SFTM: Fast Comparison of Web Documents using Similarity-based Flexible Tree Matching

TL;DR

A novel Similarity-based Flexible Tree Matching algorithm (SFTM) is proposed, which is the first algorithm to enable tree matching on real-life web documents with practical computation times and leverages node labels and local topology similarity in order to avoid any combinatorial explosion.

Abstract

Tree matching techniques have been investigated in many fields, including web data mining and extraction, as a key component to analyze the content of web documents, existing tree matching approaches, like Tree-Edit Distance (TED) or Flexible Tree Matching (FTM), fail to scale beyond a few hundreds of nodes, which is far below the average complexity of existing web online documents and applications. In this paper, we therefore propose a novel Similarity-based Flexible Tree Matching algorithm (SFTM), which is the first algorithm to enable tree matching on real-life web documents with practical computation times. In particular, we approach tree matching as an optimisation problem and we leverage node labels and local topology similarity in order to avoid any combinatorial explosion. Our practical evaluation demonstrates that our approach compares to the reference implementation of TED qualitatively, while improving the computation times by two orders of magnitude.

Figures (9)

• Figure 1: Example of a tree matching. Crossed circles are auxiliary no-match nodes enabling insertions and removals between trees.
• Figure 2: From the input trees depicted in Figure \ref{['fig:tree_matching_example']}, we build a bipartite graph $G$ representing the set of all possible matching (left) and then compute the optimal full matching (right).
• Figure 3: Steps to compute a full matching between two tree $T_1$ and $T_2$. In the top, we describe FTM and in the bottom, our algorithm: SFTM
• Figure 4: Creating the bipartite graph $G$ from two example DOMs. (1a,b) are the input DOMs, (2a,b) the extracted tokens, (3) the inverted index $T_{map}$ , (4) the neighbours dictionaries and (5) the bipartite graph $G$. For simplicity, the figure shows a matching where IDF(t) = 1 , $p = 0$ and no-match nodes are not displayed.
• Figure 5: The distribution of DOM sizes (in terms of nodes) for the complete (left, blue) and the partial (right, green) datasets.