utterance_id stringlengths 14 16 | split stringclasses 1
value | audio_path stringlengths 51 55 | task_type stringclasses 4
values | query stringlengths 3 17 | prompt stringlengths 117 375 | labels listlengths 1 5 | response_mode stringclasses 1
value | answer stringlengths 38 237 |
|---|---|---|---|---|---|---|---|---|
3081-166546-0085 | dev-clean | LibriSpeech/dev-clean/3081/166546/3081-166546-0085.flac | lexical_detection | vigour | Listen carefully to the entire clip, then report every instance of the word "vigour" as JSON objects with word, start, and end. Reply with JSON only: a list of objects {"word": <string>, "start": <seconds>, "end": <seconds>} in chronological order. If none, reply []. | [
[
0.64,
1.07,
"vigour"
]
] | offline | [{"word":"vigour","start":0.64,"end":1.07}] |
8297-275155-0000 | dev-clean | LibriSpeech/dev-clean/8297/275155/8297-275155-0000.flac | lexical_detection | arranged | You will hear a complete utterance. When it ends, list every instance of the word "arranged" you heard with word-level timestamps. Reply with JSON only: a list of objects {"word": <string>, "start": <seconds>, "end": <seconds>} in chronological order. If none, reply []. | [
[
3.13,
3.56,
"arranged"
]
] | offline | [{"word":"arranged","start":3.13,"end":3.56}] |
1988-147956-0018 | dev-clean | LibriSpeech/dev-clean/1988/147956/1988-147956-0018.flac | lexical_detection | noticed | Do not answer until the audio ends. Then list each instance of the word "noticed" as {"word", "start", "end"}. If none, reply []. | [
[
0.48,
0.9,
"noticed"
]
] | offline | [{"word":"noticed","start":0.48,"end":0.9}] |
5694-64038-0022 | dev-clean | LibriSpeech/dev-clean/5694/64038/5694-64038-0022.flac | semantic_detection | person_name | Listen to the audio in full, then emit a JSON array of every instance of a person's name with fields word, start, end (seconds). If none, reply []. | [
[
7.44,
7.67,
"jim"
]
] | offline | [{"word":"jim","start":7.44,"end":7.67}] |
652-130737-0012 | dev-clean | LibriSpeech/dev-clean/652/130737/652-130737-0012.flac | semantic_detection | cardinal | When the utterance is complete, report all instances of a number including the spoken word and time span. Reply with JSON only: a list of objects {"word": <string>, "start": <seconds>, "end": <seconds>} in chronological order. If none, reply []. | [
[
2.13,
3.31,
"eighteen ninety three"
]
] | offline | [{"word":"eighteen ninety three","start":2.13,"end":3.31}] |
8842-302203-0008 | dev-clean | LibriSpeech/dev-clean/8842/302203/8842-302203-0008.flac | semantic_detection | ordinal | Once the audio has finished, identify every instance of an ordinal number. Return a JSON list of {"word", "start", "end"} in time order. If none, reply []. | [
[
0.30000000000000004,
0.59,
"second"
],
[
2.92,
3.2,
"first"
]
] | offline | [{"word":"second","start":0.3,"end":0.59},{"word":"first","start":2.92,"end":3.2}] |
6345-93306-0025 | dev-clean | LibriSpeech/dev-clean/6345/93306/6345-93306-0025.flac | semantic_detection | money | When the utterance is complete, report all instances of a money or currency term including the spoken word and time span. Reply with JSON only: a list of objects {"word": <string>, "start": <seconds>, "end": <seconds>} in chronological order. If none, reply []. | [
[
8.74,
9.05,
"money"
]
] | offline | [{"word":"money","start":8.74,"end":9.05}] |
2428-83705-0035 | dev-clean | LibriSpeech/dev-clean/2428/83705/2428-83705-0035.flac | lexical_detection | given | Do not answer until the audio ends. Then list each instance of the word "given" as {"word", "start", "end"}. If none, reply []. | [
[
3.21,
3.5,
"given"
]
] | offline | [{"word":"given","start":3.21,"end":3.5}] |
652-130737-0002 | dev-clean | LibriSpeech/dev-clean/652/130737/652-130737-0002.flac | lexical_detection | clarets | Listen carefully to the entire clip, then report every instance of the word "clarets" as JSON objects with word, start, and end. Reply with JSON only: a list of objects {"word": <string>, "start": <seconds>, "end": <seconds>} in chronological order. If none, reply []. | [
[
1.72,
2.19,
"clarets"
]
] | offline | [{"word":"clarets","start":1.72,"end":2.19}] |
8842-304647-0006 | dev-clean | LibriSpeech/dev-clean/8842/304647/8842-304647-0006.flac | semantic_detection | body_part | You will hear a complete utterance. When it ends, list every instance of a body part you heard with word-level timestamps. Reply with JSON only: a list of objects {"word": <string>, "start": <seconds>, "end": <seconds>} in chronological order. If none, reply []. | [
[
1.9,
2.36,
"brain"
],
[
7.67,
8.16,
"hands"
]
] | offline | [{"word":"brain","start":1.9,"end":2.36},{"word":"hands","start":7.67,"end":8.16}] |
6313-76958-0004 | dev-clean | LibriSpeech/dev-clean/6313/76958/6313-76958-0004.flac | semantic_detection | body_part | Process the full audio offline. Find every instance of a body part and give the surface word plus start/end timestamps. Reply with JSON only: a list of objects {"word": <string>, "start": <seconds>, "end": <seconds>} in chronological order. If none, reply []. | [
[
3.66,
3.93,
"head"
]
] | offline | [{"word":"head","start":3.66,"end":3.93}] |
7976-105575-0013 | dev-clean | LibriSpeech/dev-clean/7976/105575/7976-105575-0013.flac | lexical_detection | center | Listen carefully to the entire clip, then report every instance of the word "center" as JSON objects with word, start, and end. Reply with JSON only: a list of objects {"word": <string>, "start": <seconds>, "end": <seconds>} in chronological order. If none, reply []. | [
