id large_stringlengths 15 15 | talk_id large_stringclasses 21
values | audio large_stringclasses 21
values | pdf large_stringclasses 21
values | offset float64 0.03 424 | duration float64 0.2 32 | speaker_id large_stringclasses 21
values | source_text_en large_stringlengths 3 421 | target_text_de large_stringlengths 3 488 | target_text_it large_stringlengths 3 491 | target_text_zh large_stringlengths 2 170 |
|---|---|---|---|---|---|---|---|---|---|---|
xhjBHGVOyQ_0038 | xhjBHGVOyQ | MCIF_DATA/LONG_AUDIO/xhjBHGVOyQ.wav | MCIF_DATA/PDF/xhjBHGVOyQ.pdf | 290.969 | 9.657 | 5 | As we can see by transfer learning from natural instruction datasets, the model can achieve much better sensitivity compared to the original OFA model. | Wie wir sehen können, kann das Modell durch Transferlernen aus natürlichen Anweisungsdatensätzen eine viel bessere Empfindlichkeit im Vergleich zum ursprünglichen OFA-Modell erreichen. | Come possiamo vedere dall'apprendimento per trasferimento da set di dati di istruzioni naturali, il modello può raggiungere una sensibilità molto più elevata rispetto al modello OFA originale. | 正如我们通过自然指令数据集的迁移学习所看到的,与原始 OFA 模型相比,该模型可以实现更优的灵敏度。 |
xhjBHGVOyQ_0039 | xhjBHGVOyQ | MCIF_DATA/LONG_AUDIO/xhjBHGVOyQ.wav | MCIF_DATA/PDF/xhjBHGVOyQ.pdf | 303.826 | 9.037 | 5 | We also can see transfer learning from natural instruction datasets can help OFA to attain much better performance on the natural instruct dataset. | Wir können auch sehen, dass Transfer-Lernen aus natürlichen Anweisungsdatensätzen OFA dabei helfen kann, eine viel bessere Leistung auf dem natürlichen Anweisungsdatensatz zu erreichen. | Possiamo anche vedere che l'apprendimento per trasferimento da set di dati di istruzioni naturali può aiutare OFA a ottenere prestazioni molto migliori sul set di dati di istruzioni naturali. | 我们还可以看到,来自自然指令数据集的迁移学习可以帮助 OFA 在自然指令数据集上实现更好的表现。 |
xhjBHGVOyQ_0040 | xhjBHGVOyQ | MCIF_DATA/LONG_AUDIO/xhjBHGVOyQ.wav | MCIF_DATA/PDF/xhjBHGVOyQ.pdf | 315.288 | 12.157 | 5 | So overall, we propose the first large scale multi-model instruction tuning dataset with significantly improved their short capability of OFA, and we explore different transfer learning technique and show their benefits. | Zusammenfassend stellen wir den ersten großen Multi-Modell-Datensatz für die Anweisungsabstimmung vor, der die kurzen Fähigkeiten von OFA signifikant verbessert, und wir untersuchen verschiedene Transfer-Learning-Techniken und zeigen deren Vorteile auf. | Quindi, nel complesso, proponiamo il primo set di dati di messa a punto delle istruzioni multi-modello su larga scala con una capacità di OFA notevolmente migliorata ed esploriamo diverse tecniche di apprendimento per trasferimento mostrandone i vantaggi. | 因此,总体而言,我们提出了第一个大规模多模型指令调整数据集,显著提高了 OFA 的短期能力,并且我们探索了不同的迁移学习技术并展示了它们的优点。 |
xhjBHGVOyQ_0041 | xhjBHGVOyQ | MCIF_DATA/LONG_AUDIO/xhjBHGVOyQ.wav | MCIF_DATA/PDF/xhjBHGVOyQ.pdf | 328.253 | 1.99 | 5 | We design a new metric called sensitivity. | Wir entwickeln eine neue Metrik, die Sensitivität. | Progettiamo una nuova metrica chiamata sensibilità. | 我们设计了一个新的指标,称为“灵敏度”。 |
xhjBHGVOyQ_0042 | xhjBHGVOyQ | MCIF_DATA/LONG_AUDIO/xhjBHGVOyQ.wav | MCIF_DATA/PDF/xhjBHGVOyQ.pdf | 331.495 | 9.731 | 5 | So one more thing, we are collecting a much larger multi-model instruction tuning dataset with around 150 additional vision language tasks and we will release them. | Und noch etwas: Wir sind gerade dabei, einen viel größeren multimodalen Anweisungsdatensatz mit etwa 150 zusätzlichen Bildverarbeitungsaufgaben zu erstellen und anschließend zu veröffentlichen. | Un'altra cosa: stiamo raccogliendo un set di dati di messa a punto delle istruzioni multi-modello molto più ampio con circa 150 attività di linguaggio visivo aggiuntive e lo renderemo disponibile. | 此外,我们正在收集一个更大的多模型指令调整数据集,其中包含大约 150 个额外的视觉语言任务,我们将发布这些数据集。 |
xhjBHGVOyQ_0043 | xhjBHGVOyQ | MCIF_DATA/LONG_AUDIO/xhjBHGVOyQ.wav | MCIF_DATA/PDF/xhjBHGVOyQ.pdf | 341.213 | 4.32 | 5 | So this is a QR code for our data and model. | Das ist der QR-Code für unsere Daten und unser Modell. | Questo è un codice QR per i nostri dati e il nostro modello. | 这是我们的数据和模型的二维码。 |
xhjBHGVOyQ_0044 | xhjBHGVOyQ | MCIF_DATA/LONG_AUDIO/xhjBHGVOyQ.wav | MCIF_DATA/PDF/xhjBHGVOyQ.pdf | 346.007 | 0.491 | 5 | Thank you. | Vielen Dank. | Grazie." | 谢谢。 |
xiSxNRoOzm_0000 | xiSxNRoOzm | MCIF_DATA/LONG_AUDIO/xiSxNRoOzm.wav | MCIF_DATA/PDF/xiSxNRoOzm.pdf | 1.432 | 5.425 | 6 | "Hello, my name is Vasudha and I'm a Computer Science PhD candidate at Stony Brook University. | Hallo, mein Name ist Vasudha und ich bin Doktorand in Informatik an der Stony Brook Universität. | "Salve, mi chiamo Vasudha e faccio il dottorato in Informatica presso la Stony Brook University. | 大家好,我叫 Vasudha,是斯托尼布鲁克大学的计算机科学博士生。 |
xiSxNRoOzm_0001 | xiSxNRoOzm | MCIF_DATA/LONG_AUDIO/xiSxNRoOzm.wav | MCIF_DATA/PDF/xiSxNRoOzm.pdf | 7.742 | 8.936 | 6 | I would like to present our work accepted into ACL 2023 as a long paper, ""Transfer Learning for Dissonance Detection: Addressing the Rare-Class Challenge."" | Ich möchte Ihnen unsere Arbeit vorstellen, die als Long Paper bei der ACL 2023 angenommen wurde: „Transfer Learning for Dissonance Detection: Addressing the Rare-Class Challenge“. | Vorrei presentare il nostro lavoro accettato in ACL 2023 come un lungo articolo, "Transfer Learning for Dissonance Detection: Addressing the Rare-Class Challenge". | 我想介绍一下我们的论文:《用于检测失调的迁移学习:应对罕见类别挑战》,这篇长篇论文已被 ACL 2023 录用。 |
xiSxNRoOzm_0002 | xiSxNRoOzm | MCIF_DATA/LONG_AUDIO/xiSxNRoOzm.wav | MCIF_DATA/PDF/xiSxNRoOzm.pdf | 20.311 | 5.719 | 6 | We begin by defining cognitive dissonance and why it is an important problem to study in language. | Wir beginnen mit einer Definition von kognitiver Dissonanz und erklären, warum sie ein wichtiges Problem bei der Untersuchung von Sprache darstellt. | Iniziamo definendo la dissonanza cognitiva e perché rappresenta un problema importante da studiare nel linguaggio. | 首先,我们来定义一下什么是认知失调,以及为什么它是语言学习中的一个重要问题。 |
