Text Chunking Research Articles

Approaches to the Metres of Alliterative Verse (review)

Reviewed by: Approaches to the Metres of Alliterative Verse W. A. Davenport Judith Jefferson and Ad Putter, eds. Approaches to the Metres of Alliterative Verse. Leeds: Leeds Texts and Monographs, New Series 17, 2009. Pp. 311. £40.00; $65.00 paper. It is difficult to make a completely satisfying book from a set of conference papers, and the editors’ choice of title for this collection (emanating [End Page 346] from a 2005 conference in Bristol) highlights the differences of perspective in the thirteen essays; they do not even claim that the contributors are discussing a single meter, though they have provided a unified bibliography (but not an index). So, although a significant number of essays take as their starting point the existence of “rules” for the composition of the alliterative long line, following Hoyt Duggan and others, and attempt further definition of the patterns of lifts and dips, the constraints that operate in the two half-lines, the presence or absence of pronounced final -e, the hierarchies of allowable alliterating words, and so on, usually by subjecting chunks of poetic text to statistical processes, others return to basic questions about linguistic history and how alliteration works, and to the consideration of individual poems. In my youth, before computers, I served a two-year apprenticeship to the scholarly alliterative trade by analyzing the syntax of every sentence of Sir Gawain and the Green Knight, and from close study of J. P. Oakden's monumental Alliterative Poetry in Middle English developed, alongside interest in the subject, a fair dose of skepticism about the rum-ramruffery of it all. How did discussion of alliterative meter get stuck in “lifts and dips” when the ear tells us that strong stresses go down, not up? Why not hops and skips, thumps and flutters, landings and scurries? Noriko Inoue takes the nine manuscript versions of The Siege of Jerusalem as the coal-face from which to chisel evidence that preference for for to + infinitive over to + infinitive is always a metrically determined choice in b-verses and that, therefore, its use in a-verses is proof that the requirement for a long medial dip is stricter than the need to observe the aa/ax pattern. But I cannot help wondering why one should trust in matters metrical a poet who is such an indifferent performer in matters narrative. Working alongside are Nikolay Yakovlev, who challenges the view that final -e had disappeared in northern dialects by the end of the fourteenth century by examining all the b-verses in Sir Gawain; Donka Minkova, who uses the b-verses of Winner and Waster and The Parliament of the Three Ages to demonstrate constraints on metricality with an impressive display of statistics, tables and code words; Geoffrey Russom, who compares Beowulf and Sir Gawain in terms of the classes of words that may bear alliteration, to demonstrate patterns of alliterative usage and the contrasts between a-verses and b-verses; and Gilbert Youmans, who examines syntactic inversions in the alliterative Morte Arthure in order to rank the poet's metrical principles in order of strictness, with observance of the ax pattern in the b-verse scoring highest, followed by the avoidance of unstressed words at the line ending. All [End Page 347] these bear witness to a good deal of meticulous labor and are within the boundaries of the modern orthodoxies of rules and constraints; yet they have a quaint look about their methodology—Youmans almost made me nostalgic for my own unreadable M.A. thesis of fifty years ago by using some familiar examples of noun/adjective inversion such as “on a stede ryche.” It is refreshing to find some essays that step outside the world of the alliterative long line, as Thomas Cable does in examining the alliterative lyrics from the Harley manuscript in terms of Duggan's theory of alliterative meter and asking whether a native or foreign model lies behind them. His conclusion that there is no single alliterative meter but various meters that use alliteration “as a cue to metrically stressed syllables, which in turn establish a rhythm” opens a door to the broader discussion of alliteration as a metrical phenomenon. Allan Gaylord...

Comparing and combining chunkers of biomedical text

Text chunking is an essential pre-processing step in information extraction systems. No comparative studies of chunking systems, including sentence splitting, tokenization and part-of-speech tagging, are available for the biomedical domain. We compared the usability (ease of integration, speed, trainability) and performance of six state-of-the-art chunkers for the biomedical domain, and combined the chunker results in order to improve chunking performance.We investigated six frequently used chunkers: GATE chunker, Genia Tagger, Lingpipe, MetaMap, OpenNLP, and Yamcha. All chunkers were integrated into the Unstructured Information Management Architecture framework. The GENIA Treebank corpus was used for training and testing. Performance was assessed for noun-phrase and verb-phrase chunking.For both noun-phrase chunking and verb-phrase chunking, OpenNLP performed best (F-scores 89.7% and 95.7%, respectively), but differences with Genia Tagger and Yamcha were small. With respect to usability, Lingpipe and OpenNLP scored best. When combining the results of the chunkers by a simple voting scheme, the F-score of the combined system improved by 3.1 percentage point for noun phrases and 0.6 percentage point for verb phrases as compared to the best single chunker. Changing the voting threshold offered a simple way to obtain a system with high precision (and moderate recall) or high recall (and moderate precision).This study is the first to compare the performance of the whole chunking pipeline, and to combine different existing chunking systems. Several chunkers showed good performance, but OpenNLP scored best both in performance and usability. The combination of chunker results by a simple voting scheme can further improve performance and allows for different precision-recall settings.

