[
  {
    "startMs": 1309,
    "text": "[music]"
  },
  {
    "startMs": 12760,
    "text": ">> Okay, hello everyone."
  },
  {
    "startMs": 16120,
    "text": "So,"
  },
  {
    "startMs": 19400,
    "text": "my name is Midam Kim. I am an ML"
  },
  {
    "startMs": 22560,
    "text": "engineer from ServiceNow and I'll be"
  },
  {
    "startMs": 25600,
    "text": "talking about a linguistic framework for"
  },
  {
    "startMs": 28680,
    "text": "voice AI."
  },
  {
    "startMs": 33240,
    "text": "So,"
  },
  {
    "startMs": 35320,
    "text": "quick background of me so you know where"
  },
  {
    "startMs": 37680,
    "text": "I'm coming from."
  },
  {
    "startMs": 39520,
    "text": "Like I said, I'm an ML engineer at"
  },
  {
    "startMs": 41120,
    "text": "ServiceNow, but I'm also a researcher,"
  },
  {
    "startMs": 43560,
    "text": "lifelong researcher, of speech"
  },
  {
    "startMs": 45600,
    "text": "communication in the wild."
  },
  {
    "startMs": 47800,
    "text": "So, my motto is doing linguistics and"
  },
  {
    "startMs": 51200,
    "text": "what I'm going to be doing today is to"
  },
  {
    "startMs": 54200,
    "text": "hand you that lens of linguistics."
  },
  {
    "startMs": 60440,
    "text": "So, have you experienced voice AI"
  },
  {
    "startMs": 63200,
    "text": "failures?"
  },
  {
    "startMs": 65239,
    "text": "Yeah, like everyone."
  },
  {
    "startMs": 66810,
    "text": ">> [laughter]"
  },
  {
    "startMs": 70240,
    "text": ">> So, I'm going to introduce an example"
  },
  {
    "startMs": 72600,
    "text": "that I experienced myself."
  },
  {
    "startMs": 75200,
    "text": "So, the bot asked me, \"Could you please"
  },
  {
    "startMs": 78000,
    "text": "spell your first name?\""
  },
  {
    "startMs": 80560,
    "text": "And then I slowly start to spell my"
  },
  {
    "startMs": 83040,
    "text": "name."
  },
  {
    "startMs": 84480,
    "text": "Yes, it is m i d a m."
  },
  {
    "startMs": 89320,
    "text": "And the bot says, \"Confirming with you,"
  },
  {
    "startMs": 91600,
    "text": "is it m i d a n?\""
  },
  {
    "startMs": 95800,
    "text": "And then I say, \"No, it is m i d a m.\""
  },
  {
    "startMs": 101200,
    "text": "Um"
  },
  {
    "startMs": 102600,
    "text": "and the bot says,"
  },
  {
    "startMs": 104400,
    "text": "\"Thank you for your correction. Happy to"
  },
  {
    "startMs": 106040,
    "text": "help you today, Madam.\""
  },
  {
    "startMs": 108840,
    "text": "And I then I"
  },
  {
    "startMs": 110040,
    "text": "get slightly annoyed, more annoyed,"
  },
  {
    "startMs": 111960,
    "text": "because my name is Midam, not Madam."
  },
  {
    "startMs": 115240,
    "text": "And then it asked me about, \"Now, what"
  },
  {
    "startMs": 117600,
    "text": "is your account number?"
  },
  {
    "startMs": 119560,
    "text": "And then, I start start getting"
  },
  {
    "startMs": 121080,
    "text": "confused. What is that account number"
  },
  {
    "startMs": 123200,
    "text": "thing?"
  },
  {
    "startMs": 124680,
    "text": "And then,"
  },
  {
    "startMs": 126800,
    "text": "I try to find uh information about that."
  },
  {
    "startMs": 130119,
    "text": "So,"
  },
  {
    "startMs": 131640,
    "text": "which one? Um it must be and I start"
  },
  {
    "startMs": 136520,
    "text": "uh"
  },
  {
    "startMs": 137400,
    "text": "slowly start spelling the account"
  },
  {
    "startMs": 140000,
    "text": "number. So, it is A X 4 5 1."
