Available online at www.sciencedirect.com Computer Speech and Language 27 (2013) 1085–1104 Bridge the gap between statistical and hand-crafted grammars  Ali Basirat, Heshaam Faili ∗ Laboratory of Natural Language and Text Processing, School of Electrical & Computer Engineering, College of Engineering, University of Tehran, Tehran, Iran Received 4 November 2011; received in revised form 19 September 2012; accepted 5 February 2013 Available online 27 February 2013 Abstract LTAG is a rich formalism for performing NLP tasks such as semantic interpretation, parsing, machine translation and information retrieval. Depend on the specific NLP task, different kinds of LTAGs for a language may be developed. Each of these LTAGs is enriched with some specific features such as semantic representation and statistical information that make them suitable to be used in that task. The distribution of these capabilities among the LTAGs makes it difficult to get the benefit from all of them in NLP applications. This paper discusses a statistical model to bridge between two kinds LTAGs for a natural language in order to benefit from the capabilities of both kinds. To do so, an HMM was trained that links an elementary tree sequence of a source LTAG onto an elementary tree sequence of a target LTAG. Training was performed by using the standard HMM training algorithm called Baum–Welch. To lead the training algorithm to a better solution, the initial state of the HMM was also trained by a novel EM-based semi-supervised bootstrapping algorithm. The model was tested on two English LTAGs, XTAG (XTAG-Group, 2001) and MICA’s grammar (Bangalore et al., 2009) as the target and source LTAGs, respectively. The empirical results confirm that the model can provide a satisfactory way for linking these LTAGs to share their capabilities together. © 2013 Elsevier Ltd. All rights reserved. Keywords: Tree adjoining grammar; LTAG; Hidden Markov model; XTAG; MICA 1. Introduction Tree adjoining grammar (TAG), which was initially introduced by Joshi et al. (1975), is a tree generating system that forms the object language by a set of derived trees. This formalism as an extension of context free grammars (CFGs) is classified in the mildly context sensitive grammars (MCSGs), which itself is a grammatical class between the context-free and context sensitive-grammars (Joshi, 1985). In the lexicalized case, the elementary structures of the lexicalized tree-adjoining grammars (LTAGs) are assigned to the lexical items of the language. These elementary structures are called elementary trees and the lexical items assigned to them are called the anchors. Each elementary tree of a LTAG defines a syntactic environment in which its anchor can appear (Bangalore and Joshi, 1999). There are two kinds of elementary trees: initial trees and auxiliary  This paper has been recommended for acceptance by E. Briscoe. ∗ Corresponding author. Tel.: +98 21 82089717; fax: +98 21 88633029. E-mail addresses: a.basirat@srbiau.ac.ir (A. Basirat), hfaili@ut.ac.ir (H. Faili). 0885-2308/$ – see front matter © 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.csl.2013.02.001