An AdaBoost Using a Weak-Learner Generating Several Weak-Hypotheses for Large Training Data of Natural Language Processing

AdaBoost is a method to create a final hypothesis by repeatedly generating a weak hypothesis in each training iteration with a given weak learner. AdaBoost-based algorithms are successfully applied to several tasks such as Natural Language Processing (NLP), OCR, and so on. However, learning on the training data consisting of large number of samples and features requires long training time. We propose a fast AdaBoost-based algorithm for learning rules represented by combination of features. Our algorithm constructs a final hypothesis by learning several weak-hypotheses at each iteration. We assign a confidence-rated value to each weak-hypothesis while ensuring a reduction in the theoretical upper bound of the training error of AdaBoost. We evaluate our methods with English POS tagging and text chunking. The experimental results show that the training speed of our algorithm are about 25 times faster than an AdaBoost-based learner, and about 50 times faster than Support Vector Machines with polynomial kernel on the average while maintaining state-of-the-art accuracy.

LFXtractor: Text chunking for long form detection from biomedical text

In this paper, we propose a novel method to detect the corresponding long forms (LFs) of short forms (SFs) from biomedical text. The proposed method is differentiated from others as follows: it incorporates lexical analysis techniques into supervised learning for extracting abbreviations; it utilises text-chunking techniques to identify LFs of abbreviations; it significantly improves recall. The experimental results show that our approach outperforms the leading abbreviation algorithms, ExtractAbbrev, ALICE and Acrophile and a collocation-based approach at least by 4.8, 6.0, 9.0 and 6.0%, respectively, in both precision and recall on the Gold Standard Development corpus.

Experimenting Oriya Text Chunking with Divide-Conquer Strategy

The traditional oriya text chunking approach identifies phrase structure or local word group by using only one model and phrases with the same types of features. Generally oriya language is a free word order language. Free word order languages have relatively unrestricted local word group or phrase structures that make the problem of chunking quite challenging It has been shown that the limitations of using only one model are that: the use of the same types of features is not suitable for all phrases.. In this paper, the divide-conquer approach is proposed and applied in the identification of phrases or local word group. This strategy divides the task of chunking into several sub-tasks according to sensitive features of each phrase and identifies different phrases in parallel. Then, a two-stage decreasing conflict strategy is used to synthesize each sub-task’s answer We argue that we might not need an explicit intermediate POS-tagging step for parsing when a sufficient amount of training material is available and word form information is used for low-frequency words. By applying and testing the approach on the public training and test corpus, the F score for arbitrary phrases identification using divide-conquer strategy achieves 91.3% compared to the previous best F score of 92.18%. Keywords: feature structure, chunking, local word grouping, parsing(LWG), free word order languages, part-of-speech tagging, morphological analysis sensitive features; divide-conquer strategy

The Stroud/ADI dementia quality framework: a cross‐national population‐level framework for assessing the quality of life impacts of services and policies for people with dementia and their family carers

To generate an internationally applicable framework for the systematic assessment of the quality of life impacts of services and policies for people with dementia and their family carers. A qualitative analysis of text. Open groups at six consecutive Alzheimer's Disease International (ADI) Conferences (1999-2005) lasting between 1.5 and 2 h. Data were: presenters' texts, transcribed points made during discussion and written contributions. Three hundred and twelve contributors of text. From 2246 chunks of text, eight interacting domains relating to quality of life emerged: Public Attitudes and Understanding; Government and Social Policy; Funding for Services; Health; Communication; Choice and Personhood; Environment; and Quality of Care leading to the development of the Stroud/ADI Dementia Quality Framework. The Stroud/ADI Dementia Quality Framework helps to structure assessment of the quality of life impacts of population level approaches in dementia impacts of population-level approaches in dementia. Information available can be mapped onto the framework. With its international approach, the Stroud/ADI Dementia Quality Framework has validity across cultures within and between countries. It is intended as a useful aid for the assessment of services and policies for people with dementia and their family carers.

The elements of fracture fixation. Second ed.