  },
  {
    "startMs": 145760,
    "text": "And then, I take time because I'm not"
  },
  {
    "startMs": 148880,
    "text": "used to reading this strange number."
  },
  {
    "startMs": 152240,
    "text": "And then, the bot cuts me off."
  },
  {
    "startMs": 154880,
    "text": "And then, it says, I couldn't find your"
  },
  {
    "startMs": 156400,
    "text": "record."
  },
  {
    "startMs": 157560,
    "text": "And then, without even trying, it asked"
  },
  {
    "startMs": 160600,
    "text": "me to repeat that again. Can you please"
  },
  {
    "startMs": 162360,
    "text": "repeat that? And then, I get super"
  },
  {
    "startMs": 164680,
    "text": "annoyed and then, I can say, can I talk"
  },
  {
    "startMs": 166959,
    "text": "to a person?"
  },
  {
    "startMs": 168160,
    "text": "I just don't want to deal with you"
  },
  {
    "startMs": 169280,
    "text": "anymore."
  },
  {
    "startMs": 170680,
    "text": "So, this is a very typical pattern of"
  },
  {
    "startMs": 173120,
    "text": "voice AI, unfortunately, at this point."
  },
  {
    "startMs": 176160,
    "text": "So, I just want to navigate how we can"
  },
  {
    "startMs": 179640,
    "text": "solve this problem"
  },
  {
    "startMs": 181720,
    "text": "with linguistics."
  },
  {
    "startMs": 186120,
    "text": "So, voice AI is booming."
  },
  {
    "startMs": 188160,
    "text": "But users are still often preferring"
  },
  {
    "startMs": 190000,
    "text": "human agents over voice agents."
  },
  {
    "startMs": 193640,
    "text": "How can we mitigate this issue?"
  },
  {
    "startMs": 197000,
    "text": "But in the first place, what are the"
  },
  {
    "startMs": 199239,
    "text": "actual problems?"
  },
  {
    "startMs": 201600,
    "text": "So, I think we can think about a"
  },
  {
    "startMs": 203239,
    "text": "fundamental frame framework to"
  },
  {
    "startMs": 205519,
    "text": "understand this into an architecture of"
  },
  {
    "startMs": 208400,
    "text": "voice AI,"
  },
  {
    "startMs": 209880,
    "text": "which is called linguistics."
  },
  {
    "startMs": 214280,
    "text": "So, as all of us already know,"
  },
  {
    "startMs": 218800,
    "text": "human communication is a joint activity,"
  },
  {
    "startMs": 221200,
    "text": "like the thing that we're doing right"
  },
  {
    "startMs": 222920,
    "text": "now."
  },
  {
    "startMs": 223959,
    "text": "So, I give you my sounds and words."
  },
  {
    "startMs": 227800,
    "text": "You hear them."
  },
  {
    "startMs": 229360,
    "text": "And then, if it is a conversation,"
  },
  {
    "startMs": 231800,
    "text": "you're going to give me your sounds and"
  },
  {
    "startMs": 233400,
    "text": "your words."
  },
  {
    "startMs": 235680,
    "text": "And then this is going back and forth"
  },
  {
    "startMs": 238200,
    "text": "through interaction."
  },
  {
    "startMs": 241320,
    "text": "And then in this process, we're"
  },
  {
    "startMs": 243000,
    "text": "continuously"
  },
  {
    "startMs": 244880,
    "text": "processing and updating our mental"
  },
  {
    "startMs": 247080,
    "text": "models."
  },
  {
    "startMs": 249280,
    "text": "So that's a joint activity"
  },
  {
    "startMs": 252720,
    "text": "for human communication."
  },
  {
    "startMs": 254800,
    "text": "And I would like to say"
  },
  {
    "startMs": 257440,
    "text": "in the voice AI human communication,"
  },
  {
    "startMs": 260600,
    "text": "it also has to be a joint activity like"
  },
  {
    "startMs": 263040,
    "text": "this."
