2b:null 2c:["$","title",null,{"children":"Speech to Text Fragment Mangling | Tim Garvin | Observable"}] 31:[] 2f:{"id":"3aca578dfe9bf892","slug":"speech-to-text-fragment-mangling","title":"Speech to Text Fragment Mangling","thumbnail":"18205f31f763853931bcdd6b9bb5c307a283674e7392b8ba4a3466f7cc71dc23","access_level":"public","create_time":"$D2019-02-14T20:12:38.029Z","update_time":"$D2019-02-20T14:44:01.716Z","view_time":null,"first_public_time":"$D2019-02-15T17:01:39.651Z","fork":false,"stars":4,"starred":false,"owner":{"type":"team","id":"bd83fd4aa63f66d4","login":"tcgarvin","name":"Tim Garvin","avatar_url":"https://avatars-next.observableusercontent.com/0a401f190f78dde3ab907b150924a42cc24ef3250274037aab2c952c0495c664","tier":"starter_2024"},"version":199,"publish_version":199,"role":null,"authored":false,"authors":[{"type":"individual","id":"3a9edae390009386","login":"tcgarvin","name":"Tim Garvin","avatar_url":"https://avatars-next.observableusercontent.com/0a401f190f78dde3ab907b150924a42cc24ef3250274037aab2c952c0495c664","tier":null}],"members":[{"type":"individual","id":"3a9edae390009386","login":"tcgarvin","name":"Tim Garvin","avatar_url":"https://avatars-next.observableusercontent.com/0a401f190f78dde3ab907b150924a42cc24ef3250274037aab2c952c0495c664","tier":null,"role":"editor"}],"comments":[],"custom_thumbnail":false,"trashed":false,"lapsed":null,"subscribed":null,"forks":1,"collection":null,"type":"notebook","body":{"title":"Speech to Text Fragment Mangling","cells":[{"id":0,"value":"md`# Speech to Text Fragment Mangling\n\n**tl;dr: Go to \"Rendering\" below, and start clicking options.**\n\nThis is a quick attempt at looking over some of the algorithms one could use to roll Azure fragments into more interesting results. Azure Batch Transcription Results are pulled from a source url, and are rendered using a number of algorithms`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","output":"$undefined","database":"$undefined"},{"id":60,"value":"sourceUrl = \"https://gist.githubusercontent.com/tcgarvin/2680b0833033ba1474f48afd030b6f84/raw/105bba3abeee6bec015c9519ffd41e7de92ea49f/sample.json\";","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","output":"$undefined","database":"$undefined"},{"id":190,"value":"md`First, a quick explaination of the algorithms implimented here, and their settings (which you can adjust!)`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","output":"$undefined","database":"$undefined"},{"id":29,"value":"md`### Algorithm 0 - No Grouping\n\nHere we just use the fragments as provided by Azure`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","output":"$undefined","database":"$undefined"},{"id":84,"value":"md`### Algorithm 1 - Group by Characters\n\nHere we grow each group until it surpasses a configurable number of paramters`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","output":"$undefined","database":"$undefined"},{"id":88,"value":"viewof algo1MaxChars = slider({\n min: 10, \n max: 500, \n step: 10, \n value: 100, \n title: \"Target Group Characters\", \n description: \"The algorithm aims for groups of roughly this size\"\n})","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","output":"$undefined","database":"$undefined"},{"id":116,"value":"md`Parameter: Max Chars: ${algo1MaxChars}`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","output":"$undefined","database":"$undefined"},{"id":131,"value":"md`### Algorithm 2 - Group by Time\n\nVery similar to group by characters, but using units of time`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","output":"$undefined","database":"$undefined"},{"id":136,"value":"viewof algo2MaxTime = slider({\n min: 1, \n max: 120, \n step: 1, \n value: 5, \n title: \"Target Group Seconds\", \n description: \"The algorithm aims for groups of roughly this time length\"\n})","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","output":"$undefined","database":"$undefined"},{"id":149,"value":"md`### Algorithm 3 - Seperated by pauses\n\nSeperate groups only by the pauses we find between