Those of a certain vintage, who learnt the use of carpentry tools through their use and through error, will find much in this book which is delightful. All their hard-learned lessons are here confirmed and put into clear instructional words. It must be astonishing to the carpentry trade, with which

The discourse of lending aid on small-scale development project websites: Dutch depreciatory diminutives

This article offers an analysis of the use of language on the websites of Dutch and Flemish small-scale development organisations active in The Gambia. The scope of this research is the websites of 66 organisations found through the hyperlink page http://gambia-hulporganisaties.startpagina.nl. The texts on the websites form a small corpus of around 375 000 words. Methodologically, a discourse-analytical perspective is assumed, heuristically assisted by corpus linguistic software. Thus, the texts are analysed simultaneously from a macro and micro-level: large-scale lexical patterns are combined with smaller-scale, contextualised, individual chunks of text. After a brief outline of the projects’ roots in tourism through self-reported histories of involvement and a cursory review of the literature on meanings and functions of diminutives in Dutch and other languages, the role of diminutives in the representation of the Third- World Other is explored. It is argued that diminutives in this context are used in an ambivalent way: diminutives express a sense of sympathy and at the same time reveal a derogative tone in descriptions of the Gambian side of the projects.

Computerizing Mathematical Text with MathLang

Mathematical texts can be computerized in many ways that capture differing amounts of the mathematical meaning. At one end, there is document imaging, which captures the arrangement of black marks on paper, while at the other end there are proof assistants (e.g., Mizar, Isabelle, Coq, etc.), which capture the full mathematical meaning and have proofs expressed in a formal foundation of mathematics. In between, there are computer typesetting systems (e.g., ▪ and Presentation MathML) and semantically oriented systems (e.g., Content MathML, OpenMath, OMDoc, etc.).The MathLang project was initiated in 2000 by Fairouz Kamareddine and Joe Wells with the aim of developing an approach for computerizing mathematical texts and knowledge which is flexible enough to connect the different approaches to computerization, which allows various degrees of formalization, and which is compatible with different logical frameworks (e.g., set theory, category theory, type theory, etc.) and proof systems. The approach is embodied in a computer representation, which we call MathLang, and associated software tools, which are being developed by ongoing work. Three Ph.D. students (Manuel Maarek (2002/2007), Krzysztof Retel (since 2004), and Robert Lamar (since 2006)) and over a dozen master's degree and undergraduate students have worked on MathLang. The project's progress and design choices are driven by the needs for computerizing real representative mathematical texts chosen from various branches of mathematics.Currently, MathLang supports entry of mathematical text either in an XML format or using the ▪ editor. Methods are provided for adding, checking, and displaying various information aspects. One aspect is a kind of weak type system that assigns categories (term, statement, noun (class), adjective (class modifier), etc.) to parts of the text, deals with binding names to meanings, and checks that a kind of grammatical sense is maintained. Another aspect allows weaving together mathematical meaning and visual presentation and can associate natural language text with its mathematical meaning. Another aspect allows identifying chunks of text, marking their roles (theorem, definition, explanation, example, section, etc.), and indicating relationships between the chunks (A uses B, A contradicts B, A follows from B, etc.). Software tool support can use this aspect to check and explain the overall logical structure of a text. Further aspects are being designed to allow adding additional formality to a text such as proof structure and details of how a human-readable proof is encoded into a fully formalized version (so far this has only been done for Mizar and started for Isabelle). A number of mathematical texts have been computerized, helping with the development of these aspects, and indicating what additional work is needed for the future. This paper surveys the past and future work of the MathLang project.

Cognitive Support for the Organization of Writing

This paper describes a novel framework and a system to support writing professional documents. Our framework focuses on aiding the process of organizing ideas and constructing logical connections between text chunks in formal documents. Most previous approaches dealing with these issues have utilized one of three types of representations: a tree, a map, and a network. However, they all have their strengths and weaknesses. The tree is superior in helping the writers to grasp the hierarchical structure of a document, but it is difficult to express vague relationships among parts of a document. The map is easy to operate, but it restricts freedom in spatial organization. The network has flexibility to express what writers want, but it requires additional graphical objects to specify relationships among parts and it cannot be naturally transformed into a linear document. In our approach, the tree and the map are integrated: the local structure of a document is defined by the map while the global structure of the document is defined by the tree. Combining both representations has the following advantages: (1) it leverages the benefits of both representations, (2) it seamlessly supports both top-down organization and bottom-up organization, and (3) it smoothly supports writing from an ill-organized state to a well-organized state. Based on the proposed framework, we have built a system called iWeaver. We describe experiments in which six users wrote six technical documents using the system. The results show that our framework is an effective aid for constructing long, structured documents.