xiSxNRoOzm_0003 | xiSxNRoOzm | MCIF_DATA/LONG_AUDIO/xiSxNRoOzm.wav | MCIF_DATA/PDF/xiSxNRoOzm.pdf | 26.514 | 15.864 | 6 | Simply put, cognitive dissonance is two beliefs or actions that are inconsistent, such as this example where a person states, ""I know that cigarettes could kill me"", and then goes on to say ""I grabbed a couple of smokes after the meeting"". | Einfach ausgedrückt handelt es sich bei kognitiver Dissonanz um zwei Überzeugungen oder Handlungen, die nicht miteinander vereinbar sind, wie in diesem Beispiel, in dem eine Person sagt: „Ich weiß, dass Zigaretten mich umbringen können“ und dann sagt: „Ich habe nach dem Meeting ein paar Zigaretten geraucht“. | In poche parole, la dissonanza cognitiva consiste in due convinzioni o azioni incoerenti, come questo esempio in cui una persona afferma: "So che le sigarette potrebbero uccidermi", e poi continua dicendo "Ho fumato un paio di sigarette dopo la riunione". | 简而言之,认知失调是指两种信念或行为之间不一致,比如这个例子,一个人说:“我知道香烟会害死我”,然后又说:“我在会后抽了几根烟”。 |
xiSxNRoOzm_0004 | xiSxNRoOzm | MCIF_DATA/LONG_AUDIO/xiSxNRoOzm.wav | MCIF_DATA/PDF/xiSxNRoOzm.pdf | 42.382 | 4.378 | 6 | This belief and action are inconsistent, and they are in dissonance. | Diese Überzeugung und diese Handlung sind inkonsistent und stehen im Widerspruch zueinander. | Questa convinzione e questa azione sono incoerenti e sono in dissonanza. | 这种信念和行为是不一致的,它们处于失调状态。 |
xiSxNRoOzm_0005 | xiSxNRoOzm | MCIF_DATA/LONG_AUDIO/xiSxNRoOzm.wav | MCIF_DATA/PDF/xiSxNRoOzm.pdf | 48.273 | 5.621 | 6 | Further mentioning that ""I don't think I could keep my job without them"" justifies the second occurrence. | Die weitere Aussage „Ich glaube nicht, dass ich ohne sie meinen Job behalten könnte“ rechtfertigt das zweite Ereignis. | L'ulteriore affermazione che "Non credo di poter mantenere il mio lavoro senza di loro" giustifica la seconda occorrenza. | 此人还提到“没有烟草,我认为我就无法保住工作”,这就是第二种情况。 |
xiSxNRoOzm_0006 | xiSxNRoOzm | MCIF_DATA/LONG_AUDIO/xiSxNRoOzm.wav | MCIF_DATA/PDF/xiSxNRoOzm.pdf | 54.584 | 2.01 | 6 | And they have a consonance relationship. | Und sie stehen in einer konsonanten Beziehung zueinander. | Queste hanno una relazione di consonanza. | 这两者之间存在共鸣关系。 |
xiSxNRoOzm_0007 | xiSxNRoOzm | MCIF_DATA/LONG_AUDIO/xiSxNRoOzm.wav | MCIF_DATA/PDF/xiSxNRoOzm.pdf | 57.767 | 8.629 | 6 | While dissonance is a very common phenomenon we experienced in daily decision making, they are really rare to find expressed in language among other kinds of discourse relations. | Während Dissonanz ein sehr häufiges Phänomen ist, das wir in der alltäglichen Entscheidungsfindung erleben, ist es sehr selten, dass sie in der Sprache unter anderen Arten von Diskursbeziehungen zum Ausdruck kommt. | Sebbene la dissonanza sia un fenomeno molto comune che sperimentiamo nel processo decisionale quotidiano, è davvero raro trovarla espressa nel linguaggio tra gli altri tipi di relazioni del discorso. | 虽然失调是我们在日常决策中遇到的一种非常常见的现象,但在其他类型的话语关系中,很少能在语言中找到它们的表达。 |
xiSxNRoOzm_0008 | xiSxNRoOzm | MCIF_DATA/LONG_AUDIO/xiSxNRoOzm.wav | MCIF_DATA/PDF/xiSxNRoOzm.pdf | 67.566 | 1.087 | 6 | So why does this matter? | Warum ist das nun wichtig? | Ma perché tutta questa importanza? | 不过,这一点为什么很重要呢? |
xiSxNRoOzm_0009 | xiSxNRoOzm | MCIF_DATA/LONG_AUDIO/xiSxNRoOzm.wav | MCIF_DATA/PDF/xiSxNRoOzm.pdf | 69.426 | 8.207 | 6 | Studying cognitive dissonance can help us understand the effects of disagreement among people, track trends and belief values, and attitude changes in population. | Die Erforschung der kognitiven Dissonanz kann uns helfen, die Auswirkungen von Meinungsverschiedenheiten zwischen Menschen zu verstehen, Trends und Überzeugungen zu verfolgen und Veränderungen in den Einstellungen von Bevölkerungsgruppen zu beobachten. | Studiare la dissonanza cognitiva può aiutarci a comprendere gli effetti del disaccordo tra le persone, a monitorare le tendenze e i valori delle credenze e i cambiamenti di atteggiamento nella popolazione. | 研究认知失调可以帮助我们理解人与人之间分歧的影响,跟踪趋势和信仰价值观,以及人群态度的变化。 |
xiSxNRoOzm_0010 | xiSxNRoOzm | MCIF_DATA/LONG_AUDIO/xiSxNRoOzm.wav | MCIF_DATA/PDF/xiSxNRoOzm.pdf | 77.618 | 6.764 | 6 | High cognitive dissonance is also related to anxiety disorders and can help understand people's mental health better. | Eine hohe kognitive Dissonanz ist auch mit Angst verbunden und kann uns helfen, die psychische Gesundheit der Menschen zu verstehen. | L'elevata dissonanza cognitiva è anche correlata ai disturbi d'ansia e può aiutare a comprendere meglio la salute mentale delle persone. | 高度的认知失调也与焦虑症有关,可以帮助我们更好地了解人们的心理健康。 |
xiSxNRoOzm_0011 | xiSxNRoOzm | MCIF_DATA/LONG_AUDIO/xiSxNRoOzm.wav | MCIF_DATA/PDF/xiSxNRoOzm.pdf | 84.409 | 8.731 | 6 | Studying dissonance expressed in language can also be beneficial in understanding extremism and polarization of vulnerable groups. | Die Untersuchung der in der Sprache ausgedrückten Dissonanz kann auch nützlich sein, um Extremismus und Polarisierung unter gefährdeten Gruppen zu verstehen. | Lo studio della dissonanza espressa nel linguaggio può anche essere utile per comprendere l'estremismo e la polarizzazione dei gruppi vulnerabili. | 研究语言中表达的失调也有助于理解弱势群体的极端主义和两极分化。 |
xiSxNRoOzm_0012 | xiSxNRoOzm | MCIF_DATA/LONG_AUDIO/xiSxNRoOzm.wav | MCIF_DATA/PDF/xiSxNRoOzm.pdf | 94.386 | 8.545 | 6 | Finally, cognitive dissonance is important to understand personal cognitive styles of individuals and helps us understand decision making processes better. | Schließlich ist kognitive Dissonanz wichtig für das Verständnis des persönlichen kognitiven Stils von Menschen und hilft uns, Entscheidungsprozesse besser zu verstehen. | Infine, la dissonanza cognitiva è importante per comprendere gli stili cognitivi personali e ci aiuta a capire meglio i processi decisionali. | 最后,认知失调对于理解个人的认知风格非常重要,它可以帮助我们更好地理解决策过程。 |
xiSxNRoOzm_0013 | xiSxNRoOzm | MCIF_DATA/LONG_AUDIO/xiSxNRoOzm.wav | MCIF_DATA/PDF/xiSxNRoOzm.pdf | 102.96 | 8.706 | 6 | To the goal of creating a cognitive dissonance resource, we conducted a large scale annotation of dissonance relations. | Um eine Ressource für kognitive Dissonanz zu schaffen, haben wir eine groß angelegte Annotation von Dissonanzbeziehungen durchgeführt. | Con l'obiettivo di creare una risorsa relativa alla dissonanza cognitiva, abbiamo condotto un'annotazione su larga scala delle relazioni di dissonanza. | 为了创建认知失调资源,我们对失调关系进行了大规模的注释。 |
xiSxNRoOzm_0014 | xiSxNRoOzm | MCIF_DATA/LONG_AUDIO/xiSxNRoOzm.wav | MCIF_DATA/PDF/xiSxNRoOzm.pdf | 112.295 | 3.857 | 6 | We used dissonance-first approach, as seen in the flow chart here. | Dabei haben wir einen Ansatz gewählt, der von der Dissonanz ausgeht, wie im folgenden Flussdiagramm dargestellt. | Abbiamo utilizzato un approccio basato sulla dissonanza, come è possibile vedere in questo diagramma di flusso. | 我们使用了“失调为先”的方法,如此处的流程图所示。 |