them. \n\nThis one ends up being a lot more interesting than I expected, but seems hard to tune, and small changes in speaker behavior or Azure audio detection would likely result in drastically different behavior`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","output":"$undefined","database":"$undefined"},{"id":155,"value":"viewof algo3MinPauseLength = slider({\n min: 0.05, \n max: 4, \n step: 0.05, \n value: 0.45, \n title: \"Pause Threshold (seconds)\", \n description: \"Split any time we see this duration of pause\"\n})","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","output":"$undefined","database":"$undefined"},{"id":167,"value":"md`### Algorithm 4 - Group by Time and Allow for Pauses\n\nWhile aiming for segments of a particular time length, also allow pauses of a certain length to cause breaks in the audio.\n\nHere, the tuning problems see in algorithm 3 should be softened by the \"max\" time length. Worse case, you end up with something a lot like algorithm 2, and best case, you get the best of algorithms 2 and 3. \n\nI favor using something like this algorithm, but more study is warranted. More study would be needed before picking this algorithm.`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","output":"$undefined","database":"$undefined"},{"id":172,"value":"viewof algo4MaxTime = slider({\n min: 1, \n max: 120, \n step: 1, \n value: 20, \n title: \"Target Group Seconds\", \n description: \"The algorithm aims for groups of roughly this time length\"\n})","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","output":"$undefined","database":"$undefined"},{"id":170,"value":"viewof algo4MinPauseLength = slider({\n min: 0.05, \n max: 4, \n step: 0.05, \n value: 0.5, \n title: \"Pause Threshold (seconds)\", \n description: \"Split any time we see this duration of pause\"\n})","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","output":"$undefined","database":"$undefined"},{"id":58,"value":"md`# Rendering\n\nChoose the Algorithm you want to use. The algorithm will be parameterized using the inputs above`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","output":"$undefined","database":"$undefined"},{"id":95,"value":"viewof r = radio({\n options: [\n { label: 'Algo 0', value: \"algo0\" },\n { label: 'Algo 1', value: \"algo1\" },\n { label: 'Algo 2', value: \"algo2\" },\n { label: 'Algo 3', value: \"algo3\" },\n { label: 'Algo 4', value: \"algo4\" }\n ],\n value: \"algo4\"\n})","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","output":"$undefined","database":"$undefined"},{"id":52,"value":"\nPreact.render(htm`\n

\n ${chosenAlgo.map((group) => \n htm`

\n ${group.map((fragment) => htm` ${fragment.text} `)}\n

`\n )} \n

\n`)","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","output":"$undefined","database":"$undefined"},{"id":105,"value":"md`### Plumbing`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","output":"$undefined","database":"$undefined"},{"id":12,"value":"resultJson = (await d3.json(sourceUrl))\n .AudioFileResults[0].SegmentResults\n .map((fragment) => {\n let text = fragment.NBest[0].Display;\n let confidence = fragment.NBest[0].Confidence;\n let offset = fragment.Offset;\n let duration = fragment.Duration;\n return {\n text, confidence, offset, duration\n };\n });","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","output":"$undefined","database":"$undefined"},{"id":17,"value":"algo0 = {\n let normalized = []\n let currentGroup = []\n let currentGroupDuration = 0;\n let timeParsed = 0; // Azure gives back measurements in something like a 10th of a nanosecond?\n for (let fragment of resultJson) { \n // Core Rollup Logic\n currentGroup.push(fragment);\n normalized.push(currentGroup);\n \n currentGroup=[];\n \n let timeParsed = fragment.offset + fragment.duration;\n }\n return normalized;\n}","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","output":"$undefined","database":"$undefined"},{"id":81,"value":"algo1 = {\n let normalized = []\n let currentGroup = []\n let currentGroupDuration = 0;\n let timeParsed = 0; // Azure gives back measurements in something like a 10th of a nanosecond?