  },
  {
    "startMs": 264480,
    "text": "Because that's the only thing that we"
  },
  {
    "startMs": 267200,
    "text": "know about human communication as a"
  },
  {
    "startMs": 269440,
    "text": "human being. We have been evolving"
  },
  {
    "startMs": 271760,
    "text": "thousands of years as communicators, and"
  },
  {
    "startMs": 274600,
    "text": "this is what we know. So we expect the"
  },
  {
    "startMs": 276720,
    "text": "same thing to bots."
  },
  {
    "startMs": 281200,
    "text": "So let me go over the failure scene of"
  },
  {
    "startMs": 284560,
    "text": "my call with the voice agent"
  },
  {
    "startMs": 287520,
    "text": "in this framework."
  },
  {
    "startMs": 289240,
    "text": "So you see there's listen"
  },
  {
    "startMs": 292120,
    "text": "and speak for each party."
  },
  {
    "startMs": 297240,
    "text": "So I start spelling my first name."
  },
  {
    "startMs": 300240,
    "text": "And then the bot did not hear that the"
  },
  {
    "startMs": 304200,
    "text": "difference between M and N correctly, so"
  },
  {
    "startMs": 307040,
    "text": "it's an"
  },
  {
    "startMs": 308480,
    "text": "SCT failure in the listening level."
  },
  {
    "startMs": 312840,
    "text": "And then the TTS applies only"
  },
  {
    "startMs": 315400,
    "text": "English-centric reading rules to my"
  },
  {
    "startMs": 317760,
    "text": "name, M I D A M, would read it as Midam"
  },
  {
    "startMs": 321200,
    "text": "in the"
  },
  {
    "startMs": 322280,
    "text": "American English version."
  },
  {
    "startMs": 325480,
    "text": "So I'm confused, but at this time I'm"
  },
  {
    "startMs": 327919,
    "text": "kind of generous because that happens a"
  },
  {
    "startMs": 329720,
    "text": "lot even with human beings. So I'm okay."
  },
  {
    "startMs": 334160,
    "text": "But then when it brought"
  },
  {
    "startMs": 335880,
    "text": "brought up account number thing"
  },
  {
    "startMs": 338160,
    "text": "because I don't know what that is,"
  },
  {
    "startMs": 340880,
    "text": "I'm confused again."
  },
  {
    "startMs": 342960,
    "text": "But I'm adaptive, I can find I can look"
  },
  {
    "startMs": 345120,
    "text": "for it."
  },
  {
    "startMs": 346280,
    "text": "So I found the number, start reading it,"
  },
  {
    "startMs": 348880,
    "text": "but"
  },
  {
    "startMs": 350120,
    "text": "the STT did not recognize the word unit"
  },
  {
    "startMs": 353040,
    "text": "correctly, so"
  },
  {
    "startMs": 354920,
    "text": "it cuts me off, and uh"
  },
  {
    "startMs": 358960,
    "text": "uh finally, it's uh eventually talked"
  },
  {
    "startMs": 362040,
    "text": "over me."
  },
  {
    "startMs": 363400,
    "text": "So, I get"
  },
  {
    "startMs": 365680,
    "text": "really irritated."
  },
  {
    "startMs": 368240,
    "text": "And then, when it asked me for the"
  },
  {
    "startMs": 369720,
    "text": "repetition of the same information, and"
  },
  {
    "startMs": 373040,
    "text": "then, it is clear that the spot is not"
  },
  {
    "startMs": 375440,
    "text": "tracking the mental model with me."
  },
  {
    "startMs": 378400,
    "text": "And then, very rudely, it's uh does not"
  },
  {
    "startMs": 382000,
    "text": "even try interactive clarification,"
  },
  {
    "startMs": 384400,
    "text": "which is a common strategy by human"
  },
  {
    "startMs": 386680,
    "text": "beings."
