark", "MuST-C.", "Experiments", "show", "that", "the", "proposed", "ConST", "consistently", "outperforms", "the", "previous", "methods,", "and", "achieves", "an", "average", "BLEU", "of", "29.4." ]
[ 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 13, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 7, 0, 9 ]
2072
[ "To", "this", "end,", "we", "propose", "ConST,", "a", "cross-modal", "contrastive", "learning", "method", "for", "end-to-end", "speech-to-text", "translation." ]
[ 0, 0, 0, 0, 0, 1, 0, 1, 2, 2, 0, 0, 11, 12, 12 ]
2073
[ "Learning", "similar", "representations", "for", "semantically", "similar", "speech", "and", "text", "is", "important", "for", "speech", "translation." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 11, 12 ]
2074
[ "How", "can", "we", "learn", "unified", "representations", "for", "spoken", "utterances", "and", "their", "written", "text?" ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
2075
[ "Abstract" ]
[ 0 ]
2076
[ "Cross-modal", "Contrastive", "Learning", "for", "Speech", "Translation" ]
[ 1, 2, 2, 0, 11, 12 ]
2077
[ "References" ]
[ 0 ]
2078
[ "Acknowledgements" ]
[ 0 ]
2079
[ "Although", "recent", "works", "on", "automatic", "selection", "of", "prompts", "and", "label", "mappings", "are", "making", "meaningful", "contribution", "to", "the", "practicability", "of", "few-shot", "learning,", "we", "believe", "more", "work", "should", "be", "done", "to", "simplify", "the", "learning", "procedure", "and", "eliminate", "human", "effort", "while", "achieving", "good", "performance." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
2080
[ "As", "shown", "in", "Figure", "2,", "64", "examples", "per", "class", "are", "enough", "to", "bring", "the", "performance", "of", "standard", "fine-tuning", "to", "the", "same", "level", "of", "prompting." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 2, 0, 0, 0, 0, 0, 0 ]
2081
[ "For", "most", "techniques,", "the", "required", "effort", "for", "finding", "good", "templates", "and", "label", "mappings,", "and", "sometimes", "training", "models", "outweighs", "the", "cost", "of", "simply", "labeling", "more", "training", "examples." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
2082
[ "In", "addition,", "complicated", "prompting", "techniques", "are", "not", "practically", "useful", "for", "real-world", "scenarios." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
2083
[ "Also,", "in-context", "learning", "cannot", "handle", "more", "training", "examples", "due", "to", "the", "maximum", "length", "limit", "of", "the", "model", "while", "AMuLaP", "without", "fine-tuning", "gets", "saturated", "quickly,", "as", "shown", "in", "Figure", "2." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
2084
[ "However,", "as", "shown", "in", "our", "Table", "2,", "with", "the", "same", "backbone", "model", ",", "GPT-3-style", "in-context", "learning", "and", "other", "methods", "that", "do", "not", "access", "the", "model", "weights", "generally", "underperform", "those", "with", "access", "to", "the", "model", "weights", "by", "a", "large", "margin." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 2, 2, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
2085
[ "Brown", "et", "al.", "(2020)", "highlights", "few-shot", "prompting", "as", "a", "way", "to", "mitigate", "their", "decision", "to", "not", "release", "the", "model", "weights." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
2086
[ "There", "is", "a", "natural", "contradiction", "between", "performance", "and", "access", "to", "the", "model", "weights." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
2087
[ "More", "broadly,", "we", "would", "like", "to", "point", "out", "some", "limitation", "and", "contradictions", "within", "current", "fewshot", "prompting", "techniques." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
2088
[ "M" ]
[ 0 ]
2089
[ "Jointly", "searching", "for", "M", "be", "a", "promising", "direction", "for", "future", "research." ]
[ 0, 0, 0, 3, 0, 0, 0, 0, 0, 0, 0 ]
2090
[ "However,", "these", "two", "variables", "are", "closely", "related", "and", "greedily", "search", "for", "the", "best", "may", "be", "then", "the", "best", "mapping", "under", "template", "T", "could", "and", "suboptimal." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
2091
[ "Similar", "to", "our", "paper,", "previous", "works", "(Schick", "et", "al.,", "2020;", "Gao", "et", "al.,", "2021)", "and", "the", "separately", "search", "for", "a", "prompt", "template", "label", "mapping", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
2092
[ "There", "is", "more", "to", "explore", "when", "considering", "the", "prompt", "template", "at", "the", "same", "time." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
2093
[ "In", "this", "paLimitations", "and", "Future", "Directions", "per,", "we", "only", "focus", "on", "the", "selection", "of", "the", "label", "mapping", "with", "a", "fixed", "prompt", "template." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
2094
[ "Thus,", "for", "the", "optimal", "performance,", "we", "find", "it", "essential", "to", "search", "k", "with", "a", "development", "set." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 3, 0, 0, 0, 0 ]
2095
[ "In", "general,", "we", "do", "not", "observe", "a", "clear", "law", "for", "choosing", "the", "best", "k", "for", "AMuLaP.", "As", "mentioned", "before,", "k", "can", "influence", "both", "the", "overall", "quality", "of", "labels", "(in", "both", "ways)", "and", "the", "training", "procedure", "(for", "fine-tuning)." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 3, 4, 1, 2, 0, 0, 0, 3, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
2096
[ "of", "a", "larger", "k." ]
[ 0, 0, 0, 3 ]
2097
[ "However,", "we", "do", "not", "observe", "significant", "improvement", "when", "continuing", "increasing", "k", "with", "labels", "selected", "by", "AMuLaP.", "As", "we", "analyze,", "increasing", "k", "harms", "the", "overall", "quality", "of", "selected", "labels", "and", "thus", "overrides", "the", "benefit" ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 3, 0, 0, 0, 0, 1, 0, 0, 0, 0, 3, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
2098
[ "This", "suggests", "our", "guess", "may", "be", "correct." ]
[ 0, 0, 0, 0, 0, 0, 0 ]
2099
[ "We", "find", "that", "for", "random", "mapping,", "more", "labels", "(i.e.,", "a", "larger", "k)", "often", "leads", "to", "better", "performance." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 3, 0, 0, 0, 0, 0 ]