[
6.79,
7.2,
"center"
]
] | offline | [{"word":"center","start":6.79,"end":7.2}] |
3000-15664-0034 | dev-clean | LibriSpeech/dev-clean/3000/15664/3000-15664-0034.flac | lexical_detection | volume | After the audio ends, extract all occurrences of the word "volume". Output each spoken word with its start and end time. Reply with JSON only: a list of objects {"word": <string>, "start": <seconds>, "end": <seconds>} in chronological order. If none, reply []. | [
[
7.01,
7.57,
"volume"
]
] | offline | [{"word":"volume","start":7.01,"end":7.57}] |
6319-57405-0004 | dev-clean | LibriSpeech/dev-clean/6319/57405/6319-57405-0004.flac | semantic_detection | kinship | Do not answer until the audio ends. Then list each instance of a family relation as {"word", "start", "end"}. If none, reply []. | [
[
2.34,
2.64,
"son"
]
] | offline | [{"word":"son","start":2.34,"end":2.64}] |
174-168635-0012 | dev-clean | LibriSpeech/dev-clean/174/168635/174-168635-0012.flac | semantic_detection | cardinal | After the audio ends, extract all occurrences of a number. Output each spoken word with its start and end time. Reply with JSON only: a list of objects {"word": <string>, "start": <seconds>, "end": <seconds>} in chronological order. If none, reply []. | [
[
1.31,
1.6400000000000001,
"six"
]
] | offline | [{"word":"six","start":1.31,"end":1.64}] |
2035-147960-0013 | dev-clean | LibriSpeech/dev-clean/2035/147960/2035-147960-0013.flac | semantic_detection | person_name | Listen to the full audio. After it finishes, find every instance of a person's name. For each match, report the spoken word and its start/end times in seconds. Reply with JSON only: a list of objects {"word": <string>, "start": <seconds>, "end": <seconds>} in chronological order. If none, reply []. | [
[
0.34,
1.1400000000000001,
"otto fuchs"
]
] | offline | [{"word":"otto fuchs","start":0.34,"end":1.14}] |
2803-154328-0023 | dev-clean | LibriSpeech/dev-clean/2803/154328/2803-154328-0023.flac | lexical_detection | kill | After listening to the whole recording, output every instance of the word "kill" with precise start and end times (seconds). Reply with JSON only: a list of objects {"word": <string>, "start": <seconds>, "end": <seconds>} in chronological order. If none, reply []. | [
[
5.46,
5.85,
"kill"
]
] | offline | [{"word":"kill","start":5.46,"end":5.85}] |
5895-34615-0016 | dev-clean | LibriSpeech/dev-clean/5895/34615/5895-34615-0016.flac | lexical_counting | called | Process the full audio offline. For every occurrence of the word "called", emit a count-word event with timestamps. Reply with JSON only: a list of objects {"word": <string>, "start": <seconds>, "end": <seconds>} in chronological order. If none, reply []. | [
[
10.42,
10.75,
"one"
],
[
13.03,
13.38,
"two"
]
] | offline | [{"word":"one","start":10.42,"end":10.75},{"word":"two","start":13.03,"end":13.38}] |
8297-275154-0027 | dev-clean | LibriSpeech/dev-clean/8297/275154/8297-275154-0027.flac | semantic_detection | person_name | After the audio ends, extract all occurrences of a person's name. Output each spoken word with its start and end time. Reply with JSON only: a list of objects {"word": <string>, "start": <seconds>, "end": <seconds>} in chronological order. If none, reply []. | [
[
5.88,
6.42,
"randal"
]
] | offline | [{"word":"randal","start":5.88,"end":6.42}] |
6319-64726-0010 | dev-clean | LibriSpeech/dev-clean/6319/64726/6319-64726-0010.flac | lexical_detection | awe | When the utterance is complete, report all instances of the word "awe" including the spoken word and time span. Reply with JSON only: a list of objects {"word": <string>, "start": <seconds>, "end": <seconds>} in chronological order. If none, reply []. | [
[
4.22,
4.6,
"awe"
]
] | offline | [{"word":"awe","start":4.22,"end":4.6}] |
3000-15664-0020 | dev-clean | LibriSpeech/dev-clean/3000/15664/3000-15664-0020.flac | lexical_detection | strangely | Listen carefully to the entire clip, then report every instance of the word "strangely" as JSON objects with word, start, and end. Reply with JSON only: a list of objects {"word": <string>, "start": <seconds>, "end": <seconds>} in chronological order. If none, reply []. | [
[
11.94,
12.8,
"strangely"
]
] | offline | [{"word":"strangely","start":11.94,"end":12.8}] |
6313-66125-0024 | dev-clean | LibriSpeech/dev-clean/6313/66125/6313-66125-0024.flac | semantic_detection | food | Listen to the audio in full, then emit a JSON array of every instance of a food name with fields word, start, end (seconds). If none, reply []. | [
[
1.81,
2.3,
"picnic"
]
] | offline | [{"word":"picnic","start":1.81,"end":2.3}] |
1272-135031-0024 | dev-clean | LibriSpeech/dev-clean/1272/135031/1272-135031-0024.flac | semantic_counting | body_part | You will hear a complete utterance. When it ends, produce a running count of a body part: each event is {"word": "one"|"two"|..., "start", "end"}. If none, reply []. | [
[
10,
10.38,
"one"
],
[
13.59,
14.05,
"two"
]
] | offline | [{"word":"one","start":10.0,"end":10.38},{"word":"two","start":13.59,"end":14.05}] |
3170-137482-0040 | dev-clean | LibriSpeech/dev-clean/3170/137482/3170-137482-0040.flac | semantic_detection | animal | Once the audio has finished, identify every instance of an animal name. Return a JSON list of {"word", "start", "end"} in time order. If none, reply []. | [
[
5.87,
6.37,
"devil"
]
] | offline | [{"word":"devil","start":5.87,"end":6.37}] |