  },
  {
    "startMs": 387919,
    "text": "So, I don't want to deal with this"
  },
  {
    "startMs": 389000,
    "text": "anymore, so I say, \"Can I talk to a"
  },
  {
    "startMs": 390840,
    "text": "person?\""
  },
  {
    "startMs": 394080,
    "text": "So,"
  },
  {
    "startMs": 395640,
    "text": "let's go over the uh the framework"
  },
  {
    "startMs": 397600,
    "text": "again. So, the these are the linguistic"
  },
  {
    "startMs": 399480,
    "text": "components that are expected and well"
  },
  {
    "startMs": 401680,
    "text": "maintained in human-to-human voi- uh"
  },
  {
    "startMs": 405520,
    "text": "uh conversation."
  },
  {
    "startMs": 407520,
    "text": "So, there are listening channels, a"
  },
  {
    "startMs": 409320,
    "text": "listening channel and speaking channel,"
  },
  {
    "startMs": 410840,
    "text": "and there are different components like"
  },
  {
    "startMs": 412240,
    "text": "sounds, words, interaction, and mental"
  },
  {
    "startMs": 414440,
    "text": "model."
  },
  {
    "startMs": 415680,
    "text": "So, the first component is, does the bot"
  },
  {
    "startMs": 418360,
    "text": "recognize the user's speech well?"
  },
  {
    "startMs": 421520,
    "text": "And all of these technical terms"
  },
  {
    "startMs": 424120,
    "text": "uh will fall under this."
  },
  {
    "startMs": 427120,
    "text": "And then, there was there's going to be"
  },
  {
    "startMs": 428760,
    "text": "this second component, which is words in"
  },
  {
    "startMs": 431640,
    "text": "the listening channel. So, does the bot"
  },
  {
    "startMs": 433760,
    "text": "understand the user's words?"
  },
  {
    "startMs": 437400,
    "text": "And then, the third one is, does the bot"
  },
  {
    "startMs": 439560,
    "text": "wait until the right timing to for its"
  },
  {
    "startMs": 442240,
    "text": "turn? It's about It's going to be about"
  },
  {
    "startMs": 444520,
    "text": "uh listening channel interaction."
  },
  {
    "startMs": 448640,
    "text": "And then, uh the last part is mental"
  },
  {
    "startMs": 451240,
    "text": "model. So, does the bot understand the"
  },
  {
    "startMs": 453600,
    "text": "user's intention"
  },
  {
    "startMs": 455520,
    "text": "in the listening part?"
  },
  {
    "startMs": 458280,
    "text": "And then, we can also go to the speaking"
  },
  {
    "startMs": 459840,
    "text": "channel, so it's going to be about"
  },
  {
    "startMs": 461360,
    "text": "pronunciation for the sound."
  },
  {
    "startMs": 463880,
    "text": "And also there's about understand the"
  },
  {
    "startMs": 466720,
    "text": "the words users are"
  },
  {
    "startMs": 469280,
    "text": "uh there's about choose the words the"
  },
  {
    "startMs": 471120,
    "text": "user can understand."
  },
  {
    "startMs": 473640,
    "text": "And in the interaction part, there's"
  },
  {
    "startMs": 475480,
    "text": "about speak with the right timing."
  },
  {
    "startMs": 478800,
    "text": "And lastly, there's about speak with the"
  },
  {
    "startMs": 481240,
    "text": "information the user actually need."
  },
  {
    "startMs": 485640,
    "text": "So, there are a lot of engineering or"
  },
  {
    "startMs": 488520,
    "text": "linguistic or cognitive science terms"
  },
  {
    "startMs": 490760,
    "text": "that are in here that that are here. Um"
  },
  {
    "startMs": 494640,
    "text": "you can see now see that all of those"
  },
  {
    "startMs": 497040,
    "text": "have their right spots in this"
  },
  {
    "startMs": 498880,
    "text": "linguistic framework."
