Can we parse without tagging?

2007

We aim to improve the performance of a syntactic parser that uses a part-of-speech (POS) tagger as a preprocessor. Pipelined parsers consisting of POS taggers and syntactic parsers have several advantages, such as the capability of domain adaptation. However the performance of such systems on raw texts tends to be disappointing as they are affected by the errors of automatic POS tagging. We attempt to compensate for the decrease in accuracy caused by automatic taggers by allowing the taggers to output multiple answers when the tags cannot be determined reliably enough. We empirically verify the effectiveness of the method using an HPSG parser trained on the Penn Treebank. Our results show that ambiguous POS tagging improves parsing if outputs of taggers are weighted by probability values, and the results support previous studies with similar intentions. We also examine the effectiveness of our method for adapting the parser to the GENIA corpus and show that the use of ambiguous POS taggers can help development of portable parsers while keeping accuracy high.

SYNTAGMA is a rule-based parsing system, structured on two levels: a general grammar and a language specific grammars. The general grammar is implemented in the program; language specific grammars are resources conceived as text files which contain a lexical database with meaning-related grammatical features, a description of constituent structures, a meaning-specific syntactic constraints database and a semantic network. Since its theoretical background is principally Tesnière's Éléments de syntaxe, SYNTAGMA's grammar emphasizes the role of argument structure (valency) in constraint satisfaction, and allows also horizontal bounds, for instance treating coordination. Notions such as traces, empty categories are derived from Generative Grammar and some solutions are close to Government & Binding Theory, although they are the result of an autonomous research. These proprieties allow SYNTAGMA to manage complex syntactic configurations and well known weak points in parsing engineering. An important resource is the semantic network, which is used by SYNTAGMA in disambiguation tasks. In contrast to statistical and data driven parsers, the system's behavior may be controlled and fine-tuned, since gaps, traces and long-term relations are structurally set and its constituent generation process is not a linear left-to-right shift and reduce, but a bottom-up, rule driven procedure.

1999

In this paper we argue in favour of an integration between statistically and syntactically based parsing by presenting data from a study of a 500,000 word corpus of Italian. Most papers present approaches on tagging which are statistically based. None of the statistically based analyses, however, produce an accuracy level comparable to the one obtained by means of linguistic rules [1]. Of course their data are strictly referred to English, with the exception of [2, 3, 4]. As to Italian, we argue that purely statistically based approaches are inefficient basically due to great sparsity of tag distribution -50% or less of unambiguous tags when punctuation is subtracted from the total count. In addition, the level of homography is also very high: readings per word are 1.7 compared to 1.07 computed for English by [2] with a similar tagset. The current work includes a syntactic shallow parser and a ATN-like grammatical function assigner that automatically classifies previously manually verified tagged corpora. In a preliminary experiment we made with automatic tagger, we obtained 99,97% accuracy in the training set and 99,03% in the test set using combined approaches: data derived from statistical tagging is well below 95% even when referred to the training set, and the same applies to syntactic tagging. As to the shallow parser, we report on a first preliminary experiment on a manually verified subset made of 10,000 words.

This paper describes work on the linguistic analysis of texts within a project devoted to knowledge acquisition from text. We focus on syntactic processing and present some key elements of the project's parser that allow it to deal successfully with technical texts. The parser is fully implemented and tested on a variety of real texts; improvements and enhancements are in progress. Because our knowledge acquisition method assumes no a priori model of the domain of the source text, the parser relies as much as possible on lexical and syntactic clues. That is why it strives for full syntactic analysis rather than some form of text skimming. We present a practical approach to four acknowledged difficult problems which to date have no generally acceptable answers: phrase attachment; time constraints for problematic input (how to avoid long and unproductive computation); parsing conjoined structures (how to preserve broad coverage without losing control of the parsing process); and the treatment of fragmentary input or fragments that are a by-product of a fallback parsing strategy. We review recent related work and conclude by listing several future work items.