xiSxNRoOzm_0015 | xiSxNRoOzm | MCIF_DATA/LONG_AUDIO/xiSxNRoOzm.wav | MCIF_DATA/PDF/xiSxNRoOzm.pdf | 116.158 | 8.89 | 6 | Tweets were passed using the PDTB parser, and pairs of discourse units were annotated according to the guidelines that are described in our paper. | Die Tweets wurden durch den PDTB-Parser geleitet und die Diskurseinheitenpaare wurden gemäß den in unserer Arbeit beschriebenen Richtlinien annotiert. | I tweet sono stati vagliati utilizzando il parser PDTB e le coppie di unità discorsive sono state annotate secondo le linee guida descritte nel nostro articolo. | 推文使用 PDTB 解析器传递,并根据我们论文中描述的指南对篇章单元对进行了注释。 |
xiSxNRoOzm_0016 | xiSxNRoOzm | MCIF_DATA/LONG_AUDIO/xiSxNRoOzm.wav | MCIF_DATA/PDF/xiSxNRoOzm.pdf | 125.734 | 5.299 | 6 | As can be seen here, dissonance was only found in 3.5% of the annotated pairs. | Wie Sie hier sehen können, wurde Dissonanz in nur 3,5 % der annotierten Paare gefunden. | Come si può vedere qui, la dissonanza è stata trovata solo nel 3,5% delle coppie annotate. | 从这里可以看出,只有 3.5% 的注释对存在失调情况。 |
xiSxNRoOzm_0017 | xiSxNRoOzm | MCIF_DATA/LONG_AUDIO/xiSxNRoOzm.wav | MCIF_DATA/PDF/xiSxNRoOzm.pdf | 134.613 | 7.965 | 6 | On collecting around 1,000 examples of discourse unit pairs, we ran training for an initial classifier trained only on 43 examples of dissonance. | Nachdem wir ca. 1.000 Beispiele von Diskurseinheitenpaaren gesammelt hatten, trainierten wir einen ersten Klassifikator, der mit nur 43 Dissonanzbeispielen trainiert wurde. | Raccogliendo circa 1.000 esempi di coppie di unità discorsive, abbiamo eseguito l'addestramento per un classificatore iniziale addestrato solo su 43 esempi di dissonanza. | 在收集了大约 1,000 个篇章单元对的例子后,我们对一个初始分类器进行了训练,该分类器仅对 43 个失调的例子进行了训练。 |
xiSxNRoOzm_0018 | xiSxNRoOzm | MCIF_DATA/LONG_AUDIO/xiSxNRoOzm.wav | MCIF_DATA/PDF/xiSxNRoOzm.pdf | 142.946 | 3.336 | 6 | To no surprise, the classifier performed not much better than chance. | Es überrascht nicht, dass der Klassifikator nicht viel besser war als der Zufall. | Non sorprende che il classificatore non abbia funzionato di gran lunga meglio del caso. | 毫不奇怪,分类器的表现并没有比随机猜测更好。 |
xiSxNRoOzm_0019 | xiSxNRoOzm | MCIF_DATA/LONG_AUDIO/xiSxNRoOzm.wav | MCIF_DATA/PDF/xiSxNRoOzm.pdf | 146.949 | 7.079 | 6 | Given the low occurrence of dissonance and absence of any prior such data set, we are facing the problem of absolute rarity. | Angesichts der Seltenheit von Dissonanzen und des Fehlens eines früheren Datensatzes dieser Art stehen wir vor dem Problem der absoluten Seltenheit. | Data la scarsa presenza di dissonanza e l'assenza di qualsiasi precedente insieme di dati di questo tipo, ci troviamo di fronte al problema della rarità assoluta. | 鉴于失调的出现率很低,并且之前没有任何此类数据集,我们面临的是绝对稀有性的问题。 |
xiSxNRoOzm_0020 | xiSxNRoOzm | MCIF_DATA/LONG_AUDIO/xiSxNRoOzm.wav | MCIF_DATA/PDF/xiSxNRoOzm.pdf | 155.481 | 14.146 | 6 | To alleviate this, we experiment over combinations of transfer learning and active learning to annotate such that more dissonant samples can be collected over lesser annotation runs, lowering the overall annotation costs while improving dissonance detection. | Um dieses Problem zu entschärfen, experimentieren wir mit Kombinationen von Transferlernen und aktivem Lernen, um die Annotation so zu gestalten, dass mehr dissonante Proben mit weniger Annotationsläufen erfasst werden können, was die Gesamtkosten der Annotation reduziert und gleichzeitig die Erkennung von Dissonanzen ... | Per ovviare a questo problema, sperimentiamo combinazioni di apprendimento per trasferimento e apprendimento attivo per eseguire l'annotazione, in modo tale che più campioni dissonanti possano essere raccolti su serie di annotazioni minori, riducendo i costi complessivi di annotazione e migliorando il rilevamento della... | 为了缓解这一问题,我们对迁移学习和主动学习的组合进行了实验,以便在较少的注释运行中收集更多的失调样本,从而降低总体注释成本,同时提高失调检测率。 |
xiSxNRoOzm_0021 | xiSxNRoOzm | MCIF_DATA/LONG_AUDIO/xiSxNRoOzm.wav | MCIF_DATA/PDF/xiSxNRoOzm.pdf | 171.855 | 9.922 | 6 | Since the initial model was not able to capture the dissonance class at all, we start the active learning process by transferring weights from closely related tasks. | Da das ursprüngliche Modell nicht in der Lage war, die Dissonanzklasse zu erfassen, beginnen wir den aktiven Lernprozess, indem wir Gewichte von eng verwandten Aufgaben übertragen. | Poiché il modello iniziale non era affatto in grado di catturare la classe di dissonanza, iniziamo il processo di apprendimento attivo trasferendo i pesi da attività strettamente correlate. | 由于初始模型根本无法捕捉到失调类别,因此我们通过从密切相关的任务中迁移权重来开始主动学习过程。 |
xiSxNRoOzm_0022 | xiSxNRoOzm | MCIF_DATA/LONG_AUDIO/xiSxNRoOzm.wav | MCIF_DATA/PDF/xiSxNRoOzm.pdf | 184.145 | 24.926 | 6 | We transfer from two different tasks: topic independent dissonance stance classification, a task that determines if two debate statements from different people are in agreement or in disagreement, irrespective of topic, called debate here, and on binary classification of expansion and comparison classes of PDTB since t... | Wir gehen von zwei verschiedenen Aufgaben aus: der themenunabhängigen Klassifikation von Dissonanzhaltungen, einer Aufgabe, die unabhängig vom Thema bestimmt, ob zwei Debattenaussagen von verschiedenen Personen übereinstimmen oder nicht, hier Debatte genannt, und der binären Klassifikation von PDTB-Expansions- und Verg... | Effettuiamo il trasferimento da due diverse attività: la classificazione della posizione di dissonanza indipendente dall'argomento, un'attività che determina se due affermazioni di dibattito di persone diverse sono d'accordo o in disaccordo, indipendentemente dall'argomento, qui chiamato dibattito, e sulla classificazi... | 我们从两个不同的任务中迁移:主题独立的失调立场分类,这个任务用于确定来自不同人的两个辩论声明是否一致,无论主题如何,在这里称之为辩论,以及 PDTB 的扩展和比较类的二元分类,因为这两者与失调和不失调的概念密切相关,我们在这里称之为 CE。 |
xiSxNRoOzm_0023 | xiSxNRoOzm | MCIF_DATA/LONG_AUDIO/xiSxNRoOzm.wav | MCIF_DATA/PDF/xiSxNRoOzm.pdf | 209.733 | 8.964 | 6 | We find that on transferring the zero-shot performance on the annotated data set is already much better than chance with the best, with AUC .62. | Wir stellen fest, dass die Zero-Shot-Performance auf dem annotierten Datensatz mit einer AUC von 0,62 bereits deutlich besser als der Zufall ist. | Notiamo che trasferendo le prestazioni zero-shot sul set di dati annotati, queste sono di gran lunga migliori rispetto al caso con il migliore, con AUC .62. | 我们发现,在迁移注释数据集上的零样本表现时,已经比随机猜测更好,最好的 AUC 为 0.62。 |