A Machine Learning Approach for an Indonesian-English Cross Language Question Answering System

We have built a CLQA (Cross Language Question Answering) system for a source language with limited data resources (e.g. Indonesian) using a machine learning approach. The CLQA system consists of four modules: question analyzer, keyword translator, passage retriever and answer finder. We used machine learning in two modules, the question classifier (part of the question analyzer) and the answer finder. In the question classifier, we classify the EAT (Expected Answer Type) of a question by using SVM (Support Vector Machine) method. Features for the classification module are basically the output of our shallow question parsing module. To improve the classification score, we use statistical information extracted from our Indonesian corpus. In the answer finder module, using an approach different from the common approach in which answer is located by matching the named entity of the word corpus with the EAT of question, we locate the answer by text chunking the word corpus. The features for the SVM based text chunking process consist of question features, word corpus features and similarity scores between the word corpus and the question keyword. In this way, we eliminate the named entity tagging process for the target document. As for the keyword translator module, we use an Indonesian-English dictionary to translate Indonesian keywords into English. We also use some simple patterns to transform some borrowed English words. The keywords are then combined in boolean queries in order to retrieve relevant passages using IDF scores. We first conducted an experiment using 2,837 questions (about 10% are used as the test data) obtained from 18 Indonesian college students. We next conducted a similar experiment using the NTCIR (NII Test Collection for IR Systems) 2005 CLQA task by translating the English questions into Indonesian. Compared to the Japanese-English and Chinese-English CLQA results in the NTCIR 2005, we found that our system is superior to others except for one system that uses a high data resource employing 3 dictionaries. Further, a rough comparison with two other Indonesian-English CLQA systems revealed that our system achieved higher accuracy score.

Encoding document information in a corpus of student writing: the British Academic Written English corpus

The information contained in a document is only partly represented by the wording of the text; in addition, features of formatting and layout can be combined to lend specific functionality to chunks of text (e.g., section headings, highlighting, enumeration through list formatting, etc.). Such functional features, although based on the ‘objective’ typographical surface of the document, are often inconsistently realised and encoded only implicitly, i.e., they depend on deciphering by a competent reader. They are characteristic of documents produced with standard text-processing tools. We discuss the representation of such information with reference to the British Academic Written English (BAWE) corpus of student writing, currently under construction at the universities of Warwick, Reading and Oxford Brookes. Assignments are usually submitted to the corpus as Microsoft Word documents and make heavy use of surface-based functional features. As the documents are to be transformed into XML-encoded corpus files, this information can only be preserved through explicit annotation, based on interpretation. We present a discussion of the choices made in the BAWE corpus and the practical requirements for a tagging interface.

High-Performance Training of Conditional Random Fields for Large-Scale Applications of Labeling Sequence Data

Conditional random fields (CRFs) have been successfully applied to various applications of predicting and labeling structured data, such as natural language tagging & parsing, image segmentation & object recognition, and protein secondary structure prediction. The key advantages of CRFs are the ability to encode a variety of overlapping, non-independent features from empirical data as well as the capability of reaching the global normalization and optimization. However, estimating parameters for CRFs is very time-consuming due to an intensive forward-backward computation needed to estimate the likelihood function and its gradient during training. This paper presents a high-performance training of CRFs on massively parallel processing systems that allows us to handle huge datasets with hundreds of thousand data sequences and millions of features. We performed the experiments on an important natural language processing task (text chunking) on large-scale corpora and achieved significant results in terms of both the reduction of computational time and the improvement of prediction accuracy.

Efficient text chunking using linear kernel with masked method

In this paper, we proposed an efficient and accurate text chunking system using linear SVM kernel and a new technique called masked method. Previous researches indicated that systems combination or external parsers can enhance the chunking performance. However, the cost of constructing multi-classifiers is even higher than developing a single processor. Moreover, the use of external resources will complicate the original tagging process. To remedy these problems, we employ richer features and propose a masked-based method to solve unknown word problem to enhance system performance. In this way, no external resources or complex heuristics are required for the chunking system. The experiments show that when training with the CoNLL-2000 chunking dataset, our system achieves 94.12 in F ( β) rate with linear. Furthermore, our chunker is quite efficient since it adopts a linear kernel SVM. The turn-around tagging time on CoNLL-2000 testing data is less than 50 s which is about 115 times than polynomial kernel SVM.