2803-154320-0004 | dev-clean | LibriSpeech/dev-clean/2803/154320/2803-154320-0004.flac | semantic_detection | body_part | You will hear a complete utterance. When it ends, list every instance of a body part you heard with word-level timestamps. Reply with JSON only: a list of objects {"word": <string>, "start": <seconds>, "end": <seconds>} in chronological order. If none, reply []. | [
[
10.98,
11.5,
"ears"
]
] | offline | [{"word":"ears","start":10.98,"end":11.5}] |
1988-147956-0020 | dev-clean | LibriSpeech/dev-clean/1988/147956/1988-147956-0020.flac | lexical_detection | blown | After listening to the whole recording, output every instance of the word "blown" with precise start and end times (seconds). Reply with JSON only: a list of objects {"word": <string>, "start": <seconds>, "end": <seconds>} in chronological order. If none, reply []. | [
[
5.18,
5.54,
"blown"
]
] | offline | [{"word":"blown","start":5.18,"end":5.54}] |
84-121550-0015 | dev-clean | LibriSpeech/dev-clean/84/121550/84-121550-0015.flac | lexical_detection | came | Do not answer until the audio ends. Then list each instance of the word "came" as {"word", "start", "end"}. If none, reply []. | [
[
6.6,
6.88,
"came"
]
] | offline | [{"word":"came","start":6.6,"end":6.88}] |
8842-302201-0014 | dev-clean | LibriSpeech/dev-clean/8842/302201/8842-302201-0014.flac | semantic_counting | body_part | After the audio ends, tally every instance of a body part. Reply with JSON only — one object per occurrence, word="one"/"two"/... plus start and end in seconds. If none, reply []. | [
[
4.77,
5.2,
"one"
],
[
7.41,
7.77,
"two"
],
[
10.67,
11.09,
"three"
]
] | offline | [{"word":"one","start":4.77,"end":5.2},{"word":"two","start":7.41,"end":7.77},{"word":"three","start":10.67,"end":11.09}] |
8842-304647-0009 | dev-clean | LibriSpeech/dev-clean/8842/304647/8842-304647-0009.flac | lexical_detection | organ | Listen carefully to the entire clip, then report every instance of the word "organ" as JSON objects with word, start, and end. Reply with JSON only: a list of objects {"word": <string>, "start": <seconds>, "end": <seconds>} in chronological order. If none, reply []. | [
[
0.53,
0.84,
"organ"
]
] | offline | [{"word":"organ","start":0.53,"end":0.84}] |
7850-73752-0003 | dev-clean | LibriSpeech/dev-clean/7850/73752/7850-73752-0003.flac | semantic_detection | person_name | Do not answer until the audio ends. Then list each instance of a person's name as {"word", "start", "end"}. If none, reply []. | [
[
7.54,
8.66,
"ferdinand armine"
],
[
11.69,
12.3,
"mirabel"
]
] | offline | [{"word":"ferdinand armine","start":7.54,"end":8.66},{"word":"mirabel","start":11.69,"end":12.3}] |
3170-137482-0033 | dev-clean | LibriSpeech/dev-clean/3170/137482/3170-137482-0033.flac | semantic_detection | cardinal | Listen to the full audio. After it finishes, find every instance of a number. For each match, report the spoken word and its start/end times in seconds. Reply with JSON only: a list of objects {"word": <string>, "start": <seconds>, "end": <seconds>} in chronological order. If none, reply []. | [
[
0.65,
0.85,
"three"
]
] | offline | [{"word":"three","start":0.65,"end":0.85}] |
1462-170145-0010 | dev-clean | LibriSpeech/dev-clean/1462/170145/1462-170145-0010.flac | lexical_detection | slipped | Listen to the full audio. After it finishes, find every instance of the word "slipped". For each match, report the spoken word and its start/end times in seconds. Reply with JSON only: a list of objects {"word": <string>, "start": <seconds>, "end": <seconds>} in chronological order. If none, reply []. | [
[
1.19,
1.51,
"slipped"
]
] | offline | [{"word":"slipped","start":1.19,"end":1.51}] |
2035-152373-0018 | dev-clean | LibriSpeech/dev-clean/2035/152373/2035-152373-0018.flac | lexical_detection | period | Once the audio has finished, identify every instance of the word "period". Return a JSON list of {"word", "start", "end"} in time order. If none, reply []. | [
[
5,
5.54,
"period"
]
] | offline | [{"word":"period","start":5.0,"end":5.54}] |
2086-149220-0023 | dev-clean | LibriSpeech/dev-clean/2086/149220/2086-149220-0023.flac | lexical_detection | fancy | Do not answer until the audio ends. Then list each instance of the word "fancy" as {"word", "start", "end"}. If none, reply []. | [
[
4.61,
5.15,
"fancy"
]
] | offline | [{"word":"fancy","start":4.61,"end":5.15}] |
84-121123-0009 | dev-clean | LibriSpeech/dev-clean/84/121123/84-121123-0009.flac | semantic_detection | profession | Listen to the audio in full, then emit a JSON array of every instance of a profession with fields word, start, end (seconds). If none, reply []. | [
[
0.89,
1.47,
"assassin"
]
] | offline | [{"word":"assassin","start":0.89,"end":1.47}] |
6241-61943-0004 | dev-clean | LibriSpeech/dev-clean/6241/61943/6241-61943-0004.flac | semantic_detection | animal | After the audio ends, extract all occurrences of an animal name. Output each spoken word with its start and end time. Reply with JSON only: a list of objects {"word": <string>, "start": <seconds>, "end": <seconds>} in chronological order. If none, reply []. | [
[
0.81,
1.19,
"ghost"
]
] | offline | [{"word":"ghost","start":0.81,"end":1.19}] |
1993-147149-0008 | dev-clean | LibriSpeech/dev-clean/1993/147149/1993-147149-0008.flac | lexical_detection | chance | Listen to the full audio. After it finishes, find every instance of the word "chance". For each match, report the spoken word and its start/end times in seconds. Reply with JSON only: a list of objects {"word": <string>, "start": <seconds>, "end": <seconds>} in chronological order. If none, reply []. | [
[
1.21,
1.63,
