Datasets:

lp
stringclasses
3 values
src
stringclasses
415 values
mt
stringlengths
1
388
ref
stringlengths
3
370
raw
int64
0
100
annotators
int64
1
1
domain
stringclasses
1 value
year
int64
2.02k
2.02k
task
stringclasses
2 values
sys_id
int64
0
34
en-zh
So, in our proposed approach we want to solve those problems in a step by step and interpretable manners.
在我们提出的方法中, 我们希望以逐步和可解释的方式解决这些问题.
在我们提出的方法中,我们希望一步一步地以可解释的方式解决这些问题。
93
1
ACL
2,023
multilingual
2
en-zh
At this time, it is necessary to detect the four labels that have been set up in advance.
现在,我们必须侦测到 那四个被陷害的强奸犯
现在,有必要检测事先设置好的四个标签。
60
1
ACL
2,023
multilingual
21
en-zh
First, it deals with two types of language.
首先,它涉及两个类型的语言,
首先,它涉及两种类型的语言。
80
1
ACL
2,023
multilingual
22
en-zh
So here we keep generating the operators until we reach the leaves, which are the quantities.
在这里,我们继续生成运算器,直到我们到达这些数量的叶子。在
所以在这里,我们继续生成运算符,直到我们到达叶子,也就是数量。
53
1
ACL
2,023
multilingual
22
en-zh
So this is the end of the talk, and questions are welcomed. Thank you.
的结束,所有问题都来了,谢谢你
我的演讲到此结束,欢迎各位提出问题。谢谢。
59
1
ACL
2,023
multilingual
22
en-zh
We hope that our work sparks interest in developing practical and reliable statutory article retrieval models.
我们希望所有工作都能激发人们对开发实用可靠的法定条文检索模型的兴趣,
我们希望我们的工作能够激发在开发实用可靠的法定条款检索模型方面的兴趣。
63
1
ACL
2,023
multilingual
21
en-zh
So in our model implementations, we first use a pretrained language model which can be BERTs or Robertas and then we encode the sentence and then we obtain these quantity representations.
在我们的模型实现中, 我们首先使用一个预先训练的语言模型, 可以是鸟类或机器人, 然后我们编码一个句子, 然后我们得到这些数量表示。
在我们的模型实现中,我们首先使用预训练语言模型,它可以是BERTs或Robertas,然后我们编码句子,然后我们得到这些数量陈述。
78
1
ACL
2,023
multilingual
21
en-zh
When we look on the small ah dataset, we can see that the performance of MTDNN decreases and the improvement of the prelim, the preliminary multitask finetuning phase decreases to one point five percent.
我们可以看到MTDNN的性能下降,而初步多任务微调阶段的改进降至1.5%.
当我们在小数据集上看时,我们可以看到MTDNN的表现下降了,初步的多任务微调阶段的改进下降到1.5%。
66
1
ACL
2,023
multilingual
2
en-zh
The state-of-the-art in multitask ah multitask finetuning called MTDNN.
然后采取数据地球的状态在多任务多任务的细节调节称为空无空完成
最先进的多任务多任务微调技术称为MTDNN。
10
1
ACL
2,023
multilingual
22
en-zh
And around eighty percent of these cited articles come from either the civil code, judicial codes, criminal investigation codes or penal codes.
其中约80%的条款来自民法,司法法,刑事调查法或刑法.
这些被引用的条款中约有80%来自民法典、司法法典、刑事调查法典或刑法典。
60
1
ACL
2,023
multilingual
2
en-zh
Alright. And after these three steps, we can actually reuse the results from the second step, and then get the ah results of the fourth step, and then finally we can obtain the dividends.
经过这三个步骤,我们实际上可以重复使用第二步的结果,然后获得第一步的结果,然后最后我们可以获得分割。
好的。经过这三个步骤,我们实际上可以重复使用第二个步骤的结果,然后得到第四个步骤的结果,最后可以得到被除数。
77
1
ACL
2,023
multilingual
22
en-zh
And we observe these two expressions actually have similar scores.
我们观察到这两个表情 实际上有相似的分数。因此,
我们观察到这两个表达式实际上有相似的分数。
80
1
ACL
2,023
multilingual
21
en-zh
So, overall our reasoner is able to significantl significantly outperform this tree based model.