xiSxNRoOzm_0024 | xiSxNRoOzm | MCIF_DATA/LONG_AUDIO/xiSxNRoOzm.wav | MCIF_DATA/PDF/xiSxNRoOzm.pdf | 221.09 | 9.685 | 6 | Further, on iteratively fine-tuning on both tasks, we find that fine-tuning of CE tasks followed by further fine-tuning on debate yields a much better zero-shot performance. | Darüber hinaus stellen wir bei der iterativen Feinabstimmung der beiden Aufgaben fest, dass die Feinabstimmung der CE-Aufgaben, gefolgt von einer weiteren Feinabstimmung der Debatte, zu einer deutlich besseren Zero-Shot-Leistung führt. | Inoltre, con il fine-tuning iterativo su entrambi i compiti, scopriamo che il fine-tuning dei compiti CE seguita da un'ulteriore messa a punto sul dibattito produce una prestazione zero-shot assai migliore. | 此外,在两个任务上的迭代微调中,我们发现微调 CE 任务,然后进一步微调辩论,可以产生更好的零样本表现。 |
xiSxNRoOzm_0025 | xiSxNRoOzm | MCIF_DATA/LONG_AUDIO/xiSxNRoOzm.wav | MCIF_DATA/PDF/xiSxNRoOzm.pdf | 231.346 | 3.117 | 6 | Thus, this is the model that we use to cold start the active learning. | Daher verwenden wir dieses Modell für den Kaltstart des aktiven Lernens. | Quindi, questo è il modello che utilizziamo per avviare a freddo l'apprendimento attivo. | 因此,这是我们用于冷启动主动学习的模型。 |
xiSxNRoOzm_0026 | xiSxNRoOzm | MCIF_DATA/LONG_AUDIO/xiSxNRoOzm.wav | MCIF_DATA/PDF/xiSxNRoOzm.pdf | 236.026 | 6.458 | 6 | Next, we determine the best method to update a model with new data from each round of active learning and annotations. | Als nächstes bestimmen wir die beste Methode, um das Modell mit neuen Daten aus jeder Runde des aktiven Lernens und der Annotation zu aktualisieren. | Successivamente, determiniamo il metodo migliore per aggiornare un modello con nuovi dati da ogni ciclo di apprendimento attivo e annotazioni. | 接下来,我们确定了使用每轮主动学习和注释中的新数据来更新模型的最佳方法。 |
xiSxNRoOzm_0027 | xiSxNRoOzm | MCIF_DATA/LONG_AUDIO/xiSxNRoOzm.wav | MCIF_DATA/PDF/xiSxNRoOzm.pdf | 243.191 | 8.265 | 6 | ""Cumulative"" accumulates all the data collected from active annotation so far, whereas ""Iterative"" updates the model by training on the latest set of data collected. | „Kumulativ“ sammelt alle bisher gesammelten Daten aus der aktiven Annotation, während „Iterativ“ das Modell durch Training mit dem zuletzt gesammelten Datensatz aktualisiert. | "Cumulative" accumula tutti i dati raccolti dall'annotazione attiva fino a quel momento, mentre "Iterative" aggiorna il modello eseguendo l'addestramento sull'ultimo set di dati raccolti. | “累积”是指迄今为止从主动注释中收集的所有数据,而“迭代”是指通过对最新收集的数据集进行训练来更新模型。 |
xiSxNRoOzm_0028 | xiSxNRoOzm | MCIF_DATA/LONG_AUDIO/xiSxNRoOzm.wav | MCIF_DATA/PDF/xiSxNRoOzm.pdf | 251.462 | 7.401 | 6 | Over the different strategies, we found that Cumulative performed equal or better than Iterative across the board. | Für die verschiedenen Strategien haben wir festgestellt, dass „Kumulativ“ immer gleich gut oder besser als „Iterativ“ funktioniert. | Tra le diverse strategie, abbiamo riscontrato che Cumulative ha avuto un rendimento uguale o migliore su tutta la linea rispetto a Iterative. | 在不同的策略中,我们发现“累积”的表现在各个方面都等于或优于“迭代”。 |
xiSxNRoOzm_0029 | xiSxNRoOzm | MCIF_DATA/LONG_AUDIO/xiSxNRoOzm.wav | MCIF_DATA/PDF/xiSxNRoOzm.pdf | 260.859 | 11.23 | 6 | Next, to improve the number of dissonance examples, we use a Probability-of-Rare-Class strategy — PRC — to select mostly the examples that are highly likely to be descended by the current model at any round of rare. | Um die Anzahl der Dissonanzbeispiele zu erhöhen, verwenden wir eine Probability-of-Rare-Class-Strategie (PRC), um in jeder Rare-Runde die meisten Beispiele auszuwählen, die mit hoher Wahrscheinlichkeit aus dem aktuellen Modell stammen. | Successivamente, per migliorare il numero di esempi di dissonanza, utilizziamo una strategia PRC (Probability-of-Rare-Class) per selezionare principalmente gli esempi che hanno un'alta probabilità di discendere dal modello attuale in qualsiasi round di rarità. | 接下来,为了增加失调示例的数量,我们使用罕见类概率策略 (PRC) 来选择大部分示例,这些示例在任何一轮罕见事件中都极有可能被当前模型所继承。 |
xiSxNRoOzm_0030 | xiSxNRoOzm | MCIF_DATA/LONG_AUDIO/xiSxNRoOzm.wav | MCIF_DATA/PDF/xiSxNRoOzm.pdf | 272.653 | 5.921 | 6 | We compare this to the other state-of-the-art AL strategies that are commonly used in the community. | Wir vergleichen diese Strategie mit anderen modernen AL-Strategien, die in der Community häufig verwendet werden. | Confrontiamo ciò con le altre strategie AL all'avanguardia comunemente usate nella comunità. | 我们将此策略与社区中常用的其他最先进的 AL 策略进行比较。 |
xiSxNRoOzm_0031 | xiSxNRoOzm | MCIF_DATA/LONG_AUDIO/xiSxNRoOzm.wav | MCIF_DATA/PDF/xiSxNRoOzm.pdf | 278.583 | 9.351 | 6 | We find that the proposed PRC strategy works better than other state-of-the-art strategies, although the difference is small. | Wir stellen fest, dass die vorgeschlagene PRC-Strategie besser abschneidet als die anderen modernen Strategien, obwohl der Unterschied gering ist. | Riteniamo che la strategia PRC proposta funzioni meglio di altre strategie all'avanguardia, sebbene la differenza sia esigua. | 我们发现,虽然差异不大,但该 PRC 策略比其他最先进的策略更有效。 |
xiSxNRoOzm_0032 | xiSxNRoOzm | MCIF_DATA/LONG_AUDIO/xiSxNRoOzm.wav | MCIF_DATA/PDF/xiSxNRoOzm.pdf | 288.38 | 3.334 | 6 | Note that the performance is significantly lower for random. | Es ist zu beachten, dass die Leistung mit Random deutlich schlechter ist. | Va notato che le prestazioni sono significativamente inferiori con riferimento al caso. | 请注意,随机策略的表现要差得多。 |
xiSxNRoOzm_0033 | xiSxNRoOzm | MCIF_DATA/LONG_AUDIO/xiSxNRoOzm.wav | MCIF_DATA/PDF/xiSxNRoOzm.pdf | 294.068 | 9.339 | 6 | On further rounds of AL with two best strategies, we improve dissonance classification AUC to 0.75, which is the best performance that we have on the task so far. | In weiteren AL-Runden mit den beiden besten Strategien verbessern wir die AUC der Dissonanzklassifikation auf 0,75, was die beste Leistung ist, die wir bisher bei dieser Aufgabe erreicht haben. | In ulteriori round di AL con le due migliori strategie, miglioriamo l'AUC di classificazione della dissonanza a 0,75, che è la migliore prestazione che abbiamo riscontrato finora sull'attività. | 在采用两种最佳策略的进一步 AL 轮次中,我们将失调分类 AUC 提高到 0.75,这是我们迄今为止在该任务中取得的最佳表现。 |
