Efficient Convolution Kernels for Dependency and Constituent Syntactic Trees

Abstract

Abstract. In this paper, we provide a study on the use of tree kernels to encode syntactic parsing information in natural language learning. In particular, we propose a new convolution kernel, namely the Partial Tree (PT) kernel, to fully exploit dependency trees. We also propose an efficient algorithm for its computation which is futhermore sped-up by applying the selection of tree nodes with non-null kernel. The experiments with Support Vector Machines on the task of semantic role labeling and question classification show that (a) the kernel running time is linear on the average case and (b) the PT kernel improves on the other tree kernels when applied to the appropriate parsing paradigm.

Semantic Sentiment Analysis of Twitter

“…Apart from simply combining various features, they also designed a tree representation of tweets to combine many categories of features in one succinct representation. A partial tree kernel [8] was used to calculate the similarity between two trees. They found that the most important features are those that combine prior polarity of words with their POS tags.…”

“…Rizzo and Troncy [10] evaluated the use of five popular entity extraction tools on a dataset of news articles, including AlchemyAPI, DBPedia Spotlight, 7 Extractiv, 8 OpenCalais and Zemanta. Their experimental results showed that AlchemyAPI performs best for entity extraction and semantic concept mapping.…”

Detect Rumors in Microblog Posts Using Propagation Structure via Kernel Learning

“…Kernel methods are designed to evaluate the similarity between two objects, and tree kernel specifically addresses structured data which has been successfully applied for modeling syntactic information in many natural language tasks such as syntactic parsing (Collins and Duffy, 2001), question-answering (Moschitti, 2006), semantic analysis (Moschitti, 2004), relation extraction (Zhang et al, 2008) and machine translation (Sun et al, 2010). These kernels are not suitable for modeling the social media propagation structures because the nodes are not given as discrete values like part-of-speech tags, but are represented as high dimensional real-valued vectors.…”

“…Then, we propose a kernel-based data-driven method called Propagation Tree Kernel (PTK) to generate relevant features (i.e., subtrees) automatically for estimating the similarity between two propagation trees. Unlike traditional tree kernel (Moschitti, 2006;Zhang et al, 2008) for modeling syntactic structure based on parse tree, our propagation tree consists of nodes corresponding to microblog posts, each represented as a continuous vector, and edges representing the direction of propagation and providing the context to individual posts. The basic idea is to find and capture the salient substructures in the propagation trees indicative of rumors.…”

Learning Methods in Multi-grained Query Answering

“…". Clustering based on Syntactic Tree Kernels [11] is a possible method to use the syntactic features of the queries.…”