总的来说,我们的理智能够从这个时间模型中看到一个显著的输出,
总的来说,我们的推理能够显著优于这个基于树的模型。
67
1
ACL
2,023
multilingual
22
en-zh
Common natural language for the questions and complex legal language for the statutes.
共同的自然语言问题和复杂的法律语言的
针对问题本身的通用自然语言,以及针对法规使用的复杂法律语言。
78
1
ACL
2,023
multilingual
22
en-zh
So, these slides actually visualize the evolution of the state where we keep adding expression to the current state.
这个谎言实际上可视化了我们不断向当前状态添加表达的状态的演变。
这些幻灯片实际上可视化了我们不断向当前状态添加表达式的状态的演变。
81
1
ACL
2,023
multilingual
22
en-zh
So this is the end of the talk, and questions are welcomed. Thank you.
演讲和提问到此结束。我欢迎。谢谢。
我的演讲到此结束,欢迎各位提出问题。谢谢。
58
1
ACL
2,023
multilingual
21
en-zh
But we can see the absolute numbers on MathQA or SVAMP are not really high.
但我们可以看到数学上的绝对木材,或者我们不是真的高。
但我们可以看到,MathQA或SVAMP上的绝对数字并不高。
20
1
ACL
2,023
multilingual
22
en-zh
After the feature extraction phase, there is a feature generation phase when we use the extracted features to generate a small number of new features.
在特征提取阶段之后。有一个特征生成阶段, 我们使用提取的特征 生成一小部分新特征。
在特征提取阶段之后,还有一个特征生成阶段,我们使用提取的特征来生成少量的新特征。
90
1
ACL
2,023
multilingual
21
en-zh
And the second thing is that, as mentioned, because the probability distribution is unbalanced between different time steps, so it's also pretty challenging to apply beam search strategy.
第二件事是提到的,因为财产分布在不同的时间步骤之间是平衡的,所以还相当具有挑战性来应用雷射搜索雷射搜索策略。这是谈话
第二件事是,如前所述,由于不同时间步骤之间的概率分布不平衡,因此应用波束搜索策略也非常具有挑战性。
71
1
ACL
2,023
multilingual
22
en-zh
Note that for training, we experiment with the two flavors of the bi-encoder architecture.
请注意,为了训练, 我们尝试了两年一度建筑的两种口味。
请注意,针对训练,我们尝试使用了双编码器架构的两种风格。
63
1
ACL
2,023
multilingual
21
en-zh
For those samples without unused quantities, so the overall performance is actually higher than the, the performance is actually higher than the overall performance.
对于那些没有未使用量的样本,所以整体性能实际上高于性能实际上
对于那些没有未使用数量的样本,其整体表现实际上高于整体表现。
81
1
ACL
2,023
multilingual
2
en-zh
And we also think math word problem is a straightforward application to evaluate such reasoning abilities.
我们还认为形象问题是评估这种推理能力的简单应用.
我们还认为,数学文字问题是一个用来评估这种推理能力的简明应用。
51
1
ACL
2,023
multilingual
2
en-zh
Finally, a conclusion.
终于得出结论了
最后,一个结论。
89
1
ACL
2,023
multilingual
21
en-zh
These results indicate that CEAS and CAS are significantly affected.
这些结果表明C-E-A-S和C-E-A-S具有显著的有效性.
这些结果表明CEAS和CAS受到严重影响。
43
1
ACL
2,023
multilingual
2
en-zh
Each feature represents the likelihood for each class.
表了每个类的概率.
每个特征表示每个类的可能性。
75
1
ACL
2,023
multilingual
2
en-zh
First generate ah features in the number of classes of the original dataset.
FeSTE生成了原始数据集的类数的特征.
首先,在原始数据集的类别数量中生成特征。
56
1
ACL
2,023
multilingual
2
en-zh
And as we said, we are interested in assessing what capabilities the vision and language models have after pretraining.
正如我们所说的,我们有兴趣评估视觉和语言模型在预训练后具有哪些能力
正如我们所说,我们有兴趣评估视觉和语言模型在预训练后具有哪些能力。
80
1
ACL
2,023
multilingual
22
en-zh
These results indicate that CEAS and CAS are significantly affected.
这些结果表明,T、S和C、S显著有效。G’
这些结果表明CEAS和CAS受到严重影响。
59
1
ACL
2,023
multilingual
21
en-zh
Release notes are labeled as improvements or fixes, etc.
