Transitivity of transformation matrices to bridge word vector spaces over 1000 years
Transitivity of transformation matrices to bridge word vector spaces over 1000 years
复制标题
DOI：
10.1007/s11227-020-03584-5
复制
发表时间：
2021-02
期刊：
影响因子：
--
通讯作者：
Katsurou Takahashi;Hiroaki Ohshima
中科院分区：
文献类型：
--
作者：
Katsurou Takahashi;Hiroaki Ohshima
我们提出了一种同义词搜索方法来解决随时间推移的问题，通过一个已知域中的实例查询未知域中的信息。“21世纪初的布什”与“80年代的里根”似乎是一种天然的联系，因为“布什”和“里根”是这几十年的美国总统。其抽象形式为“AinB”，类似于“CinD”。我们通过在Skip-gram模型中使用随时间变化的变换矩阵词向量来解决这个问题。这个问题的例子如下所示。例如，给出的句子“21世纪初的布什”与“80年代的新”相似。我们搜索句子中的适当实体。因此，我们重点研究变换矩阵之间的传递性。我们的方法是通过得到单词嵌入模型之间的转换矩阵，将“B”的单词嵌入模型中的“A”的矢量表示转换为“D”的单词嵌入模型中的“X”的矢量表示。讨论了前人工作的参数，并利用共现聚类法对转换矩阵的选词进行了改进。我们的目标是寻找分离超过1000年的同义词。然而，在“2000年代”和“1000年代”之间，有几个常见的稳定义词。因此，在这种情况下，由于常用词聚类超过10万个，所以很难使用共现聚类法。这就是为什么我们使用传递关系(2000s，X)(1500s，Y)，(1500s，Y)(1000s，Z)(2000s，X)(1000s，Z)(在本文中，传递性是)来解决分离了1000年的问题。我们进行了实验作为演示来解决问题，并对nDCG和MRR进行了评估。
We proposed a synonym search method to solveproblem over time with a query by an example in a known domain for information in an unknown domain. It seems a natural relation that “Bushin the2000s” is similar to “Reaganin the1980s” because “Bush” and “Reagan” are the president of theUSAin these decades. The abstraction is “AinB” which is similar to “CinD.” We solve theproblem over time by using transformation matrix word vectors over time in the Skip-gram model. The example of theproblem is as below. For instance, the sentence “Bushin the2000s” is similar to “Xin the1980s” is given. We search for the appropriate entity toXof the sentence. Therefore, we focus on the transitivity between the transformation matrix. Our approach is to convert the vector representation of “A” in the model of the word embedding model of the “B” to the vector representation of “X” in the model of the word embedding model of the “D” by getting the transformation matrix between word embedding models. We discuss the parameters of previous work and improve choosing words to make transformation matrix using co-occurrence cluster. Our aim is to search for synonyms in which there are more than the 1000 years of separation. However, there are a few common stable meaning words between the “2000s” and the “1000s.” Therefore, in the situation, there is difficulty to use co-occurrence cluster because the clusters of common words are more than 100,000. That is why, we use the transitive relation (2000s,X)(1500s,Y) , (1500s,Y)(1000s,Z)(2000s,X)(1000s,Z) (in the abstruction, the transitivity is) to solve theproblem over the 1000 years of separation. We had experiments as the demonstration to solve theproblem and evaluate nDCG and MRR.