\n for (let fragment of resultJson) { \n // Core Rollup Logic\n currentGroup.push(fragment);\n \n if (currentGroup.reduce((acc, next) => acc + next.text.length, 0) > algo1MaxChars) {\n normalized.push(currentGroup);\n currentGroup=[];\n }\n \n let timeParsed = fragment.offset + fragment.duration;\n }\n \n if (currentGroup.length > 0) {\n normalized.push(currentGroup);\n }\n \n return normalized;\n}","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","output":"$undefined","database":"$undefined"},{"id":139,"value":"algo2 = {\n let normalized = []\n let currentGroup = []\n let currentGroupDuration = 0;\n let timeParsed = 0; // Azure gives back measurements in something like a 10th of a nanosecond?\n for (let fragment of resultJson) { \n // Core Rollup Logic\n currentGroup.push(fragment);\n \n if (currentGroup.reduce((acc, next) => acc + next.duration / 10000000, 0) > algo2MaxTime) {\n normalized.push(currentGroup);\n currentGroup=[];\n }\n \n let timeParsed = fragment.offset + fragment.duration;\n }\n \n if (currentGroup.length > 0) {\n normalized.push(currentGroup);\n }\n \n return normalized;\n}","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","output":"$undefined","database":"$undefined"},{"id":146,"value":"algo3 = {\n let normalized = []\n let currentGroup = []\n let currentGroupDuration = 0;\n let timeParsed = 0; // Azure gives back measurements in something like a 10th of a nanosecond?\n for (let fragment of resultJson) { \n // Core Rollup Logic\n if (timeParsed + algo3MinPauseLength * 10000000 < fragment.offset) {\n normalized.push(currentGroup);\n currentGroup=[];\n }\n \n currentGroup.push(fragment);\n\n timeParsed = fragment.offset + fragment.duration;\n }\n \n if (currentGroup.length > 0) {\n normalized.push(currentGroup);\n }\n \n return normalized;\n}","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","output":"$undefined","database":"$undefined"},{"id":175,"value":"algo4 = {\n let normalized = []\n let currentGroup = []\n let currentGroupDuration = 0;\n let timeParsed = 0; // Azure gives back measurements in something like a 10th of a nanosecond?\n for (let fragment of resultJson) { \n // Core Rollup Logic\n if (timeParsed + algo4MinPauseLength * 10000000 < fragment.offset) {\n normalized.push(currentGroup);\n currentGroup=[];\n }\n \n currentGroup.push(fragment);\n \n if (currentGroup.reduce((acc, next) => acc + next.duration / 10000000, 0) > algo4MaxTime) {\n normalized.push(currentGroup);\n currentGroup=[];\n }\n\n timeParsed = fragment.offset + fragment.duration;\n }\n \n if (currentGroup.length > 0) {\n normalized.push(currentGroup);\n }\n \n return normalized;\n}","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","output":"$undefined","database":"$undefined"},{"id":108,"value":"chosenAlgo = {\n let choices = {algo0, algo1, algo2, algo3, algo4};\n return choices[r];\n}","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","output":"$undefined","database":"$undefined"},{"id":78,"value":"html``","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","output":"$undefined","database":"$undefined"},{"id":69,"value":"md`### Libraries`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","output":"$undefined","database":"$undefined"},{"id":10,"value":"d3 = require(\"d3\");","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","output":"$undefined","database":"$undefined"},{"id":47,"value":"htm = (await require(\"htm@2/dist/htm.umd.js\")).bind(Preact.h)","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","output":"$undefined","database":"$undefined"},{"id":44,"value":"Preact = require(\"preact@8/dist/preact.min.js\").catch(() => window.preact)","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","output":"$undefined","database":"$undefined"},{"id":97,"value":"import {radio, slider} from \"@jashkenas/inputs\"","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","output":"$undefined","database":"$undefined"}],"files":"$W31","stdlib":"1","titling":"auto"}} 30:"$2f:body"