xiSxNRoOzm_0034 | xiSxNRoOzm | MCIF_DATA/LONG_AUDIO/xiSxNRoOzm.wav | MCIF_DATA/PDF/xiSxNRoOzm.pdf | 304.852 | 4.966 | 6 | We also check the feasibility of each strategy for annotation quality and costs to annotators. | Wir untersuchen auch die Durchführbarkeit jeder Strategie in Bezug auf die Qualität der Kommentare und die Kosten für die Annotatoren. | Verifichiamo anche la fattibilità di ciascuna strategia per la qualità delle annotazioni e i costi per gli annotatori. | 我们还检查了每种策略在注释质量和注释成本方面的可行性。 |
xiSxNRoOzm_0035 | xiSxNRoOzm | MCIF_DATA/LONG_AUDIO/xiSxNRoOzm.wav | MCIF_DATA/PDF/xiSxNRoOzm.pdf | 310.186 | 4.522 | 6 | We find that PRC has the highest percentage of dissonance and works best for rare class. | Wir stellen fest, dass PRC den höchsten Prozentsatz an Dissonanzen aufweist und am besten bei der seltenen Klasse funktioniert. | Riteniamo che PRC abbia la più alta percentuale di dissonanza e funzioni meglio per la classe rara. | 我们发现,PRC 的失调百分比最高,并且对稀有类别效果最佳。 |
xiSxNRoOzm_0036 | xiSxNRoOzm | MCIF_DATA/LONG_AUDIO/xiSxNRoOzm.wav | MCIF_DATA/PDF/xiSxNRoOzm.pdf | 315.136 | 3.795 | 6 | However, the annotators also find the examples difficult. | Allerdings finden die Annotatoren die Beispiele auch schwierig. | Tuttavia, anche gli annotatori trovano difficili gli esempi. | 然而,注释者也发现这些例子很难。 |
xiSxNRoOzm_0037 | xiSxNRoOzm | MCIF_DATA/LONG_AUDIO/xiSxNRoOzm.wav | MCIF_DATA/PDF/xiSxNRoOzm.pdf | 321.268 | 9.585 | 6 | In summary, we find that PRC is a simple AL strategy for rare class acquisition and cold starting AL with appropriately designed transfer learning task and help significantly. | Zusammenfassend stellen wir fest, dass PRC eine einfache AL-Strategie ist, um seltene Klassen und Kaltstart-AL mit einer gut gestalteten Transfer-Lernaufgabe zu erfassen, und dass sie sehr hilfreich ist. | In sintesi, riteniamo che PRC sia una semplice strategia di AL per l'acquisizione di classi rare e l'avvio a freddo di AL con un'attività di apprendimento per trasferimento opportunamente progettata e un aiuto significativo. | 总结来说,我们发现,PRC 是一种用于稀有类别获取和冷启动 AL 的简单 AL 策略,它具有适当设计的迁移学习任务,并能够提供显著帮助。 |
xiSxNRoOzm_0038 | xiSxNRoOzm | MCIF_DATA/LONG_AUDIO/xiSxNRoOzm.wav | MCIF_DATA/PDF/xiSxNRoOzm.pdf | 331.318 | 8.384 | 6 | We also find that iterative update is useful for transfer learning from a different domain, whereas in domain active annotations benefit from cumulative update. | Wir stellen auch fest, dass die iterative Aktualisierung für das Transferlernen aus einer anderen Domain nützlich ist, während aktive Annotationen in der Domain von der kumulativen Aktualisierung profitieren. | Inoltre, crediamo che l'aggiornamento iterativo è utile per l'apprendimento per trasferimento da un dominio diverso, mentre le annotazioni attive nel dominio beneficiano dell'aggiornamento cumulativo. | 我们还发现,迭代更新对于来自不同领域的迁移学习很有用,而在领域内活跃的注释则受益于累积更新。 |
xiSxNRoOzm_0039 | xiSxNRoOzm | MCIF_DATA/LONG_AUDIO/xiSxNRoOzm.wav | MCIF_DATA/PDF/xiSxNRoOzm.pdf | 341.175 | 2.791 | 6 | These are the links to our core data set and our paper. | Dies sind die Links zu unserem Kerndatensatz und unserer Arbeit. | Questi sono i link al nostro set di dati di base e al nostro articolo. | 这些是我们核心数据集和论文的链接。 |
xiSxNRoOzm_0040 | xiSxNRoOzm | MCIF_DATA/LONG_AUDIO/xiSxNRoOzm.wav | MCIF_DATA/PDF/xiSxNRoOzm.pdf | 345.039 | 2.312 | 6 | Feel free to get in touch with us if you have any questions. | Wenn Sie Fragen haben, zögern Sie bitte nicht, uns zu kontaktieren. | Non esitate a contattarci in caso di domande. | 如果大家有任何疑问,欢迎随时与我们联系。 |
xiSxNRoOzm_0041 | xiSxNRoOzm | MCIF_DATA/LONG_AUDIO/xiSxNRoOzm.wav | MCIF_DATA/PDF/xiSxNRoOzm.pdf | 348.046 | 0.511 | 6 | Thank you." | Vielen Dank. | Grazie." | 谢谢。 |
miPjvjWOvI_0000 | miPjvjWOvI | MCIF_DATA/LONG_AUDIO/miPjvjWOvI.wav | MCIF_DATA/PDF/miPjvjWOvI.pdf | 0.031 | 0.899 | 7 | "Hi! | Hallo! | "Ciao! | 嗨! |
miPjvjWOvI_0001 | miPjvjWOvI | MCIF_DATA/LONG_AUDIO/miPjvjWOvI.wav | MCIF_DATA/PDF/miPjvjWOvI.pdf | 1.656 | 8.583 | 7 | I'm going to talk about our work on ""Resolving Indirect Referring Expressions for Entity Selection"", in which we introduce the AltEntities Corpus. | Ich werde über unsere Arbeit „Resolving Indirect Referring Expressions for Entity Selection“ sprechen, in der wir das AltEntities-Korpus vorstellen. | Parlerò del nostro lavoro su "Resolving Indirect Referring Expressions for Entity Selection", in cui introduciamo il corpus AltEntities. | 我将谈谈我们的论文《实体选择的间接指称表达式解析》,其中我们介绍了 AltEntities 语料库。 |
miPjvjWOvI_0002 | miPjvjWOvI | MCIF_DATA/LONG_AUDIO/miPjvjWOvI.wav | MCIF_DATA/PDF/miPjvjWOvI.pdf | 11.066 | 6.299 | 7 | My name is Javad Hosseini and this is a joint work with Filip Radlinski, Silvia Pareti, and Annie Louis. | Mein Name ist Javad Hosseini und diese Zusammenarbeit entstand mit Filip Radlinski, Silvia Pareti und Annie Louis. | Mi chiamo Javad Hosseini e questo è un lavoro congiunto con Filip Radlinski, Silvia Pareti e Annie Louis. | 我叫 Javad Hosseini,这是我与 Filip Radlinski、Silvia Pareti 和 Annie Louis 合作编写的论文。 |
miPjvjWOvI_0003 | miPjvjWOvI | MCIF_DATA/LONG_AUDIO/miPjvjWOvI.wav | MCIF_DATA/PDF/miPjvjWOvI.pdf | 19.318 | 3.411 | 7 | Our goal is to understand users’ language when they want to make a choice. | Unser Ziel ist es, die Sprache der Benutzer zu verstehen, wenn sie eine Wahl treffen. | Il nostro obiettivo è comprendere il linguaggio degli utenti quando intendono compiere una scelta. | 我们的目标是了解用户在做出选择时的语言。 |
miPjvjWOvI_0004 | miPjvjWOvI | MCIF_DATA/LONG_AUDIO/miPjvjWOvI.wav | MCIF_DATA/PDF/miPjvjWOvI.pdf | 23.69 | 2.202 | 7 | Consider this alternative question. | Betrachten Sie diese Alternativen-Frage. | Consideriamo questa domanda alternativa. | 请考虑这个选择问题。 |
miPjvjWOvI_0005 | miPjvjWOvI | MCIF_DATA/LONG_AUDIO/miPjvjWOvI.wav | MCIF_DATA/PDF/miPjvjWOvI.pdf | 25.886 | 3.622 | 7 | ""Did you mean 'Easy on Me' or 'I Gotta Feeling'?"" | „Meinten Sie 'Easy on Me' oder 'I Gotta Feeling'?“ | "Intendevi 'Easy on Me' o 'I Gotta Feeling'?" | “你是说‘Easy on Me’还是‘I Gotta Feeling’?” |
miPjvjWOvI_0006 | miPjvjWOvI | MCIF_DATA/LONG_AUDIO/miPjvjWOvI.wav | MCIF_DATA/PDF/miPjvjWOvI.pdf | 29.408 | 3.438 | 7 | Here, a user wants to select between one of these two songs. | Hier möchte ein Benutzer zwischen diesen beiden Liedern wählen. | Qui, un utente vuole scegliere una tra queste due canzoni. | 在这一示例中,用户想在这两首歌中选择一首。 |