"chance"
]
] | offline | [{"word":"chance","start":1.21,"end":1.63}] |
422-122949-0025 | dev-clean | LibriSpeech/dev-clean/422/122949/422-122949-0025.flac | semantic_detection | cardinal | Listen to the audio in full, then emit a JSON array of every instance of a number with fields word, start, end (seconds). If none, reply []. | [
[
3.9,
4.1,
"one"
]
] | offline | [{"word":"one","start":3.9,"end":4.1}] |
5895-34622-0013 | dev-clean | LibriSpeech/dev-clean/5895/34622/5895-34622-0013.flac | semantic_counting | person_name | You will hear a complete utterance. When it ends, produce a running count of a person's name: each event is {"word": "one"|"two"|..., "start", "end"}. If none, reply []. | [
[
1.72,
2.23,
"one"
],
[
3.47,
3.95,
"two"
]
] | offline | [{"word":"one","start":1.72,"end":2.23},{"word":"two","start":3.47,"end":3.95}] |
652-129742-0018 | dev-clean | LibriSpeech/dev-clean/652/129742/652-129742-0018.flac | lexical_detection | ice | Once the audio has finished, identify every instance of the word "ice". Return a JSON list of {"word", "start", "end"} in time order. If none, reply []. | [
[
2.13,
2.32,
"ice"
]
] | offline | [{"word":"ice","start":2.13,"end":2.32}] |
2428-83699-0034 | dev-clean | LibriSpeech/dev-clean/2428/83699/2428-83699-0034.flac | semantic_detection | cardinal | Process the full audio offline. Find every instance of a number and give the surface word plus start/end timestamps. Reply with JSON only: a list of objects {"word": <string>, "start": <seconds>, "end": <seconds>} in chronological order. If none, reply []. | [
[
3.59,
3.99,
"several"
]
] | offline | [{"word":"several","start":3.59,"end":3.99}] |
1988-147956-0010 | dev-clean | LibriSpeech/dev-clean/1988/147956/1988-147956-0010.flac | semantic_detection | body_part | You will hear a complete utterance. When it ends, list every instance of a body part you heard with word-level timestamps. Reply with JSON only: a list of objects {"word": <string>, "start": <seconds>, "end": <seconds>} in chronological order. If none, reply []. | [
[
2.81,
3.34,
"eyes"
]
] | offline | [{"word":"eyes","start":2.81,"end":3.34}] |
2428-83705-0022 | dev-clean | LibriSpeech/dev-clean/2428/83705/2428-83705-0022.flac | semantic_detection | animal | Listen to the full audio. After it finishes, find every instance of an animal name. For each match, report the spoken word and its start/end times in seconds. Reply with JSON only: a list of objects {"word": <string>, "start": <seconds>, "end": <seconds>} in chronological order. If none, reply []. | [
[
0.37,
0.8,
"bird"
]
] | offline | [{"word":"bird","start":0.37,"end":0.8}] |
8297-275154-0007 | dev-clean | LibriSpeech/dev-clean/8297/275154/8297-275154-0007.flac | lexical_detection | visit | Listen to the audio in full, then emit a JSON array of every instance of the word "visit" with fields word, start, end (seconds). If none, reply []. | [
[
2.88,
3.2800000000000002,
"visit"
]
] | offline | [{"word":"visit","start":2.88,"end":3.28}] |
6241-61943-0003 | dev-clean | LibriSpeech/dev-clean/6241/61943/6241-61943-0003.flac | semantic_detection | kinship | Do not answer until the audio ends. Then list each instance of a family relation as {"word", "start", "end"}. If none, reply []. | [
[
0.53,
0.79,
"uncle"
]
] | offline | [{"word":"uncle","start":0.53,"end":0.79}] |
6295-244435-0015 | dev-clean | LibriSpeech/dev-clean/6295/244435/6295-244435-0015.flac | semantic_detection | person_name | Do not answer until the audio ends. Then list each instance of a person's name as {"word", "start", "end"}. If none, reply []. | [
[
0.26,
0.98,
"colonel kenton"
]
] | offline | [{"word":"colonel kenton","start":0.26,"end":0.98}] |
6313-76958-0005 | dev-clean | LibriSpeech/dev-clean/6313/76958/6313-76958-0005.flac | semantic_detection | ordinal | After listening to the whole recording, output every instance of an ordinal number with precise start and end times (seconds). Reply with JSON only: a list of objects {"word": <string>, "start": <seconds>, "end": <seconds>} in chronological order. If none, reply []. | [
[
1.6099999999999999,
1.97,
"second"
]
] | offline | [{"word":"second","start":1.61,"end":1.97}] |
2086-149220-0015 | dev-clean | LibriSpeech/dev-clean/2086/149220/2086-149220-0015.flac | semantic_detection | food | When the utterance is complete, report all instances of a food name including the spoken word and time span. Reply with JSON only: a list of objects {"word": <string>, "start": <seconds>, "end": <seconds>} in chronological order. If none, reply []. | [
[
7.28,
7.7,
"food"
]
] | offline | [{"word":"food","start":7.28,"end":7.7}] |
1919-142785-0016 | dev-clean | LibriSpeech/dev-clean/1919/142785/1919-142785-0016.flac | semantic_detection | food | Listen to the full audio. After it finishes, find every instance of a food name. For each match, report the spoken word and its start/end times in seconds. Reply with JSON only: a list of objects {"word": <string>, "start": <seconds>, "end": <seconds>} in chronological order. If none, reply []. | [
[
1.9300000000000002,
2.2800000000000002,
"lemon"
]
] | offline | [{"word":"lemon","start":1.93,"end":2.28}] |
251-118436-0014 | dev-clean | LibriSpeech/dev-clean/251/118436/251-118436-0014.flac | semantic_detection | color | Listen to the audio in full, then emit a JSON array of every instance of a color name with fields word, start, end (seconds). If none, reply []. | [
[
1.72,
2.14,
"golden"
]
] | offline | [{"word":"golden","start":1.72,"end":2.14}] |
