论文A Tree-Structured Decoder for Image-to-Markup Generation中gtd文件详解

\int	1	<s>	0	Start
3	2	\int	1	Below
6	3	\int	1	Above
\int	4	\int	1	Right
2	5	\int	4	Below
4	6	\int	4	Above
2	7	\int	4	Right
d	8	2	7	Right
x	9	d	8	Right
d	10	x	9	Right
y	11	d	10	Right
=	12	y	11	Right
2	13	=	12	Right
\cdot	14	2	13	Right
(	15	\cdot	14	Right
6	16	(	15	Right
-	17	6	16	Right
3	18	-	17	Right
)	19	3	18	Right
\cdot	20	)	19	Right
(	21	\cdot	20	Right
4	22	(	21	Right
-	23	4	22	Right
2	24	-	23	Right
)	25	2	24	Right
=	26	)	25	Right
1	27	=	26	Right
2	28	1	27	Right
</s>	29	2	28	End

每一行代表公式树中一个节点,最左一列表示公式树中一个节点符号,如\int表示积分;左起第二列表示树中每一节点符号的位置编码,缺少位置编码会导致父节点歧义;中间一列表示最左侧节点符号的父节点;右起第二列表示父节点编号;最右节点表示每一行中最左边子节点与中间父节点的关系。
二、树形结构表示(画得有点丑)
论文A Tree-Structured Decoder for Image-to-Markup Generation中gtd文件详解_第1张图片

你可能感兴趣的:(算法,计算机视觉,人工智能,深度学习)