  },
  {
    "startMs": 503200,
    "text": "And importantly, these components are"
  },
  {
    "startMs": 504880,
    "text": "interdependent,"
  },
  {
    "startMs": 506440,
    "text": "not separate or uh independent from each"
  },
  {
    "startMs": 509880,
    "text": "other. They're interdependent and"
  },
  {
    "startMs": 511760,
    "text": "they're aligned. So, when you want to do"
  },
  {
    "startMs": 514360,
    "text": "good things about sounds,"
  },
  {
    "startMs": 517039,
    "text": "you have to think about words level."
  },
  {
    "startMs": 519719,
    "text": "And then when you want to do good things"
  },
  {
    "startMs": 521240,
    "text": "about these sounds and words,"
  },
  {
    "startMs": 523599,
    "text": "you also have to uh account for"
  },
  {
    "startMs": 526200,
    "text": "interaction, so turn taking or turn"
  },
  {
    "startMs": 528680,
    "text": "detection."
  },
  {
    "startMs": 530320,
    "text": "And then finally, you want to uh have"
  },
  {
    "startMs": 533480,
    "text": "good uh task completion, which is the"
  },
  {
    "startMs": 536280,
    "text": "goal of these mental model uh layer."
  },
  {
    "startMs": 539520,
    "text": "Then you have to have all of these."
  },
  {
    "startMs": 542520,
    "text": "Without all of those, without any of"
  },
  {
    "startMs": 544000,
    "text": "those, any of those components, your"
  },
  {
    "startMs": 546800,
    "text": "voice agent will fail."
  },
  {
    "startMs": 549880,
    "text": "And then finally,"
  },
  {
    "startMs": 551760,
    "text": "uh it has to be well aligned. All of"
  },
  {
    "startMs": 553600,
    "text": "these have to be well aligned."
  },
  {
    "startMs": 557080,
    "text": "And additionally, you have to keep your"
  },
  {
    "startMs": 560120,
    "text": "mind keep in mind that"
  },
  {
    "startMs": 562480,
    "text": "this is happening on the timeline."
  },
  {
    "startMs": 566240,
    "text": "What I mean by that is it is silently"
  },
  {
    "startMs": 569960,
    "text": "tracked. Unlike in chat, in chat you see"
  },
  {
    "startMs": 573120,
    "text": "the history of what was said"
  },
  {
    "startMs": 575560,
    "text": "uh as text."
  },
  {
    "startMs": 577200,
    "text": "But in voice agent experience,"
  },
  {
    "startMs": 580320,
    "text": "uh, you say something, and the bot says"
  },
  {
    "startMs": 582680,
    "text": "something, you go back and forth,"
  },
  {
    "startMs": 585840,
    "text": "and then see, all these waveforms, the"
  },
  {
    "startMs": 589160,
    "text": "air via the vibration in the, uh, in the"
  },
  {
    "startMs": 591400,
    "text": "air, they're all gone."
  },
  {
    "startMs": 593760,
    "text": "And only the user's mental model is the"
  },
  {
    "startMs": 596720,
    "text": "thing that's left, and that matters."
  },
  {
    "startMs": 600480,
    "text": "So, sounds, words, interactions vanish"
  },
  {
    "startMs": 603000,
    "text": "the moment they're spoken,"
  },
  {
    "startMs": 604960,
    "text": "but the mental model proceeds and grows"
  },
  {
    "startMs": 607080,
    "text": "over the timeline."
  },
  {
    "startMs": 609120,
    "text": "So, this is what you have to"
  },
  {
    "startMs": 612680,
    "text": "target"
  },
  {
    "startMs": 614160,
    "text": "for user satisfaction."
  },
  {
    "startMs": 616560,
    "text": "And then, what can we do"
  },
  {
    "startMs": 619080,
    "text": "for the bot to meet the standard of the"
  },
  {
    "startMs": 622440,
    "text": "user?"