777-126732-0034 | dev-clean | LibriSpeech/dev-clean/777/126732/777-126732-0034.flac | semantic_detection | person_name | Listen to the audio in full, then emit a JSON array of every instance of a person's name with fields word, start, end (seconds). If none, reply []. | [
[
2.06,
2.7,
"michaelis"
]
] | offline | [{"word":"michaelis","start":2.06,"end":2.7}] |
1993-147966-0000 | dev-clean | LibriSpeech/dev-clean/1993/147966/1993-147966-0000.flac | semantic_counting | date | Listen to the full audio. Count every occurrence of a date. After it finishes, for each occurrence in order output a JSON object whose word is the count word ("one", "two", ...) with that occurrence's start/end times. Reply with JSON only: a list of objects {"word": <string>, "start": <seconds>, "end": <seconds>} in ch... | [
[
1.26,
1.73,
"one"
],
[
3.3,
4.06,
"two"
]
] | offline | [{"word":"one","start":1.26,"end":1.73},{"word":"two","start":3.3,"end":4.06}] |
174-84280-0013 | dev-clean | LibriSpeech/dev-clean/174/84280/174-84280-0013.flac | lexical_detection | freely | You will hear a complete utterance. When it ends, list every instance of the word "freely" you heard with word-level timestamps. Reply with JSON only: a list of objects {"word": <string>, "start": <seconds>, "end": <seconds>} in chronological order. If none, reply []. | [
[
9.72,
10.23,
"freely"
]
] | offline | [{"word":"freely","start":9.72,"end":10.23}] |
8842-302196-0012 | dev-clean | LibriSpeech/dev-clean/8842/302196/8842-302196-0012.flac | lexical_detection | vita | After the audio ends, extract all occurrences of the word "vita". Output each spoken word with its start and end time. Reply with JSON only: a list of objects {"word": <string>, "start": <seconds>, "end": <seconds>} in chronological order. If none, reply []. | [
[
1.92,
2.21,
"vita"
]
] | offline | [{"word":"vita","start":1.92,"end":2.21}] |
5694-64038-0010 | dev-clean | LibriSpeech/dev-clean/5694/64038/5694-64038-0010.flac | semantic_detection | plant | When the utterance is complete, report all instances of a plant name including the spoken word and time span. Reply with JSON only: a list of objects {"word": <string>, "start": <seconds>, "end": <seconds>} in chronological order. If none, reply []. | [
[
2.41,
2.88,
"grass"
]
] | offline | [{"word":"grass","start":2.41,"end":2.88}] |
2277-149897-0022 | dev-clean | LibriSpeech/dev-clean/2277/149897/2277-149897-0022.flac | semantic_detection | person_name | After the audio ends, extract all occurrences of a person's name. Output each spoken word with its start and end time. Reply with JSON only: a list of objects {"word": <string>, "start": <seconds>, "end": <seconds>} in chronological order. If none, reply []. | [
[
6.22,
6.67,
"carrie"
]
] | offline | [{"word":"carrie","start":6.22,"end":6.67}] |
777-126732-0006 | dev-clean | LibriSpeech/dev-clean/777/126732/777-126732-0006.flac | lexical_detection | added | Listen to the audio in full, then emit a JSON array of every instance of the word "added" with fields word, start, end (seconds). If none, reply []. | [
[
3.27,
3.45,
"added"
]
] | offline | [{"word":"added","start":3.27,"end":3.45}] |
2277-149874-0019 | dev-clean | LibriSpeech/dev-clean/2277/149874/2277-149874-0019.flac | lexical_detection | intervals | Listen to the audio in full, then emit a JSON array of every instance of the word "intervals" with fields word, start, end (seconds). If none, reply []. | [
[
5.15,
5.76,
"intervals"
]
] | offline | [{"word":"intervals","start":5.15,"end":5.76}] |
3081-166546-0028 | dev-clean | LibriSpeech/dev-clean/3081/166546/3081-166546-0028.flac | semantic_detection | cardinal | You will hear a complete utterance. When it ends, list every instance of a number you heard with word-level timestamps. Reply with JSON only: a list of objects {"word": <string>, "start": <seconds>, "end": <seconds>} in chronological order. If none, reply []. | [
[
0.67,
0.9500000000000001,
"one"
]
] | offline | [{"word":"one","start":0.67,"end":0.95}] |
2086-149220-0026 | dev-clean | LibriSpeech/dev-clean/2086/149220/2086-149220-0026.flac | semantic_detection | body_part | After the audio ends, extract all occurrences of a body part. Output each spoken word with its start and end time. Reply with JSON only: a list of objects {"word": <string>, "start": <seconds>, "end": <seconds>} in chronological order. If none, reply []. | [
[
2.2,
2.63,
"eyes"
]
] | offline | [{"word":"eyes","start":2.2,"end":2.63}] |
1673-143396-0004 | dev-clean | LibriSpeech/dev-clean/1673/143396/1673-143396-0004.flac | semantic_counting | person_name | Listen to the audio in full, then emit a JSON array of count events for a person's name (word, start, end). If none, reply []. | [
[
4.5600000000000005,
5.16,
"one"
],
[
12.13,
12.64,
"two"
]
] | offline | [{"word":"one","start":4.56,"end":5.16},{"word":"two","start":12.13,"end":12.64}] |
7976-110124-0024 | dev-clean | LibriSpeech/dev-clean/7976/110124/7976-110124-0024.flac | lexical_detection | vain | After the audio ends, extract all occurrences of the word "vain". Output each spoken word with its start and end time. Reply with JSON only: a list of objects {"word": <string>, "start": <seconds>, "end": <seconds>} in chronological order. If none, reply []. | [
[
0.99,
1.29,
"vain"
]
] | offline | [{"word":"vain","start":0.99,"end":1.29}] |
3000-15664-0028 | dev-clean | LibriSpeech/dev-clean/3000/15664/3000-15664-0028.flac | semantic_detection | plant | Process the full audio offline. Find every instance of a plant name and give the surface word plus start/end timestamps. Reply with JSON only: a list of objects {"word": <string>, "start": <seconds>, "end": <seconds>} in chronological order. If none, reply []. | [
