Extreme Tomato

Vacuosity incarnate.

As of Thu 09/09/10.

{\rtf1\ansi\deff0\adeflang1025 {\fonttbl{\f0\froman\fprq2\fcharset0 Times New Roman;}{\f1\froman\fprq2\fcharset0 Times New Roman;}{\f2\fswiss\fprq2\fcharset0 Arial;}{\f3\fnil\fprq0\fcharset0 Times New Roman;}{\f4\fnil\fprq2\fcharset0 DejaVu Sans;}{\f5\fnil\fprq2\fcharset0 Tahoma;}{\f6\fnil\fprq0\fcharset0 Tahoma;}} {\colortbl;\red0\green0\blue0;\red128\green128\blue128;} {\stylesheet{\s1\cf0{\*\hyphen2\hyphlead2\hyphtrail2\hyphmax0}\rtlch\af5\afs24\lang255\ltrch\dbch\af4\langfe255\hich\f0\fs24\lang2057\loch\f0\fs24\lang2057\snext1 Normal;} {\s2\sb240\sa120\keepn\cf0{\*\hyphen2\hyphlead2\hyphtrail2\hyphmax0}\rtlch\afs28\lang255\ltrch\dbch\langfe255\hich\f2\fs28\lang2057\loch\f2\fs28\lang2057\sbasedon1\snext3 Heading;} {\s3\sa120\cf0{\*\hyphen2\hyphlead2\hyphtrail2\hyphmax0}\rtlch\af5\afs24\lang255\ltrch\dbch\af4\langfe255\hich\f0\fs24\lang2057\loch\f0\fs24\lang2057\sbasedon1\snext3 Body Text;} {\s4\sa120\cf0{\*\hyphen2\hyphlead2\hyphtrail2\hyphmax0}\rtlch\af6\afs24\lang255\ltrch\dbch\af4\langfe255\hich\f0\fs24\lang2057\loch\f0\fs24\lang2057\sbasedon3\snext4 List;} {\s5\sb120\sa120\cf0{\*\hyphen2\hyphlead2\hyphtrail2\hyphmax0}\rtlch\af6\afs24\lang255\ai\ltrch\dbch\af4\langfe255\hich\f0\fs24\lang2057\i\loch\f0\fs24\lang2057\i\sbasedon1\snext5 caption;} {\s6\cf0{\*\hyphen2\hyphlead2\hyphtrail2\hyphmax0}\rtlch\af6\afs24\lang255\ltrch\dbch\af4\langfe255\hich\f0\fs24\lang2057\loch\f0\fs24\lang2057\sbasedon1\snext6 Index;} } {\info{\creatim\yr2009\mo3\dy26\hr18\min53}{\revtim\yr0\mo0\dy0\hr0\min0}{\printim\yr0\mo0\dy0\hr0\min0}{\comment StarWriter}{\vern3000}}\deftab709 {\*\pgdsctbl {\pgdsc0\pgdscuse195\pgwsxn11905\pghsxn16837\marglsxn1134\margrsxn1134\margtsxn1134\margbsxn1134\pgdscnxt0 Standard;}} \paperh16837\paperw11905\margl1134\margr1134\margt1134\margb1134\sectd\sbknone\pgwsxn11905\pghsxn16837\marglsxn1134\margrsxn1134\margtsxn1134\margbsxn1134\ftnbj\ftnstart1\ftnrstcont\ftnnar\aenddoc\aftnrstcont\aftnstart1\aftnnrlc \page\pard\plain \ltrpar\s3\cf0{\*\hyphen2\hyphlead2\hyphtrail2\hyphmax0}\sa120\ql\rtlch\af5\afs24\lang255\ltrch\dbch\af4\langfe255\hich\f0\fs24\lang2057\loch\f0\fs24\lang2057{\rtlch \ltrch\loch\f0\fs24\lang2057\i0\b0\ltrch\hich\f3\loch\f3 Adaptive document structuring} \par \pard\plain \ltrpar\s3\cf0{\*\hyphen2\hyphlead2\hyphtrail2\hyphmax0}\sa120\rtlch\af5\afs24\lang255\ltrch\dbch\af4\langfe255\hich\f0\fs24\lang2057\loch\f0\fs24\lang2057{\rtlch \ltrch\loch\f0\fs24\lang2057\i0\b0\ltrch\hich\f3\loch\f3 Traditionally computational and corpus linguistics has dealt with corpora represented in a simple form without document structuring information, such as paragraphs, headings, sub-headings, page breaks, sections, chapters etc. Even with the event of more co mplex encoding schemes such as the Text Encoding Initiative (TEI) and Corpus Encoding Standard (CES), the document structure tends to be coded by hand. Many corpora are still simple text documents, with no formatting or layout data.