礁石注释被标记为改进,但面部等等。
发行说明被标记为优化或修复等。
41
1
ACL
2,023
multilingual
21
en-zh
For example, though it's not impossible, it is statistically less likely for plants to cut a man than a man to cut plants, and large vision and language models could pick up on this.
例如,虽然这不是不可能的,但在统计学上,植物削减人类的可能性比人类削减植物的可能性更小,大视力和语言模型可能会从中获取。
例如,虽然这不是不可能的,但从统计学上来说,植物砍人的可能性比人砍植物的可能性要小,较大的视觉和语言模型可以发现这一点。
95
1
ACL
2,023
multilingual
22
en-zh
This extra information is not used in the present work, but might be of interest for future research on legal information retrieval or legal text classification.
这些额外的信息在本研究中没有使用,但可能对未来的法律信息检索或法律文本分类研究有兴趣.
这些额外的信息未在当前工作中使用,但可能对今后关于法律信息检索或法律文本分类的研究有所帮助。
90
1
ACL
2,023
multilingual
2
en-zh
Ah, if we have a large number of operators or constants, the memory consumption could be pretty high.
如果我们有大量的运算符或常数, 记忆消耗可能相当高.
如果我们有大量 数量的运算符或常量,内存消耗可能相当高。
90
1
ACL
2,023
multilingual
2
en-zh
Lastly, we passed the legal references and filtered out the questions whose references were not articles in one of the codes of law we considered.
最后,我们通过了法律参考文献, 并过滤了那些参考文献 不是法典中文章的问题。我们考虑了。
最后,我们浏览了法律参考文献,并过滤掉参考文献不包含在我们所考虑的其中任何一部法律条款的问题。
79
1
ACL
2,023
multilingual
21
en-zh
Our experiments show that vision and language models identify named objects and their presence in images well, as shown by the existence piece, but struggle to ground their interdependence and relationships in visual scenes when forced to respect linguistic indicators.
我们的实验表明视觉和语言模型很好地识别出存在于图像中的命名物体, 正如存在部分所示, 但当被迫尊重语言指标时, 很难在视觉场景中建立它们的相互依赖和关系.
我们的实验表明,视觉和语言模型能很好地识别命名对象及其在图片中的存在(如“存在性”部分所示),但在被迫尊重语言指标时,却很难在视觉场景中建立它们的相互依存性和关系。
91
1
ACL
2,023
multilingual
2
en-zh
To analyze the text, we use the current state-of-the-art of text analysis, which are transformer based language models as BERT, GPT, XLNet and etc.
为了分析文本,我们使用当前最先进的文本分析,
为了分析文本,我们使用了目前最先进的文本分析方法,即基于转换器的语言模型,如BERT、GPT、XLNet等。
50
1
ACL
2,023
multilingual
2
en-zh
Here are the results.
结果。
结果如下。
95
1
ACL
2,023
multilingual
21
en-zh
This information retrieval task comes with its own set of challenges.
这种信息检索任务也带来了其自身的挑战.
这种信息检索 任务有其自身的一系列挑战。
87
1
ACL
2,023
multilingual
2
en-zh
Furthermore, CAS methods cannot generate accurate release notes without additional information.
此外,C-E-A-S方法不能在没有附加信息的情况下生成准确的释放说明.
此外,如果没有额外的信息,CAS方法就无法生成准确的发行说明。
63
1
ACL
2,023
multilingual
2
en-zh
When you finetune the language model over the n minus one datasets.
当你在数据集中发现语言中,此外,在数据集中anda,
当你在n减1数据集上微调语言模型时。
56
1
ACL
2,023
multilingual
22
en-zh
When its goal is to classify universities into low ranking universities and high-ranking universities.
当它的目标是将大学分为低排名大学和高排名大学时。作为知识基础,
当它的目标是将大学分为低排名的大学和高排名的大学时。
69
1
ACL
2,023
multilingual
21
en-zh
Hello, my name is Kamezawa from the University of Tokyo.
大家好,我是东京大学的卡米泽弗
大家好,我是东京大学的Kamezawa。
100
1
ACL
2,023
multilingual
21
en-zh
Hello, my name is Kamezawa from the University of Tokyo.
的名字是来自土耳其大学的警长,提出了
大家好,我是东京大学的Kamezawa。
17
1
ACL
2,023
multilingual
22
en-zh
Before diving into the main contribution of this work, let's first describe the problem of statutory article retrieval.