miPjvjWOvI_0007 | miPjvjWOvI | MCIF_DATA/LONG_AUDIO/miPjvjWOvI.wav | MCIF_DATA/PDF/miPjvjWOvI.pdf | 34.4 | 9.946 | 7 | The most obvious thing is to use a direct reference, for example by saying the name of the song ""Easy on Me"" or its position, ""the first one"". | Am naheliegendsten ist es, einen direkten Bezug zu verwenden, etwa den Namen des Liedes, „Easy on Me“, oder seine Position, „das erste“. | La cosa più ovvia è usare un riferimento diretto, ad esempio pronunciando il nome della canzone "Easy on Me" o la sua posizione, "la prima". | 最明显的方法是使用直接指称,例如说出歌曲名称“Easy on Me”或其位置“第一首”。 |
miPjvjWOvI_0008 | miPjvjWOvI | MCIF_DATA/LONG_AUDIO/miPjvjWOvI.wav | MCIF_DATA/PDF/miPjvjWOvI.pdf | 44.471 | 6.622 | 7 | But sometimes an indirect reference is more appropriate to have a more natural conversation. | Manchmal ist jedoch ein indirekter Bezug besser geeignet, um ein natürlicheres Gespräch zu führen. | Ma a volte un riferimento indiretto è più appropriato per avere una conversazione più naturale. | 但有时候,间接指称更适合进行更自然的对话。 |
miPjvjWOvI_0009 | miPjvjWOvI | MCIF_DATA/LONG_AUDIO/miPjvjWOvI.wav | MCIF_DATA/PDF/miPjvjWOvI.pdf | 51.566 | 3.636 | 7 | This could happen when the user cannot remember the name of the song. | Dies kann der Fall sein, wenn sich der Benutzer nicht an den Namen des Liedes erinnern kann. | Questo potrebbe accadere quando l'utente non riesce a ricordare il nome della canzone. | 当用户不记得歌曲的名称时,就可能会出现这种情况。 |
miPjvjWOvI_0010 | miPjvjWOvI | MCIF_DATA/LONG_AUDIO/miPjvjWOvI.wav | MCIF_DATA/PDF/miPjvjWOvI.pdf | 56.566 | 4.934 | 7 | Or the pronunciations are too similar to each other and hard to disambiguate. | Oder die Aussprachen sind zu ähnlich und schwer zu unterscheiden. | Oppure le pronunce sono troppo simili tra loro e difficili da disambiguare. | 或者两首歌的名字发音太过相似,难以区分。 |
miPjvjWOvI_0011 | miPjvjWOvI | MCIF_DATA/LONG_AUDIO/miPjvjWOvI.wav | MCIF_DATA/PDF/miPjvjWOvI.pdf | 62.751 | 3.458 | 7 | Or when the user wants to specify a preference. | Oder wenn der Benutzer eine Vorliebe ausdrücken möchte. | O, ancora, quando l'utente vuole specificare una preferenza. | 或者,当用户想指定偏好时,也可能会出现这种情况。 |
miPjvjWOvI_0012 | miPjvjWOvI | MCIF_DATA/LONG_AUDIO/miPjvjWOvI.wav | MCIF_DATA/PDF/miPjvjWOvI.pdf | 67.016 | 8.284 | 7 | Here are some examples of indirect references for example, ""the newer one"" or ""the song that's not energetic."" | Hier sind einige Beispiele für indirekte Verweise wie „das neueste“ oder „das Lied, das nicht energiegeladen ist“. | Ecco alcuni esempi di riferimenti indiretti, ad esempio "la più recente" o "la canzone non vivace". | 以下是一些间接指称的例子,例如“较新的那一首”或“比较抒情的那一首”。 |
miPjvjWOvI_0013 | miPjvjWOvI | MCIF_DATA/LONG_AUDIO/miPjvjWOvI.wav | MCIF_DATA/PDF/miPjvjWOvI.pdf | 75.294 | 7.337 | 7 | This is an important problem in conversational systems and also for benchmarking LLMs' entity understanding. | Dies ist ein wichtiges Problem für Konversationssysteme und auch für das Benchmarking des Entity Understanding von LLMs. | Questo è un problema importante nei sistemi conversazionali e anche per il benchmarking della comprensione dell'entità dei LLM. | 这是对话系统中的一个重要问题,也是对 LLM 实体理解进行基准测试的一个重要问题。 |
miPjvjWOvI_0014 | miPjvjWOvI | MCIF_DATA/LONG_AUDIO/miPjvjWOvI.wav | MCIF_DATA/PDF/miPjvjWOvI.pdf | 83.056 | 9.391 | 7 | We're not aware of a larger-scale public data set for the task, so we collect one using crowd annotation. | Uns ist kein großer öffentlicher Datensatz für diese Aufgabe bekannt, also sammeln wir selbst einen mit Hilfe von Crowd-Annotation. | Non siamo a conoscenza di alcun set di dati pubblici su larga scala per l'attività, quindi ne raccogliamo uno utilizzando l'annotazione della folla. | 我们没有发现针对这项任务的大规模公共数据集,因此我们使用众包注释来收集数据集。 |
miPjvjWOvI_0015 | miPjvjWOvI | MCIF_DATA/LONG_AUDIO/miPjvjWOvI.wav | MCIF_DATA/PDF/miPjvjWOvI.pdf | 92.431 | 5.839 | 7 | Our data set covers three different domains: music, books, and recipes. | Unser Datensatz umfasst drei verschiedene Domains: Musik, Bücher und Rezepte. | Il nostro set di dati copre tre diversi domini: musica, libri e ricette. | 我们的数据集涵盖三个不同的领域:音乐、书籍和食谱。 |
miPjvjWOvI_0016 | miPjvjWOvI | MCIF_DATA/LONG_AUDIO/miPjvjWOvI.wav | MCIF_DATA/PDF/miPjvjWOvI.pdf | 100.225 | 7.139 | 7 | Our data set collection methodology emphasizes informality using a cartoon completion setup. | Unsere Methode der Datensammlung betont den informellen Charakter, indem wir einen Cartoon als Ergänzung verwenden. | La nostra metodologia di raccolta dei dati sottolinea l'informalità utilizzando una configurazione di completamento a fumetti. | 我们的数据集收集方法强调使用卡通完成设置来达到非正式性。 |
miPjvjWOvI_0017 | miPjvjWOvI | MCIF_DATA/LONG_AUDIO/miPjvjWOvI.wav | MCIF_DATA/PDF/miPjvjWOvI.pdf | 107.372 | 3.154 | 7 | The cartoon has three speech bubbles. | Der Cartoon besteht aus drei Sprechblasen. | Il fumetto ha tre nuvolette. | 这幅漫画有三个对话框。 |
miPjvjWOvI_0018 | miPjvjWOvI | MCIF_DATA/LONG_AUDIO/miPjvjWOvI.wav | MCIF_DATA/PDF/miPjvjWOvI.pdf | 110.533 | 5.476 | 7 | In the first bubble, Bob says, ""Remember that song we were listening to yesterday?"" | In der ersten Sprechblase sagt Bob: „Erinnerst du dich an das Lied, das wir gestern gehört haben?“ | Nella prima, Bob dice: "Ricordi quella canzone che stavamo ascoltando ieri?" | 在第一个气泡中,Bob 说:“记得我们昨天听的那首歌吗?” |
miPjvjWOvI_0019 | miPjvjWOvI | MCIF_DATA/LONG_AUDIO/miPjvjWOvI.wav | MCIF_DATA/PDF/miPjvjWOvI.pdf | 117.442 | 3.431 | 7 | And with that, Bob sets the dialogue context. | Damit gibt Bob den Kontext für den Dialog vor. | In tal modo, Bob imposta il contesto del dialogo. | 这样,Bob 就设置了对话的上下文。 |
miPjvjWOvI_0020 | miPjvjWOvI | MCIF_DATA/LONG_AUDIO/miPjvjWOvI.wav | MCIF_DATA/PDF/miPjvjWOvI.pdf | 122.549 | 8.168 | 7 | In the second speech bubble, Alice says, ""Do you mean 'Easy on Me' or 'I Gotta Feeling'?"" | In der zweiten Sprechblase sagt Alice: „Meinst du ‚Easy on Me‘ oder ‚I Gotta Feeling‘?“ | Nella seconda, Alice dice: "Intendi 'Easy on Me' o 'I Gotta Feeling'?" | 在第二个气泡中,Alice 说:“你是说‘Easy on Me’还是‘I Gotta Feeling’?” |
miPjvjWOvI_0021 | miPjvjWOvI | MCIF_DATA/LONG_AUDIO/miPjvjWOvI.wav | MCIF_DATA/PDF/miPjvjWOvI.pdf | 130.725 | 1.673 | 7 | Which is the alternative question. | Das ist die Alternativfrage. | Che è la domanda alternativa. | 这是一个选择问题。 |