  },
  {
    "startMs": 625160,
    "text": "So, what we can do, uh, would include,"
  },
  {
    "startMs": 628160,
    "text": "of course, choosing good ASR models or"
  },
  {
    "startMs": 631360,
    "text": "configurations and do some"
  },
  {
    "startMs": 632880,
    "text": "post-processing,"
  },
  {
    "startMs": 634480,
    "text": "uh, choosing good TTS models,"
  },
  {
    "startMs": 636080,
    "text": "configurations, and pre-processing,"
  },
  {
    "startMs": 638840,
    "text": "and, uh, carefully curate the vocabulary"
  },
  {
    "startMs": 642720,
    "text": "that can be shared between the bot and"
  },
  {
    "startMs": 644280,
    "text": "the user,"
  },
  {
    "startMs": 645920,
    "text": "and do good job of a turn-to-turn"
  },
  {
    "startMs": 648640,
    "text": "detection, latency, and turn-taking."
  },
  {
    "startMs": 652800,
    "text": "Um, and very importantly, we have to, it"
  },
  {
    "startMs": 656040,
    "text": "would be great if we can do good emotion"
  },
  {
    "startMs": 657640,
    "text": "detection and handling, and context"
  },
  {
    "startMs": 660560,
    "text": "retention, and by context, what I mean"
  },
  {
    "startMs": 662320,
    "text": "is context about all of these."
  },
  {
    "startMs": 667880,
    "text": "And importantly,"
  },
  {
    "startMs": 669960,
    "text": "uh, it has to be dynamic because things"
  },
  {
    "startMs": 672400,
    "text": "are always changing, uh, throughout over"
  },
  {
    "startMs": 674640,
    "text": "the course of the call. So, we would"
  },
  {
    "startMs": 677280,
    "text": "have to do this management dynamically"
  },
  {
    "startMs": 679839,
    "text": "along the timeline"
  },
  {
    "startMs": 681680,
    "text": "for different kinds of people."
  },
  {
    "startMs": 683960,
    "text": "So, kids or different kinds of people"
  },
  {
    "startMs": 686880,
    "text": "like these will have different"
  },
  {
    "startMs": 688440,
    "text": "expectations that we have to satisfy."
  },
  {
    "startMs": 692000,
    "text": "Uh, not just when they're happy, but"
  },
  {
    "startMs": 694320,
    "text": "also when they're not happy."
  },
  {
    "startMs": 696680,
    "text": "So, only then you can pursue a dynamic"
  },
  {
    "startMs": 699920,
    "text": "and truly scalable orchestration of"
  },
  {
    "startMs": 701800,
    "text": "voice AI."
  },
  {
    "startMs": 703120,
    "text": "So, it's a very difficult job to do."
  },
  {
    "startMs": 708240,
    "text": "We always say that voice is the most"
  },
  {
    "startMs": 710200,
    "text": "natural way of communication, but it is"
  },
  {
    "startMs": 713000,
    "text": "not actually not easy. Behind the scene,"
  },
  {
    "startMs": 715680,
    "text": "it is thanks to this linguistic"
  },
  {
    "startMs": 717200,
    "text": "orchestration."
  },
  {
    "startMs": 719200,
    "text": "When your bot is not good at it,"
  },
  {
    "startMs": 721960,
    "text": "it's a catastrophic failure."
  },
  {
    "startMs": 726960,
    "text": "Um, so paying attention to this"
  },
  {
    "startMs": 729760,
    "text": "linguistic framework would have lots of"
  },
  {
    "startMs": 732040,
    "text": "business implications because then you"
  },
  {
    "startMs": 734840,
    "text": "can uh"
  },
  {
    "startMs": 737120,
    "text": "decrease all of these user frustration,"
  },
  {
    "startMs": 739280,
    "text": "task failures, live agent escalation, or"
  },
  {
    "startMs": 742280,
    "text": "abandoned calls, or silent failures."
  },
  {
    "startMs": 748000,
    "text": "So, in ServiceNow, we have made a a good"
  },
  {
    "startMs": 752440,
    "text": "uh benchmark end-to-end benchmark called"
  },
  {
    "startMs": 754520,
    "text": "Eva bench. So, you can try that to"
  },
  {
    "startMs": 756000,
    "text": "diagnose your voice agent's uh status."