在深入探讨这项工作的主要贡献之前, 让我们先描述一下法定文章检索的问题.
在深入探讨这项工作的主要贡献之前,让我们首先描述法定条款检索的问题。
72
1
ACL
2,023
multilingual
2
en-zh
So we further investigate the results on SVAMP.
我们进一步调查了SWAM的结果.
我们进一步研究SVAMP的结果。
62
1
ACL
2,023
multilingual
2
en-zh
High coverage of CEAS can be achieved probably because the classifier can focus on selecting relevant commit messages for each class.
的高覆盖率可以实现,可能是因为崩溃失败可以专注于每个草地的分散事件。
CEAS的高覆盖率之所以能够实现,可能是因为分类器可以专注于为每个类别选择相关的提交信息。
16
1
ACL
2,023
multilingual
22
en-zh
And as we said, we are interested in assessing what capabilities the vision and language models have after pretraining.
正如我们所说,我们有兴趣评估 视觉和语言模型在预训练后 有什么能力。
正如我们所说,我们有兴趣评估视觉和语言模型在预训练后具有哪些能力。
92
1
ACL
2,023
multilingual
21
en-zh
Overall, the finetuned bi-encoder significantly outperforms all the other baselines.
的 Beyonc 线显著优于所有其他基线。
总体而言,微调的双编码器明显优于所有其他基线。
68
1
ACL
2,023
multilingual
21
en-zh
The classwise extractive then abstractive summarization model consists of two neural modules.
填字游戏,提取和抽象。总结模块由两个神经模块组成:
这个先按类别抽取然后抽象总结的模型是由两个神经模块组成的。
59
1
ACL
2,023
multilingual
21
en-zh
And I will present our paper, Few-Shot Tabular Data Enrichment Using Fine-Tuned Transformers Architectures.
我会介绍我们的论文几个短页是使用Intune转换器架构的丰富
我将介绍我们的论文《使用微调转换器架构的少样本表格式数据充实》。
70
1
ACL
2,023
multilingual
22
en-zh
So, in our prediction we find some of the intermediate values are actually negatives.
在我们的预测中,我们发现一些中间值实际上是负值.
在我们的预测中,我们发现一些中间值实际上是负数。
98
1
ACL
2,023
multilingual
2
en-zh
Glyph's training data is fairly small, about five thousand, and will be shown in the experiments described below.
等效转换数据相当小 — — 大约5000个 — — 将在下面描述的实验中展示。
Glyph的训练数据相当小,约为五千,并将在下面描述的实验中显示。
47
1
ACL
2,023
multilingual
21
en-zh
Where we perform operator from q_i to q_j, and such expression is actually directed.
我们执行从qi到qj的运算符, 这种表达式实际上是指向的.所以
我们执行从q_i到q_j的运算符,这样的表达式实际上是定向的。
91
1
ACL
2,023
multilingual
2
en-zh
I will be presenting my joint work with Jerry which is about a New Dataset for Statutory Article Retrieval.
我将介绍我与杰里的共同工作, 这是一个新的数据集
我将展示我与Jerry的合作结果,它是针对法定条款检索的新数据集。
63
1
ACL
2,023
multilingual
2
en-zh
It ah it extracts the text from the knowledge base, which in this example is the abstract of the Wikipedia page.
从知识库中提取文本, 在这个例子中,是维基百科页面的摘要。
它从知识库中提取文本,在本例子中,它是维基百科页面的摘要。
81
1
ACL
2,023
multilingual
21
en-zh
We created a heuristic matching rule to get the previous and next versions.
我们创建了一个人道主义的匹配DOO 来得到前和下一个版本。
我们创建了一个启发式匹配规则来获取上一个和下一个版本。
61
1
ACL
2,023
multilingual
21
en-zh
With the lexical methods above, the Siamese bi-encoders evaluated in a zero shot setup in the middle, and the finetuned bi-encoders below.
在中间的零 statset中评估的西雅姆 bon 方位上方的语法方法和下方的精细调制 bann
上面是词汇方法,中间是在零样本设定中评估的siamese双编码器,下面是微调的双编码器。
0
1
ACL
2,023
multilingual
22
en-zh
Which can be explained by the fact that those codes focused less on individuals and their concerns.