[
1.48,
2.02,
"woods"
]
] | offline | [{"word":"woods","start":1.48,"end":2.02}] |
6295-64301-0027 | dev-clean | LibriSpeech/dev-clean/6295/64301/6295-64301-0027.flac | lexical_detection | confusion | Do not answer until the audio ends. Then list each instance of the word "confusion" as {"word", "start", "end"}. If none, reply []. | [
[
1.85,
2.45,
"confusion"
]
] | offline | [{"word":"confusion","start":1.85,"end":2.45}] |
1462-170142-0020 | dev-clean | LibriSpeech/dev-clean/1462/170142/1462-170142-0020.flac | lexical_detection | bit | After listening to the whole recording, output every instance of the word "bit" with precise start and end times (seconds). Reply with JSON only: a list of objects {"word": <string>, "start": <seconds>, "end": <seconds>} in chronological order. If none, reply []. | [
[
0.43,
0.63,
"bit"
]
] | offline | [{"word":"bit","start":0.43,"end":0.63}] |
3536-8226-0010 | dev-clean | LibriSpeech/dev-clean/3536/8226/3536-8226-0010.flac | lexical_detection | followers | Listen to the audio in full, then emit a JSON array of every instance of the word "followers" with fields word, start, end (seconds). If none, reply []. | [
[
1.5,
2.12,
"followers"
]
] | offline | [{"word":"followers","start":1.5,"end":2.12}] |
2035-152373-0009 | dev-clean | LibriSpeech/dev-clean/2035/152373/2035-152373-0009.flac | semantic_counting | city | Listen to the full audio. Count every occurrence of a city name. After it finishes, for each occurrence in order output a JSON object whose word is the count word ("one", "two", ...) with that occurrence's start/end times. Reply with JSON only: a list of objects {"word": <string>, "start": <seconds>, "end": <seconds>} ... | [
[
1.26,
1.71,
"one"
],
[
5.72,
6.13,
"two"
],
[
16.28,
16.82,
"three"
]
] | offline | [{"word":"one","start":1.26,"end":1.71},{"word":"two","start":5.72,"end":6.13},{"word":"three","start":16.28,"end":16.82}] |
777-126732-0041 | dev-clean | LibriSpeech/dev-clean/777/126732/777-126732-0041.flac | semantic_detection | person_name | Do not answer until the audio ends. Then list each instance of a person's name as {"word", "start", "end"}. If none, reply []. | [
[
0.33,
0.77,
"stevie"
]
] | offline | [{"word":"stevie","start":0.33,"end":0.77}] |
2412-153954-0000 | dev-clean | LibriSpeech/dev-clean/2412/153954/2412-153954-0000.flac | semantic_detection | cardinal | After the audio ends, extract all occurrences of a number. Output each spoken word with its start and end time. Reply with JSON only: a list of objects {"word": <string>, "start": <seconds>, "end": <seconds>} in chronological order. If none, reply []. | [
[
1.62,
1.8,
"one"
]
] | offline | [{"word":"one","start":1.62,"end":1.8}] |
3170-137482-0041 | dev-clean | LibriSpeech/dev-clean/3170/137482/3170-137482-0041.flac | lexical_detection | scruples | When the utterance is complete, report all instances of the word "scruples" including the spoken word and time span. Reply with JSON only: a list of objects {"word": <string>, "start": <seconds>, "end": <seconds>} in chronological order. If none, reply []. | [
[
5.9,
6.66,
"scruples"
]
] | offline | [{"word":"scruples","start":5.9,"end":6.66}] |
6319-64726-0015 | dev-clean | LibriSpeech/dev-clean/6319/64726/6319-64726-0015.flac | semantic_detection | body_part | Listen to the audio in full, then emit a JSON array of every instance of a body part with fields word, start, end (seconds). If none, reply []. | [
[
4.22,
4.55,
"heads"
]
] | offline | [{"word":"heads","start":4.22,"end":4.55}] |
3081-166546-0045 | dev-clean | LibriSpeech/dev-clean/3081/166546/3081-166546-0045.flac | semantic_detection | cardinal | When the utterance is complete, report all instances of a number including the spoken word and time span. Reply with JSON only: a list of objects {"word": <string>, "start": <seconds>, "end": <seconds>} in chronological order. If none, reply []. | [
[
4.99,
5.2,
"one"
]
] | offline | [{"word":"one","start":4.99,"end":5.2}] |
1919-142785-0003 | dev-clean | LibriSpeech/dev-clean/1919/142785/1919-142785-0003.flac | semantic_detection | plant | Listen to the full audio. After it finishes, find every instance of a plant name. For each match, report the spoken word and its start/end times in seconds. Reply with JSON only: a list of objects {"word": <string>, "start": <seconds>, "end": <seconds>} in chronological order. If none, reply []. | [
[
0.5,
1.17,
"long pepper"
],
[
2.46,
2.86,
"black"
]
] | offline | [{"word":"long pepper","start":0.5,"end":1.17},{"word":"black","start":2.46,"end":2.86}] |
2902-9006-0015 | dev-clean | LibriSpeech/dev-clean/2902/9006/2902-9006-0015.flac | semantic_detection | kinship | Do not answer until the audio ends. Then list each instance of a family relation as {"word", "start", "end"}. If none, reply []. | [
[
10.36,
10.77,
"sister"
]
] | offline | [{"word":"sister","start":10.36,"end":10.77}] |
1988-147956-0018 | dev-clean | LibriSpeech/dev-clean/1988/147956/1988-147956-0018.flac | semantic_detection | body_part | You will hear a complete utterance. When it ends, list every instance of a body part you heard with word-level timestamps. Reply with JSON only: a list of objects {"word": <string>, "start": <seconds>, "end": <seconds>} in chronological order. If none, reply []. | [
[
2.44,
2.7800000000000002,
"hands"
]
] | offline | [{"word":"hands","start":2.44,"end":2.78}] |
