Learning to extract information from semi-structured text using a discriminative context free grammar

Abstract

In recent work, conditional Markov chain models (CMM) have been used to extract information from semi-structured text (one example is the Conditional Random Field [10]). Applications range from finding the author and title in research papers to finding the phone number and street address in a web page. The CMM framework combines a priori knowledge encoded as features with a set of labeled training data to learn an efficient extraction process. We will show that similar problems can be solved more effectively by learning a discriminative context free grammar from training data. The grammar has several distinct advantages: long range, even global, constraints can be used to disambiguate entity labels; training data is used more efficiently; and a set of new more powerful features can be introduced. The grammar based approach also results in semantic information (encoded in the form of a parse tree) which could be used for IR applications like question answering. The specific problem we consider is of extracting personal contact, or address, information from unstructured sources such as documents and emails. While linear-chain CMMs perform reasonably well on this task, we show that a statistical parsing approach results in a 50% reduction in error rate. This system also has the advantage of being interactive, similar to the system described in [9]. In cases where there are multiple errors, a single user correction can be propagated to correct multiple errors automatically. Using a discriminatively trained grammar, 93.71% of all tokens are labeled correctly (compared to 88.43% for a CMM) and 72.87% of records have all tokens labeled correctly (compared to 45.29% for the CMM).

Information Extraction from Hypertext Mark-Up Language Web Pages

“…Paul Viola and Mukund Narasimhand [15] , present a classification algorithm based on discriminatively trained Context Free Grammar (CFG) to extract information from HTML text. The challenge is in converting the HTML information of customer (which is already available in an unstructured form on web sites and in email) into the regularized or schematized form required by a database system.…”

Relaxation Labeling for Selecting and Exploiting Efficiently Non-local Dependencies in Sequence Labeling

“…As baseline models we used a Maximum Entropy classifier (the local classifier described in Section 3.1), standard linear-chain CRF 5 and a grammar-based extraction approach similar to the ones presented by [8] or [22]. Because of its computational complexity, the SVM'ISO approach we described in Section 2.2 cannot be used on our corpora.…”

“…Another popular idea consists in capturing interaction among labels in a hierarchical approach [7,8]. For instance, in an Information Extraction task, [8] proposes to use a Context Free Grammar to escape the "linear tyranny" of chain-models.…”

“…For instance, in an Information Extraction task, [8] proposes to use a Context Free Grammar to escape the "linear tyranny" of chain-models. Within this framework, inference amounts to syntactic parsing.…”

“…The first one [7,8] relies on a grammar-based formalism to model non-local relationships by introducing a hierarchy of hidden variables, while the second one proposes alternative inference procedures like Gibbs sampling [9] or Integer Linear Programming [10]. In most of these methods, approximate inference algorithms are used to allow tractable inference with long-range dependencies.…”