  },
  {
    "startMs": 763400,
    "text": "Um, key takeaways."
  },
  {
    "startMs": 765960,
    "text": "So, voice AI is a joint activity between"
  },
  {
    "startMs": 770200,
    "text": "the bot and the user, not just a"
  },
  {
    "startMs": 772080,
    "text": "pipeline."
  },
  {
    "startMs": 774240,
    "text": "And we must serve users' needs in"
  },
  {
    "startMs": 775880,
    "text": "multiple layers real time."
  },
  {
    "startMs": 779760,
    "text": "It's not that I have given you a fix"
  },
  {
    "startMs": 781480,
    "text": "today because there's nothing like that."
  },
  {
    "startMs": 784320,
    "text": "It just uh the fix is in you and your"
  },
  {
    "startMs": 787560,
    "text": "system."
  },
  {
    "startMs": 789920,
    "text": "But, what I have given you is today is"
  },
  {
    "startMs": 794040,
    "text": "the linguistic framework you can try to"
  },
  {
    "startMs": 796360,
    "text": "diagnose your system"
  },
  {
    "startMs": 798560,
    "text": "and to build your system upon."
  },
  {
    "startMs": 801560,
    "text": "You can try Eva, but also you can learn"
  },
  {
    "startMs": 804000,
    "text": "linguistics and hire linguists."
  },
  {
    "startMs": 807680,
    "text": "Um, another thing I want to remind you"
  },
  {
    "startMs": 809880,
    "text": "of is that business implications are"
  },
  {
    "startMs": 812920,
    "text": "linguistic implications and vice versa"
  },
  {
    "startMs": 815920,
    "text": "in this voice AI scene. Because voice is"
  },
  {
    "startMs": 819480,
    "text": "fundamentally a linguistic and very"
  },
  {
    "startMs": 821600,
    "text": "human and cognitive experience."
  },
  {
    "startMs": 826000,
    "text": "I would like to ask you a longer term"
  },
  {
    "startMs": 828200,
    "text": "question."
  },
  {
    "startMs": 830720,
    "text": "Speakers adapt. So, I"
  },
  {
    "startMs": 834400,
    "text": "I'm pretty sure that in this talk in my"
  },
  {
    "startMs": 838000,
    "text": "talk with you guys today, you have"
  },
  {
    "startMs": 840440,
    "text": "learned something about me, about my"
  },
  {
    "startMs": 842480,
    "text": "speaking style, what kind of accents I"
  },
  {
    "startMs": 844800,
    "text": "speak, what kind of words I'm using. So,"
  },
  {
    "startMs": 847640,
    "text": "next time I see you guys in person, you"
  },
  {
    "startMs": 850400,
    "text": "would find it more comfortable to talk"
  },
  {
    "startMs": 852280,
    "text": "to me because you have paid attention to"
  },
  {
    "startMs": 854320,
    "text": "me."
  },
  {
    "startMs": 855360,
    "text": "Right? So, speakers are always adapting."
  },
  {
    "startMs": 857600,
    "text": "So, the user will be adapting to your"
  },
  {
    "startMs": 860560,
    "text": "voice agent throughout the call. So, is"
  },
  {
    "startMs": 864440,
    "text": "your system ready for them to"
  },
  {
    "startMs": 867920,
    "text": "use you better, use it your voice agent"
  },
  {
    "startMs": 869760,
    "text": "better the next time?"
  },
  {
    "startMs": 871480,
    "text": "And"
  },
  {
    "startMs": 872680,
    "text": "language is always change. So, is your"
  },
  {
    "startMs": 875320,
    "text": "voice agent ready for language change in"
  },
  {
    "startMs": 878280,
    "text": "1 year or 6 months even?"
  },
  {
    "startMs": 884360,
    "text": "So, thank you."
  },
  {
    "startMs": 887535,
    "text": ">> [applause]"
  }
]