2277-149897-0004 | dev-clean | LibriSpeech/dev-clean/2277/149897/2277-149897-0004.flac | semantic_detection | disease | Process the full audio offline. Find every instance of a disease name and give the surface word plus start/end timestamps. Reply with JSON only: a list of objects {"word": <string>, "start": <seconds>, "end": <seconds>} in chronological order. If none, reply []. | [
[
4.34,
4.51,
"ill"
]
] | offline | [{"word":"ill","start":4.34,"end":4.51}] |
8297-275155-0030 | dev-clean | LibriSpeech/dev-clean/8297/275155/8297-275155-0030.flac | semantic_detection | person_name | Listen to the full audio. After it finishes, find every instance of a person's name. For each match, report the spoken word and its start/end times in seconds. Reply with JSON only: a list of objects {"word": <string>, "start": <seconds>, "end": <seconds>} in chronological order. If none, reply []. | [
[
1.44,
1.99,
"randal"
]
] | offline | [{"word":"randal","start":1.44,"end":1.99}] |
3576-138058-0028 | dev-clean | LibriSpeech/dev-clean/3576/138058/3576-138058-0028.flac | semantic_detection | person_name | After the audio ends, extract all occurrences of a person's name. Output each spoken word with its start and end time. Reply with JSON only: a list of objects {"word": <string>, "start": <seconds>, "end": <seconds>} in chronological order. If none, reply []. | [
[
0.43,
0.97,
"claudia"
]
] | offline | [{"word":"claudia","start":0.43,"end":0.97}] |
6313-66129-0013 | dev-clean | LibriSpeech/dev-clean/6313/66129/6313-66129-0013.flac | lexical_detection | dog | Listen carefully to the entire clip, then report every instance of the word "dog" as JSON objects with word, start, and end. Reply with JSON only: a list of objects {"word": <string>, "start": <seconds>, "end": <seconds>} in chronological order. If none, reply []. | [
[
2.05,
2.42,
"dog"
]
] | offline | [{"word":"dog","start":2.05,"end":2.42}] |
5895-34622-0016 | dev-clean | LibriSpeech/dev-clean/5895/34622/5895-34622-0016.flac | lexical_detection | ursus | Listen to the full audio. After it finishes, find every instance of the word "ursus". For each match, report the spoken word and its start/end times in seconds. Reply with JSON only: a list of objects {"word": <string>, "start": <seconds>, "end": <seconds>} in chronological order. If none, reply []. | [
[
0.21,
0.61,
"ursus"
]
] | offline | [{"word":"ursus","start":0.21,"end":0.61}] |
251-137823-0005 | dev-clean | LibriSpeech/dev-clean/251/137823/251-137823-0005.flac | lexical_detection | faber | You will hear a complete utterance. When it ends, list every instance of the word "faber" you heard with word-level timestamps. Reply with JSON only: a list of objects {"word": <string>, "start": <seconds>, "end": <seconds>} in chronological order. If none, reply []. | [
[
3.81,
4.27,
"faber"
]
] | offline | [{"word":"faber","start":3.81,"end":4.27}] |
2035-152373-0004 | dev-clean | LibriSpeech/dev-clean/2035/152373/2035-152373-0004.flac | semantic_detection | city | After the audio ends, extract all occurrences of a city name. Output each spoken word with its start and end time. Reply with JSON only: a list of objects {"word": <string>, "start": <seconds>, "end": <seconds>} in chronological order. If none, reply []. | [
[
1.48,
2.03,
"sparta"
]
] | offline | [{"word":"sparta","start":1.48,"end":2.03}] |
6241-61943-0002 | dev-clean | LibriSpeech/dev-clean/6241/61943/6241-61943-0002.flac | lexical_detection | wind | Listen carefully to the entire clip, then report every instance of the word "wind" as JSON objects with word, start, and end. Reply with JSON only: a list of objects {"word": <string>, "start": <seconds>, "end": <seconds>} in chronological order. If none, reply []. | [
[
2.14,
2.5700000000000003,
"wind"
]
] | offline | [{"word":"wind","start":2.14,"end":2.57}] |
5694-64038-0017 | dev-clean | LibriSpeech/dev-clean/5694/64038/5694-64038-0017.flac | semantic_detection | body_part | Listen to the audio in full, then emit a JSON array of every instance of a body part with fields word, start, end (seconds). If none, reply []. | [
[
1.06,
1.47,
"pulse"
],
[
3.7,
4.1,
"heart"
]
] | offline | [{"word":"pulse","start":1.06,"end":1.47},{"word":"heart","start":3.7,"end":4.1}] |
3853-163249-0050 | dev-clean | LibriSpeech/dev-clean/3853/163249/3853-163249-0050.flac | lexical_counting | ready | After the audio ends, tally every instance of the word "ready". Reply with JSON only — one object per occurrence, word="one"/"two"/... plus start and end in seconds. If none, reply []. | [
[
2.17,
2.46,
"one"
],
[
3.6,
3.9699999999999998,
"two"
]
] | offline | [{"word":"one","start":2.17,"end":2.46},{"word":"two","start":3.6,"end":3.97}] |
5536-43363-0008 | dev-clean | LibriSpeech/dev-clean/5536/43363/5536-43363-0008.flac | lexical_detection | relatives | Process the full audio offline. Find every instance of the word "relatives" and give the surface word plus start/end timestamps. Reply with JSON only: a list of objects {"word": <string>, "start": <seconds>, "end": <seconds>} in chronological order. If none, reply []. | [
[
2.84,
3.29,
"relatives"
]
] | offline | [{"word":"relatives","start":2.84,"end":3.29}] |
3576-138058-0022 | dev-clean | LibriSpeech/dev-clean/3576/138058/3576-138058-0022.flac | lexical_counting | thee | After listening to the whole recording, output timed count words for the word "thee" (one, two, three, ...) as JSON. If none, reply []. | [
[
18.71,
18.89,
"one"
],
[
24.13,
24.45,
"two"
]
] | offline | [{"word":"one","start":18.71,"end":18.89},{"word":"two","start":24.13,"end":24.45}] |