miPjvjWOvI_0022 | miPjvjWOvI | MCIF_DATA/LONG_AUDIO/miPjvjWOvI.wav | MCIF_DATA/PDF/miPjvjWOvI.pdf | 132.382 | 9.691 | 7 | And in the third speech bubble, Bob uses an indirect reference to select one of these entities, for example, ""the newer one."" | Und in der dritten Sprechblase verwendet Bob einen indirekten Bezug, um eine der Entitäten auszuwählen, zum Beispiel „das neuere“. | E nella terza, Bob usa un riferimento indiretto per selezionare una di queste entità, ad esempio "la più recente". | 在第三个气泡中,Bob 使用间接指称来选择其中一个实体,例如“较新的那一首”。 |
miPjvjWOvI_0023 | miPjvjWOvI | MCIF_DATA/LONG_AUDIO/miPjvjWOvI.wav | MCIF_DATA/PDF/miPjvjWOvI.pdf | 142.059 | 6.721 | 7 | We provide the first and second speech bubbles automatically, but the third one is filled in by the annotator. | Die erste und zweite Sprechblase werden von uns automatisch zur Verfügung gestellt, die dritte Sprechblase wird vom Annotator ausgefüllt. | La prima e la seconda nuvoletta vengono generate automaticamente, mentre la terza deve essere compilata dall'annotatore. | 我们会自动提供第一个和第二个对话框,但第三个对话框由注释者填写。 |
miPjvjWOvI_0024 | miPjvjWOvI | MCIF_DATA/LONG_AUDIO/miPjvjWOvI.wav | MCIF_DATA/PDF/miPjvjWOvI.pdf | 150.712 | 5.075 | 7 | The first speech bubble is chosen from a few manual prompts per domain. | Die erste Sprechblase wird aus einer Anzahl manueller Prompts pro Domain ausgewählt. | La prima viene scelta tra alcuni suggerimenti manuali per dominio. | 第一个对话框是从每个域中的几个手动提示中选择的。 |
miPjvjWOvI_0025 | miPjvjWOvI | MCIF_DATA/LONG_AUDIO/miPjvjWOvI.wav | MCIF_DATA/PDF/miPjvjWOvI.pdf | 157.041 | 4.757 | 7 | The second one, which is the alternative question is generated as follows. | Die zweite Sprechblase, die Alternativfrage, wird wie folgt generiert. | La seconda, che è la domanda alternativa, viene generata come segue. | 第二个是选择问题,它是这样生成的: |
miPjvjWOvI_0026 | miPjvjWOvI | MCIF_DATA/LONG_AUDIO/miPjvjWOvI.wav | MCIF_DATA/PDF/miPjvjWOvI.pdf | 163.071 | 5.263 | 7 | We always use a simple template. | Wir benutzen immer eine einfache Vorlage. | Usiamo sempre un modello semplice. | 我们总是使用简单的模板。 |
miPjvjWOvI_0027 | miPjvjWOvI | MCIF_DATA/LONG_AUDIO/miPjvjWOvI.wav | MCIF_DATA/PDF/miPjvjWOvI.pdf | 168.427 | 1.487 | 7 | Do you mean A or B? | Meinst du A oder B? | Intendi A o B? | 你是说 A 还是 B? |
miPjvjWOvI_0028 | miPjvjWOvI | MCIF_DATA/LONG_AUDIO/miPjvjWOvI.wav | MCIF_DATA/PDF/miPjvjWOvI.pdf | 171.168 | 3.135 | 7 | Where A and B are samples from Wikipedia. | Wobei A und B Beispiele aus Wikipedia sind. | Dove A e B sono campioni da Wikipedia. | A 和 B 是来自维基百科的样本。 |
miPjvjWOvI_0029 | miPjvjWOvI | MCIF_DATA/LONG_AUDIO/miPjvjWOvI.wav | MCIF_DATA/PDF/miPjvjWOvI.pdf | 176.141 | 2.473 | 7 | Here are the different sampling methods we've used. | Hier sehen Sie die verschiedenen Stichprobenverfahren, die wir verwendet haben. | Ecco i diversi metodi di campionamento che abbiamo utilizzato. | 以下是我们使用的不同抽样方法。 |
miPjvjWOvI_0030 | miPjvjWOvI | MCIF_DATA/LONG_AUDIO/miPjvjWOvI.wav | MCIF_DATA/PDF/miPjvjWOvI.pdf | 179.122 | 7.766 | 7 | When we move higher in the list, the entities become more similar to each other and it's usually harder to make the disambiguation. | Je höher wir in der Liste kommen, desto ähnlicher werden die Begriffe und desto schwieriger wird es, eine eindeutige Antwort zu finden. | Quando ci spostiamo più in alto nell'elenco, le entità diventano più simili tra loro e di solito è più difficile disambiguarle. | 当我们在列表中向上移动时,实体彼此变得更加相似,通常更难以辨认。 |
miPjvjWOvI_0031 | miPjvjWOvI | MCIF_DATA/LONG_AUDIO/miPjvjWOvI.wav | MCIF_DATA/PDF/miPjvjWOvI.pdf | 187.881 | 2.15 | 7 | The first one is uniform at random. | Die erste ist zufällig gleich. | La prima è casualmente uniforme. | 第一个情况是随机统一。 |
miPjvjWOvI_0032 | miPjvjWOvI | MCIF_DATA/LONG_AUDIO/miPjvjWOvI.wav | MCIF_DATA/PDF/miPjvjWOvI.pdf | 190.8 | 6.219 | 7 | The second one is when the entities have similar titles, for example, two books with the name ""The Return"". | Die zweite ist, wenn die Entitäten ähnliche Titel haben, zum Beispiel zwei Bücher mit dem Titel „Die Rückkehr“. | La seconda è quando le entità hanno titoli simili, ad esempio due libri con il nome "The Return". | 第二种情况是实体的标题相似,例如,两本名为《The Return》的书。 |
miPjvjWOvI_0033 | miPjvjWOvI | MCIF_DATA/LONG_AUDIO/miPjvjWOvI.wav | MCIF_DATA/PDF/miPjvjWOvI.pdf | 197.031 | 3.754 | 7 | The third one is when they have similar descriptions on Wikipedia. | Die dritte ist, wenn sie ähnliche Beschreibungen in Wikipedia haben. | La terza è quando presentano descrizioni simili su Wikipedia. | 第三种情况是它们在维基百科上有相似的描述。 |
miPjvjWOvI_0034 | miPjvjWOvI | MCIF_DATA/LONG_AUDIO/miPjvjWOvI.wav | MCIF_DATA/PDF/miPjvjWOvI.pdf | 201.373 | 5.585 | 7 | And finally when they have similar info boxes or attributes on Wikipedia. | Und schließlich, wenn sie ähnliche Infoboxen oder Attribute auf Wikipedia haben. | E, infine, quando hanno riquadri informativi o attributi simili su Wikipedia. | 最后是它们在维基百科上具有相似的信息框或属性。 |
miPjvjWOvI_0035 | miPjvjWOvI | MCIF_DATA/LONG_AUDIO/miPjvjWOvI.wav | MCIF_DATA/PDF/miPjvjWOvI.pdf | 206.943 | 8.129 | 7 | For example, the same genre or the same artist for a song. | Zum Beispiel das gleiche Genre oder der gleiche Interpret bei einem Lied. | Ad esempio, lo stesso genere o lo stesso artista per una canzone. | 例如,歌曲的类型或艺术家相同。 |
miPjvjWOvI_0036 | miPjvjWOvI | MCIF_DATA/LONG_AUDIO/miPjvjWOvI.wav | MCIF_DATA/PDF/miPjvjWOvI.pdf | 215.037 | 10.204 | 7 | When we show this alternative question to the annotators, they know the name of these entities, but they don't necessarily know about the entities. | Wenn wir den Annotatoren diese Alternativfrage stellen, kennen sie die Namen dieser Entitäten, aber sie wissen nicht unbedingt etwas von den Entitäten. | Quando mostriamo questa domanda alternativa agli annotatori, conoscono il nome di queste entità, ma non necessariamente le entità stesse. | 当我们向注释者展示这个选择问题时,他们知道这些实体的名称,但他们不一定知道这些实体。 |
miPjvjWOvI_0037 | miPjvjWOvI | MCIF_DATA/LONG_AUDIO/miPjvjWOvI.wav | MCIF_DATA/PDF/miPjvjWOvI.pdf | 226.128 | 4.716 | 7 | So what we do is that we show some background knowledge about the two entities. | Wir zeigen also etwas Hintergrundwissen über die beiden Entitäten. | Quindi, quello che facciamo è mostrare alcune conoscenze di base sulle due entità. | 我们要做的是展示关于这两个实体的一些背景知识。 |