这可以由这些守则不太关注个人和他们的关注点来解释.
这可以解释为,这些法典较少关注个人及个人关心的问题。
70
1
ACL
2,023
multilingual
2
en-zh
So the nice thing here is that we can easily add constraints to control this search this search space.
我们可以很容易地添加限制来控制这个搜索空间.
这里的好处是我们可以轻松地添加约束条件,来控制这个搜索这个搜索空间。
70
1
ACL
2,023
multilingual
2
en-zh
The classwise extractive then abstractive summarization model consists of two neural modules.
分类智能提取比抽象总结模型由两个神经模块组成,
这个先按类别抽取然后抽象总结的模型是由两个神经模块组成的。
64
1
ACL
2,023
multilingual
2
en-zh
Here are the results.
这里是结果
结果如下。
100
1
ACL
2,023
multilingual
22
en-zh
Foiling basically means that we take the caption of an image and produce a foil by altering the caption such that it does not describe the image anymore.
ing基本上意味着我们采取图像的标题,并通过改变标题,以便它不再描述图像产生一个标题。
干扰的意思是,我们获取一个图像的标题,通过改变标题,使其不再描述图像中的物体,从而造成干扰。
81
1
ACL
2,023
multilingual
22
en-zh
A classifier using BERT or CodeBERT and a generator using BART.
一个使用BERT或codeBERT的分类器和一个使用BERT的生成器.
一个是使用BERT或CodeBERT的分类器,另一个是使用BART的生成器。
93
1
ACL
2,023
multilingual
2
en-zh
We experiment with mean, max and CLS pooling as well as product and cosine for computing similarities.
我们试验了平均MAX和CASS聚合以及狗产品和COS标志来计算
我们试验了均值、最大值和CLS集合,以及计算相似性的乘积和余弦。
56
1
ACL
2,023
multilingual
22
en-zh
So we got a message, the accuracies on these tasks and specific benchmarks are increasing steadily.
因此,我们得到了一个消息。这些特定任务基准的准确性正在稳步提高。
因此,我们得到了一个信息,这些任务的准确性和特定基准正在稳步提升。
94
1
ACL
2,023
multilingual
21
en-zh
It uses one architecture for all tasks and datasets.
它使用一个架构来处理所有任务数据集.
它使用一个架构用于所有任务和数据集。
75
1
ACL
2,023
multilingual
2
en-zh
Well, during the last years, we have seen an explosion of transformer based vision and language models pretrained on large amounts of image text pairs.
在过去的几年里, 我们看到了基于转换器的视觉和语言模型的爆炸性增长,
那是因为,在过去的几年里,我们看到了基于转换器的视觉和语言模型在大量的图像文本对上进行预训练的爆炸性增长。
77
1
ACL
2,023
multilingual
2
en-zh
So, we need to ah feature extraction phase ah which includes text analysis.
我们需要一个包含文字分析
因此,我们需要进行特征提取阶段,其中包括文本分析。
95
1
ACL
2,023
multilingual
2
en-zh
First, we started by compiling a large corpus of legal articles.
首先,我们从收集大量法律文章开始。
首先,我们从汇编大量语料库的法律条款开始。
90
1
ACL
2,023
multilingual
21
en-zh
First, we started by compiling a large corpus of legal articles.
我们开始通过编写一个大批忠实的文章,
首先,我们从汇编大量语料库的法律条款开始。
84
1
ACL
2,023
multilingual
22
en-zh
Ah, if we have a large number of operators or constants, the memory consumption could be pretty high.
如果我们有较大的操作员数量或含有凝固剂,则内存消耗量可能相当高。
如果我们有大量 数量的运算符或常量,内存消耗可能相当高。
55
1
ACL
2,023
multilingual
22
en-zh
But our performance increased to eleven percent compared to the target task finetuning alone.
但我们的成绩增长了11%, 相比目标任务微调.
但与单独的目标任务微调相比,我们的表现提高到了11%。
89
1
ACL
2,023
multilingual
2
en-zh
Our framework FeSTE is exactly this automatic process.
我们的框架首先就是这个自动化过程。讓
我们的框架FeSTE正是这个自动过程。
75
1
ACL
2,023
multilingual
22
en-zh
This is quite similar to relation extraction.
这与关系抽取非常相似。
这与关系提取 非常相似。
100
1
ACL
2,023
multilingual
22
en-zh
So finally we can obtain this final expression q_3 times q_4.