Proc. of the ECAI98, 1998

Several (and successfull) Information Extraction systems have recently replaced the core parsing components with shallow but more ecient recognizers. In this paper we argue that the absence of an underlying grammatical recognizer, given the complex nature of several (non-english) languages, is a strong limitation for text processing functionalities, like those an IE system needs. We propose a robust and ecient syntactic recognizer mainly aimed to capture grammatical information crucial for several linguistic and non linguistic inferences. The proposed system is based on a novel architecture exploiting two major principles: lexicalization and stratication of the parsing process. As several linguistic theories (e.g. HPSG) and parsing frameworks (e.g. LTAG, SLTAG, lexicalized probabilistic parsing) suggest, lexicon-driven systems ensure the suitable forms of grammatical control for many complex phenomena. In our system an analysis guided by information on typical verb projections (e.g. verb subcategorization structures) is coupled with extended locality constraints (i.e. recognition of clause boundaries). Furthermore, stratication is also employed. A cascade of processing steps starts from chunk recognition and proceeds through clause analysis to dependency detection. Recognition of chunks allows to minimize the input ambiguity to the remaining phases. The resulting system is thus robust against ungrammatical phenomena (e.g. complex clause embedding, misspellings, unknown words). Eciency is also retained, although ambiguous phenomena (multiple PP attachments) are recognized.

Proceedings of SNRFAI99, Bilbao, …, 1999

In this paper, we present APOLN (Analizador Parcial de Oraciones en Lenguaje Natural): a partial parser of unrestricted natural language sentences based on finite-state techniques. Partial parsing has been used in several applications: syntactic parsing of unrestricted texts, data extraction systems, machine translation, solving the attachment ambiguity, speech recognition systems, text summarization, etc. The main attractiveness of partial parsing is that is able to handle unrestricted sentences, that contain lexical errors or that present constructions not accepted by the defined grammar. Partial parsing is an alternative to the definition of wide coverage grammars whose definition is an expensive and complex task and that present well-known problems such as overgeneration, undergeneration and ambiguity. We present APOLN as a tool that can be used to construct syntactically annotated corpora from lexically tagged corpora. We also present the results (precision and recall rates) of applying APOLN on unrestricted Spanish corpora, and how tagging errors influence the performance of the parser.

This paper provides a critical review of two highly esteemed parsers: the Stanford and the Berkeley Parsers. The parsers are examined in terms of the constituency. Both parsers prove efficiency, especially with complex sentences, yet Berkeley's parser provides more efficient accuracy. Both parsers could not deal with common syntactic challenges, namely garden path, elliptical sentences, and topicalization. It is recommended that both parsers provide additional user-friendly interfaces and services.

2008

This paper presents a comparative evaluation of several state-of-the-art English parsers based on different frameworks. Our approach is to measure the impact of each parser when it is used as a component of an information extraction system that performs protein-protein interaction (PPI) identification in biomedical papers. We evaluate eight parsers (based on dependency parsing, phrase structure parsing, or deep parsing) using five different parse representations. We run a PPI system with several combinations of parser and parse representation, and examine their impact on PPI identification accuracy. Our experiments show that the levels of accuracy obtained with these different parsers are similar, but that accuracy improvements vary when the parsers are retrained with domain-specific data.

2018

Information retrieval (IR) has become one of the most popular Natural Language Processing (NLP) applications. Part of speech (PoS) parsing and tagging plays an important role in IR systems. A broad range of PoS parsers and taggers tools have been proposed with the aim of helping to find a solution for the information retrieval problems, but most of these are tools based on generic NLP tags which do not capture domain-related information. In this research, we present a domain-specific parsing and tagging approach that uses not only generic PoS tags but also domain-specific PoS tags, grammatical rules, and domain knowledge. Experimental results show that our approach has a good level of accuracy when applying it to different domains.

Computational linguistics, 1999

In this paper, we have proposed novel methods for robust parsing that integrate the flexibility of linguistically motivated lexical descriptions with the robustness of statistical techniques. Our thesis is that the computation of linguistic structure can be localized iflexical items are associated with rich descriptions (supertags) that impose complex constraints in a local context. The supertags are designed such that only those elements on which the lexical item imposes constraints appear within a given supertag. Further, each lexical item is associated with as many supertags as the number of different syntactic contexts in which the lexical item can appear. This makes the number of different descriptions for each lexical item much larger than when the descriptions are less complex, thus increasing the local ambiguity for a parser. But this local ambiguity can be resolved by using statistical distributions of supertag co-occurrences collected from a corpus of parses. We have explored these ideas in the context of the Lexicalized Tree-Adjoining Grammar (LTAG) framework. The supertags in LTAG combine both phrase structure information and dependency information in a single representation. Supertag disambiguation results in a representation that is effectively a parse (an almost parse), and the parser need "only" combine the individual supertags. This method of parsing can also be used to parse sentence fragments such as in spoken utterances where the disambiguated supertag sequence may not combine into a single structure.