174-50561-0000 | dev-clean | LibriSpeech/dev-clean/174/50561/174-50561-0000.flac | lexical_detection | captive | You will hear a complete utterance. When it ends, list every instance of the word "captive" you heard with word-level timestamps. Reply with JSON only: a list of objects {"word": <string>, "start": <seconds>, "end": <seconds>} in chronological order. If none, reply []. | [
[
2.95,
3.52,
"captive"
]
] | offline | [{"word":"captive","start":2.95,"end":3.52}] |
2277-149874-0012 | dev-clean | LibriSpeech/dev-clean/2277/149874/2277-149874-0012.flac | semantic_detection | kinship | You will hear a complete utterance. When it ends, list every instance of a family relation you heard with word-level timestamps. Reply with JSON only: a list of objects {"word": <string>, "start": <seconds>, "end": <seconds>} in chronological order. If none, reply []. | [
[
2.07,
2.47,
"husband"
]
] | offline | [{"word":"husband","start":2.07,"end":2.47}] |
1462-170138-0025 | dev-clean | LibriSpeech/dev-clean/1462/170138/1462-170138-0025.flac | semantic_detection | person_name | Process the full audio offline. Find every instance of a person's name and give the surface word plus start/end timestamps. Reply with JSON only: a list of objects {"word": <string>, "start": <seconds>, "end": <seconds>} in chronological order. If none, reply []. | [
[
3.86,
4.33,
"hilda"
]
] | offline | [{"word":"hilda","start":3.86,"end":4.33}] |
2035-147961-0019 | dev-clean | LibriSpeech/dev-clean/2035/147961/2035-147961-0019.flac | semantic_detection | ordinal | Listen to the audio in full, then emit a JSON array of every instance of an ordinal number with fields word, start, end (seconds). If none, reply []. | [
[
0.49,
0.77,
"first"
]
] | offline | [{"word":"first","start":0.49,"end":0.77}] |
652-129742-0012 | dev-clean | LibriSpeech/dev-clean/652/129742/652-129742-0012.flac | semantic_detection | food | Once the audio has finished, identify every instance of a food name. Return a JSON list of {"word", "start", "end"} in time order. If none, reply []. | [
[
0.51,
1.11,
"salad"
],
[
2.87,
3.34,
"apples"
],
[
5.4,
5.89,
"celery"
],
[
8.39,
8.99,
"walnuts"
]
] | offline | [{"word":"salad","start":0.51,"end":1.11},{"word":"apples","start":2.87,"end":3.34},{"word":"celery","start":5.4,"end":5.89},{"word":"walnuts","start":8.39,"end":8.99}] |
5694-64029-0022 | dev-clean | LibriSpeech/dev-clean/5694/64029/5694-64029-0022.flac | semantic_detection | weather | You will hear a complete utterance. When it ends, list every instance of a weather term you heard with word-level timestamps. Reply with JSON only: a list of objects {"word": <string>, "start": <seconds>, "end": <seconds>} in chronological order. If none, reply []. | [
[
0.67,
1.45,
"hail storm"
]
] | offline | [{"word":"hail storm","start":0.67,"end":1.45}] |
3752-4944-0047 | dev-clean | LibriSpeech/dev-clean/3752/4944/3752-4944-0047.flac | lexical_detection | boy's | When the utterance is complete, report all instances of the word "boy's" including the spoken word and time span. Reply with JSON only: a list of objects {"word": <string>, "start": <seconds>, "end": <seconds>} in chronological order. If none, reply []. | [
[
5.11,
5.49,
"boy's"
]
] | offline | [{"word":"boy's","start":5.11,"end":5.49}] |
6319-275224-0007 | dev-clean | LibriSpeech/dev-clean/6319/275224/6319-275224-0007.flac | lexical_detection | grieved | Do not answer until the audio ends. Then list each instance of the word "grieved" as {"word", "start", "end"}. If none, reply []. | [
[
1.21,
1.67,
"grieved"
]
] | offline | [{"word":"grieved","start":1.21,"end":1.67}] |
3536-8226-0030 | dev-clean | LibriSpeech/dev-clean/3536/8226/3536-8226-0030.flac | lexical_detection | opinion | Do not answer until the audio ends. Then list each instance of the word "opinion" as {"word", "start", "end"}. If none, reply []. | [
[
1.22,
1.76,
"opinion"
]
] | offline | [{"word":"opinion","start":1.22,"end":1.76}] |
2428-83705-0022 | dev-clean | LibriSpeech/dev-clean/2428/83705/2428-83705-0022.flac | lexical_detection | bird | Once the audio has finished, identify every instance of the word "bird". Return a JSON list of {"word", "start", "end"} in time order. If none, reply []. | [
[
0.37,
0.8,
"bird"
]
] | offline | [{"word":"bird","start":0.37,"end":0.8}] |
3000-15664-0040 | dev-clean | LibriSpeech/dev-clean/3000/15664/3000-15664-0040.flac | lexical_detection | wilderness | Do not answer until the audio ends. Then list each instance of the word "wilderness" as {"word", "start", "end"}. If none, reply []. | [
[
3.34,
3.96,
"wilderness"
]
] | offline | [{"word":"wilderness","start":3.34,"end":3.96}] |
6313-66125-0006 | dev-clean | LibriSpeech/dev-clean/6313/66125/6313-66125-0006.flac | semantic_detection | cardinal | Listen carefully to the entire clip, then report every instance of a number as JSON objects with word, start, and end. Reply with JSON only: a list of objects {"word": <string>, "start": <seconds>, "end": <seconds>} in chronological order. If none, reply []. | [
[
1.88,
2.12,
"two"
]
] | offline | [{"word":"two","start":1.88,"end":2.12}] |
6345-93302-0025 | dev-clean | LibriSpeech/dev-clean/6345/93302/6345-93302-0025.flac | lexical_detection | really | Once the audio has finished, identify every instance of the word "really". Return a JSON list of {"word", "start", "end"} in time order. If none, reply []. | [
[
0.44,
0.65,
"really"
]
] | offline | [{"word":"really","start":0.44,"end":0.65}] |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.