我们可以得到最后的表达式Q3乘以Q4.
我们最终可以得到最后这个表达式q_3乘以q_4。
78
1
ACL
2,023
multilingual
2
en-zh
The remaining references were matched and converted to the corresponding article ids from our corpus.
其余参考资料被匹配并转换为相应的文章。《Orcopus》的灵感
其余的参考文献被匹配,并转换为我们语料库中相应的条款ID。
65
1
ACL
2,023
multilingual
21
en-zh
Our experiments show that the proposed method generates less noisy release notes at higher coverage than the baselines.
我们的实验表明,所提出的质量产生较少噪声的珊瑚礁,而
我们的实验表明,与基线相比,所提出的方法在更高的覆盖率下产生了更少干扰的发行说明。
27
1
ACL
2,023
multilingual
21
en-zh
Second, we use natural language inference or short NLI to filter out foils that could be still describing the image, since when constructing foils we need to ensure that they fail to describe the image.
第二,我们使用自然语言推断或短肛门来过滤可能仍然描述图像的叶子,因为在构建叶子时,我们需要确保未能描述图像。
其次,我们使用自然语言推断或简短NLI来过滤可能仍在描述图像的干扰词,因为在构建干扰词时,我们需要确保它们无法描述图像。
60
1
ACL
2,023
multilingual
22
en-zh
So this figure is taken from the PaLM paper where they perform prompting to solve the network problem in the few shot learning scenario.
这张图是从Pound的论文中提取的, 在这里他们执行提示以解决几个镜头学习场景中的元问题.
这个数字取自PaLM的论文,他们在这个论文中进行了提示,以解决少样本学习情况下的网络问题。
42
1
ACL
2,023
multilingual
2
en-zh
The image shows a release note for version two point six point four of the vuejs library.
图像显示了Vue.js库的2.6.4版本的RnSAM
这个图像显示的是vuejs库的2.6.4版本的发行说明。
40
1
ACL
2,023
multilingual
2
en-zh
The second is Glyph, recently announced in twenty twenty.
第二個是嚴重的這篇文章,它可以被觀察到
第二个是Glyph ,最近在2020年宣布。
3
1
ACL
2,023
multilingual
22
en-zh
First, we started by compiling a large corpus of legal articles.
首先,我们编制了大量的法律文章.
首先,我们从汇编大量语料库的法律条款开始。
88
1
ACL
2,023
multilingual
2
en-zh
I'm Allan from ByteDance AI Lab, and this is a joint work with Jierui Li from the University of Texas at Austin and Wei Lu from SUTD.
el 從 bios 是,這是從稅務 astin 大學和我們從突然抽取的工作。
我是ByteDance 人工智能实验室的Allan,以下是我与德克萨斯大学奥斯汀分校的Jierui Li和SUTD的Wei Lu的合作成果。
0
1
ACL
2,023
multilingual
22
en-zh
So for example, here in the second step, ah we can obtain these divisors which is twenty seven.
例子,在第二步中, 我们可以得到这个设备,27个,
例如,在第二步中,我们可以得到这些除数,即27。
76
1
ACL
2,023
multilingual
21
en-zh
And the last thing is that the underlying mechanism does not only apply to network problem solving tasks but also other tasks that involve multi step reasoning.
最后一点是, 底层机制不仅适用于记忆问题解决任务,但
最后一点是,底层机制不仅适用于解决任务的网络问题,也适用于涉及多步骤推理的其他任务。
40
1
ACL
2,023
multilingual
2
en-zh
Each of them has been labeled by experienced jurists with references to relevant articles from a corpus of more than twenty-two thousand six hundred legal articles from Belgian codes of law.
经验丰富的法学家用比利时法典中超过22,600个法律条款的相关条款进行了标记.
每个问题都由经验丰富的法学家标记,并参考了比利时法典中超过22,600篇法律条款的语料库中的相关条款。
86
1
ACL
2,023
multilingual
2
en-zh
Then it reformulated the task into a pairwise sentence classification task.
然后它将任务重新定义为一个 pille senden piac分类任务。
然后,它将任务重新表述为一个成对的句子分类任务。
43
1
ACL
2,023
multilingual
22
en-zh
We experiment with context independent text encoder, namely word2vec and fastText, and context dependent embedding models, namely Roberta and more specifically CamemBERT which is a French Roberta model.