A robust multilingual portable phrase chunking system

Automatic text chunking aims to recognize grammatical phrase structures in natural language text. Text chunking provides downstream syntactic information for further analysis, which is also an important technology in the area of text mining (TM) and natural language processing (NLP). Existing chunking systems make use of external knowledge, e.g. grammar parsers, or integrate multiple learners to achieve higher performance. However, the external knowledge is almost unavailable in many domains and languages. Besides, employing multiple learners does not only complicate the system architecture, but also increase training and testing time costs. In this paper, we present a novel phrase chunking model based on the proposed mask method without employing external knowledge and multiple learners. The mask method could automatically derive more training examples from the original training data, which significantly improves system performance. We had evaluated our method in different chunking tasks and languages in comparison to previous studies. The experimental results show that our method achieves state of the art performance in chunking tasks. In two English chunking tasks, i.e., shallow parsing and base-chunking, our method achieves 94.22 and 93.23 in F(@b=1) rates. When porting to Chinese, the F(@b=1) rate is 92.30. Also, our chunker is quite efficient. The complete chunking time of a 50K-words is less than 10s.

English Text Chunking Based on Headword Extending and the Evaluation of Relative-Degree

English Text Chunking Based on Headword Extending and the Evaluation of Relative-Degree

ER Vets: Life in an Animal Emergency Room (review)

Reviewed by: ER Vets: Life in an Animal Emergency Room Deborah Stevenson Jackson, Donna M. ER Vets: Life in an Animal Emergency Room. Houghton, 2005 [96p] illus. with photographs ISBN 0-618-43663-4$17.00 Reviewed from galleys R Gr. 4-7 Going ER one better by making the patients cute and fuzzy, ER Vets focuses on the state-of-the-art emergency vet clinic at Colorado State University. The book provides some information about the history of veterinary medicine in general and some of these hospital staffers in particular, but the main stories here are those of the animals. Jackson, who researched on-site at the hospital and took most of the book's photographs herself, intersperses accounts of specific cases—a dog hit by a car, a foal with oxygen deprivation, a hypothermic snake—with discussion of emergency-room veterinary care in general, while sidebars provide information about topics such as common emergency-room abbreviations and human foods with devastating pet consequences. Though the non-linear organization makes this more a detailed scrapbook than a narrative (and the important issue of cost is never touched upon), the book makes effective use of its exciting and inviting material, making clear that the high-pressure, wildly varied work in the emergency room is indeed a specialization that requires particular abilities as well as interest in animals. The pictures are key in this photodocumentary, and while the setting sometimes limits photographic options, the images are informative and affecting; reproduced x-rays and 'scope pictures add crucial internal views and familiarize readers with the vets' own tools. The heavily visual layout is occasionally a little overbusy, but the captions and chapter titles keep text chunks easily manageable. This will be an easy sell to animal lovers of various reading levels, and they may also get some ideas about future career options. Extensive end matter includes a list of suggested reading, relevant websites, a glossary, source notes, and an index. Copyright © 2005 The Board of Trustees of the University of Illinois

日本語話し言葉の係り受け解析と文境界推定の相互作用による高精度化

This paper describes methods to detect dependencies between Japanese phrasal units called bunsetsus, and sentence boundaries in a spontaneous speech corpus. In spontaneous monologues, the biggest problem with dependency structure analysis is that sentence boundaries are ambiguous. In this paper, we propose two methods for improving the accuracy of sentence boundary detection in spontaneous Japanese speech: One is based on statistical machine translation using dependency information and the other is based on text chunking using SVM. An F-measure of 84.9 was achieved for the accuracy of sentence boundary detection by using the proposed methods. The accuracy of dependency structure analysis was also improved from 75.2% to 77.2% by using automatically detected sentence boundaries. Furthermore, the accuracy of dependency structure analysis and that of sentence boundary detection were improved by interactively using the counterpart results.

Combined Multiple Classifiers Based on a Stacking Algorithm and Their Application to Chinese Text Chunking

Combined Multiple Classifiers Based on a Stacking Algorithm and Their Application to Chinese Text Chunking

A tool for taking class notes

Students still take class notes using pencil and paper—although digital documents are more legible, easier to search in and easier to edit—in part because of the lack of software to support note-taking. Class notes are characterized by free spatial organization, many small chunks of text, and a dense mix of text and graphic elements. These characteristics imply that a note-taking system should use pen, keyboard and mouse-or-equivalent; allow the swift entry of text at any desired position; and minimize the need to switch between input tools. A system with these properties was built and used by 10 subjects in a controlled study and by four users in their classes. Some users preferred our system to pencil and paper, suggesting that taking class notes with the computer is feasible.