miPjvjWOvI_0038 | miPjvjWOvI | MCIF_DATA/LONG_AUDIO/miPjvjWOvI.wav | MCIF_DATA/PDF/miPjvjWOvI.pdf | 231.53 | 12.161 | 7 | For songs, we simply show a Google search link to each song and then ask the annotators to listen to at least some of each song, and read about each song. | Bei den Songs zeigen wir einfach einen Google-Suchlink zu jedem Song und bitten die Annotatoren, sich zumindest einen Teil des Lieds anzuhören und etwas über das Lied zu lesen. | Per le canzoni, mostriamo semplicemente un link di ricerca di Google per ogni canzone e poi chiediamo agli annotatori di ascoltare alcune parti e di leggere qualcosa su ciascuna. | 对于歌曲,我们只是显示每首歌的谷歌搜索链接,然后要求注释者至少听一听每首歌的一部分,并阅读每首歌的信息。 |
miPjvjWOvI_0039 | miPjvjWOvI | MCIF_DATA/LONG_AUDIO/miPjvjWOvI.wav | MCIF_DATA/PDF/miPjvjWOvI.pdf | 243.707 | 8.572 | 7 | Here's for example, the Google search result for the song ""Easy on Me."" | Hier ist zum Beispiel das Google-Suchergebnis für das Lied „Easy on Me“. | Ecco ad esempio il risultato della ricerca su Google per la canzone "Easy on Me". | 例如,这是歌曲“Easy on Me”的谷歌搜索结果。 |
miPjvjWOvI_0040 | miPjvjWOvI | MCIF_DATA/LONG_AUDIO/miPjvjWOvI.wav | MCIF_DATA/PDF/miPjvjWOvI.pdf | 252.268 | 6.316 | 7 | For the recipes and books domain, we show some background text from Wikipedia. | Bei den Domains Rezepte und Bücher zeigen wir einige Hintergrundtexte aus Wikipedia. | Per il dominio delle ricette e dei libri, mostriamo alcuni testi di sfondo da Wikipedia. | 对于食谱和书籍领域,我们会展示来自维基百科的一些背景文本。 |
miPjvjWOvI_0041 | miPjvjWOvI | MCIF_DATA/LONG_AUDIO/miPjvjWOvI.wav | MCIF_DATA/PDF/miPjvjWOvI.pdf | 259.713 | 10.302 | 7 | For recipes, we additionally show their images, again from Wikipedia, so that the annotators know how they look like. | Bei den Rezepten zeigen wir auch Bilder, ebenfalls aus Wikipedia, damit die Annotatoren wissen, wie sie aussehen. | Per le ricette, mostriamo anche le loro immagini, tratte sempre da Wikipedia, in modo che gli annotatori sappiano che aspetto hanno. | 对于食谱,我们还会展示它们的图像(也是来自维基百科),以便注释者知道它们的外观。 |
miPjvjWOvI_0042 | miPjvjWOvI | MCIF_DATA/LONG_AUDIO/miPjvjWOvI.wav | MCIF_DATA/PDF/miPjvjWOvI.pdf | 270.023 | 11.289 | 7 | Then, we asked the annotators to pick one of these entities, for example, here's the first one, and describe them using three to five indirect referring expressions. | Dann haben wir die Annotatoren gebeten, eine dieser Entitäten auszuwählen, zum Beispiel hier ist die erste, und sie mit drei bis fünf indirekten Bezugsausdrücken zu beschreiben. | Successivamente, abbiamo chiesto agli annotatori di scegliere una di queste entità, ecco ad esempio la prima, e descriverle usando da tre a cinque espressioni di riferimento indirette. | 然后,我们要求注释者从这些实体中选择一个,例如,这是第一个,并使用三到五个间接指称表达式来描述它们。 |
miPjvjWOvI_0043 | miPjvjWOvI | MCIF_DATA/LONG_AUDIO/miPjvjWOvI.wav | MCIF_DATA/PDF/miPjvjWOvI.pdf | 282.064 | 2.05 | 7 | For example, the one with the piano music. | Zum Beispiel „das Lied mit Klaviermusik“. | Ad esempio, quella con la musica per pianoforte. | 例如,有钢琴曲伴奏的。 |
miPjvjWOvI_0044 | miPjvjWOvI | MCIF_DATA/LONG_AUDIO/miPjvjWOvI.wav | MCIF_DATA/PDF/miPjvjWOvI.pdf | 284.806 | 2.212 | 7 | Here are some examples from our dataset. | Hier sind einige Beispiele aus unserem Datensatz. | Ecco alcuni esempi dal nostro set di dati. | 以下是我们数据集中的一些例子。 |
miPjvjWOvI_0045 | miPjvjWOvI | MCIF_DATA/LONG_AUDIO/miPjvjWOvI.wav | MCIF_DATA/PDF/miPjvjWOvI.pdf | 287.747 | 11.131 | 7 | For example, ""the one without words"", ""not the one with the 12 year old boy"", or ""the fictional one"", or ""comes from Azerbaijan"", and so on. | Zum Beispiel „das ohne Worte“, „nicht das mit dem 12-jährigen Jungen“, oder „das fiktive“, oder „kommt aus Aserbaidschan“, und so weiter. | Ad esempio, "quella senza parole", "non quella con il ragazzo di 12 anni", o "quella immaginaria", o "viene dall'Azerbaigian", e così via. | 例如,“没有文字的”、“不是那个有 12 岁男孩的”、“虚构的”、“来自阿塞拜疆”等等。 |
miPjvjWOvI_0046 | miPjvjWOvI | MCIF_DATA/LONG_AUDIO/miPjvjWOvI.wav | MCIF_DATA/PDF/miPjvjWOvI.pdf | 298.865 | 8.502 | 7 | The AltEntities Corpus has 6,000 alternative questions across three domains, and it has 42,000 indirect referring expressions. | Das AltEntities-Korpus enthält 6.000 Alternativfragen aus drei Domains und 42.000 indirekt verweisende Ausdrücke. | L'AltEntities Corpus consta di 6.000 domande alternative in tre domini e di 42.000 espressioni di riferimento indirette. | AltEntities 语料库在三个领域中有 6,000 个选择问题,并有 42,000 个间接指称表达式。 |
miPjvjWOvI_0047 | miPjvjWOvI | MCIF_DATA/LONG_AUDIO/miPjvjWOvI.wav | MCIF_DATA/PDF/miPjvjWOvI.pdf | 308.418 | 3.974 | 7 | Results with T5 XL model are summarized below. | Die Ergebnisse mit dem T5 XL Modell sind unten zusammengefasst. | I risultati con il modello T5 XL sono riassunti di seguito. | 下面总结了使用 T5 XL 模型的结果。 |
miPjvjWOvI_0048 | miPjvjWOvI | MCIF_DATA/LONG_AUDIO/miPjvjWOvI.wav | MCIF_DATA/PDF/miPjvjWOvI.pdf | 313.078 | 9.606 | 7 | If the language model has access to the exact same background knowledge as the annotators, then the accuracy is really high, it's around 92 to 95%. | Wenn das Sprachmodell Zugriff auf genau dasselbe Hintergrundwissen hat wie die Annotatoren, dann ist die Genauigkeit wirklich hoch, sie liegt zwischen 92 und 95 %. | Se il modello linguistico ha accesso alle stesse identiche conoscenze di base degli annotatori, allora l'accuratezza è decisamente alta, intorno al 92-95%. | 如果语言模型可以访问与注释者完全相同的背景知识,那么准确率就会非常高,大约在 92% 到 95% 之间。 |
miPjvjWOvI_0049 | miPjvjWOvI | MCIF_DATA/LONG_AUDIO/miPjvjWOvI.wav | MCIF_DATA/PDF/miPjvjWOvI.pdf | 323.858 | 1.508 | 7 | But this is not realistic. | Dies ist jedoch nicht realistisch. | Ma ciò non è realistico. | 但这是不现实的。 |
miPjvjWOvI_0050 | miPjvjWOvI | MCIF_DATA/LONG_AUDIO/miPjvjWOvI.wav | MCIF_DATA/PDF/miPjvjWOvI.pdf | 327.816 | 10.301 | 7 | If the language model has access to some partially overlapping background knowledge, then the accuracy is between 82 to 87%, which is more realistic. | Wenn das Sprachmodell Zugang zu Hintergrundwissen hat, das sich teilweise überschneidet, liegt die Genauigkeit zwischen 82 und 87 %, was realistischer ist. | Se il modello linguistico ha accesso ad alcune conoscenze di base parzialmente sovrapposte, l'accuratezza è compresa tra l'82 e l'87%, il che è più realistico. | 如果语言模型可以访问一些部分重叠的背景知识,那么准确率在 82% 到 87% 之间,这更符合实际。 |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.