情况下被直接应用.我们试验了文本独立编码器,即Word2Vec和FastText,以及文本依赖嵌入模型,即Robota,更具体地说是Camembert,这是法国Robota的模型.
我们尝试了与上下文无关的文本编码器,即word2vec和fastText,以及上下文相关的嵌入模型,即Roberta,更具体地说是CamemBERT,它是个法语Roberta模型。
69
1
ACL
2,023
multilingual
2
en-zh
Such a system could provide a free professional legal help service for unskilled humans.
这种系统可以为无技能的人提供免费的专业法律援助服务。
这样一个系统可以为非技术人员提供免费的专业法律援助服务。
70
1
ACL
2,023
multilingual
22
en-zh
For example, for BERT, we improve seven points and then for the Roberta base model we actually improved two points.
例如,对于鸟类,我们提高了7分.对于基于罗伯塔的模型, 我们实际上提高了2个点
例如,对于BERT,我们提高了7分。然后,对于Roberta基础模型,我们实际上提高了两分。
88
1
ACL
2,023
multilingual
2
en-zh
We have set up a task that takes the commit messages as input and outputs a labeled release notes.
我们设置了一个任务, 将提交消息作为输入并输出标记的发布说明.
我们设置了一个任务,将提交信息作为输入,并输出一个标签化的发行说明。
93
1
ACL
2,023
multilingual
2
en-zh
So there are many tasks.
很多任务清
存在很多的任务。
68
1
ACL
2,023
multilingual
22
en-zh
We modeled the classwise abstractive summarization approach by two different methods.
我们通过两种不同的方法建模草本阻塞性总结方法。
我们通过两种不同的方法对分类抽象总结方法进行建模。
60
1
ACL
2,023
multilingual
21
en-zh
Why did we do the trouble in setting up this benchmark?
为什么在过去几年中,我们在设置这个基准方面
我们为什么要费尽心思设立这个基准呢?
86
1
ACL
2,023
multilingual
22
en-zh
We assume the precision of quantities are known.
我们假设数量的精度是已知的,
我们假设数量的精确度是已知的。
100
1
ACL
2,023
multilingual
2
en-zh
What is it that a vision and language transformer understood when assigning a high score for this image and this sentence to match?
视觉和语言转换器在为此图像和这个句子分配高分和这个句子分配低分时理
视觉和语言转换器在为这个图像和这个句子分配高分时,所理解的是什么呢?
84
1
ACL
2,023
multilingual
22
en-zh
Siamese, which uses a unique word embedding model that maps the query and article together in a shared dense vector space, and two-tower, which uses two independent word embedding models that encode the query and article separately into different embedding spaces.
Siamese使用一个独特的词嵌入模型,将查询和文章映射到一个共享的密集向量空间中,Tutawa使用两个独立的词嵌入模型,将查询和文章单独编码到不同的嵌入空间.
第一种是siamese,它使用一个独特的单词嵌入模型,将查询和条款一起映射到一个共享的密集向量空间;另一种是双塔,它使用两个独立的单词嵌入模型,将查询和条款分别编码到不同的嵌入空间。
60
1
ACL
2,023
multilingual
2
en-zh
These were set based on previous research and other factors.
这些都是基于此前的研究和其他因素。
这些都是基于之前的研究和其他因素设定的。
63
1
ACL
2,023
multilingual
21
en-zh
So, in this example there are four tasks in the training set, so MTDNN maintain four heads as you can see at the image.
在这个例子中, 训练集中有四个任务。因此,空的DNA和保持四个头, 正如你所看到的图片。
因此,在这个例子中,训练集里有四个任务,所以MTDNN维持四个头,正如在图像上看到的那样。
92
1
ACL
2,023
multilingual
21
en-zh
Then it reformulated the task into a pairwise sentence classification task.
然后它把任务重新表述成句子 每个分类任务,
然后,它将任务重新表述为一个成对的句子分类任务。
85
1
ACL
2,023
multilingual
21
en-zh
And in fact we do not use beam search, in contrast all previous approaches are using beam search.
事实上,我们不使用光束搜索.相比之下,显而易见的方法是使用光束搜索.很好.
事实上,我们不使用波束搜索。相反,所有之前的方法都使用波束搜索。
72
1
ACL
2,023
multilingual
2