} \par \pard\plain \ltrpar\s3\cf0{\*\hyphen2\hyphlead2\hyphtrail2\hyphmax0}\sa120\rtlch\af5\afs24\lang255\ltrch\dbch\af4\langfe255\hich\f0\fs24\lang2057\loch\f0\fs24\lang2057{\rtlch \ltrch\loch\f0\fs24\lang2057\i0\b0\ltrch\hich\f3\loch\f3 Today, many applications of Natural Language Processing could be improved if the document structure was retained, e.g. machine translation, document summarisation, information extraction and retrieval. However, even where documents are born-digital the log ical structure is not always encoded in a machine-understandable manner (e.g. for the semantic web), rather it tends to be represented through font formatting features. Moreover, many document collections involve the conversion of formats (i.e. web-scrapin g, transcription or text-to-speech) which entails the loss of logical document structure. Vast quantities of archive and historical data are now being digitised e.g. Academic Journal back issues, Early English Books Online, British Library Nineteenth Centu ry Newspapers etc. The conversion of these is now partially automated, employing OCR techniques. This process is still far from effortless, and remains impractical for many real-world applications \'96 even where data are already digitised they are often sto red in layout-oriented formats.} \par \pard\plain \ltrpar\s3\cf0{\*\hyphen2\hyphlead2\hyphtrail2\hyphmax0}\sa120\rtlch\af5\afs24\lang255\ltrch\dbch\af4\langfe255\hich\f0\fs24\lang2057\loch\f0\fs24\lang2057{\rtlch \ltrch\loch\f0\fs24\lang2057\i0\b0\ltrch\hich\f3\loch\f3 Document structure analysis is capable of informing this process of digitisation, but the algorithms need to be able to adapt to their context (both in terms of historical relevance and the origin of the document). Understanding the changes in document str ucture relative to time, format and context will allow for more accurate analysis, summarisation and categorisation, allowing vast catalogs of existing data to be processed and used as corpora.} \par \pard\plain \ltrpar\s3\cf0{\*\hyphen2\hyphlead2\hyphtrail2\hyphmax0}\sa120\rtlch\af5\afs24\lang255\ltrch\dbch\af4\langfe255\hich\f0\fs24\lang2057\loch\f0\fs24\lang2057{\rtlch \ltrch\loch\f0\fs24\lang2057\i0\b0\ltrch\hich\f3\loch\f3 I wish to work on developing statistical methods of text analysis that use and encode document-level structural features in order to:} \par \pard\plain \ltrpar\s3\cf0{\*\hyphen2\hyphlead2\hyphtrail2\hyphmax0}\sa120\rtlch\af5\afs24\lang255\ltrch\dbch\af4\langfe255\hich\f3\fs24\lang2057\loch\f3\fs24\lang2057 {\rtlch \ltrch\loch\f3\fs24\lang2057\i0\b0 * Inform resolution of text flow in OCR and other format-conversion mechanisms (including web as corpus and historical text),} \par \pard\plain \ltrpar\s3\cf0{\*\hyphen2\hyphlead2\hyphtrail2\hyphmax0}\sa120\rtlch\af5\afs24\lang255\ltrch\dbch\af4\langfe255\hich\f0\fs24\lang2057\loch\f0\fs24\lang2057 {\rtlch \ltrch\loch\f3\fs24\lang2057\i0\b0 * Refine and aid existing analysis techniques when used with data which exhibit disparate style (such as historical and modern academic sources),} \par \pard\plain \ltrpar\s3\cf0{\*\hyphen2\hyphlead2\hyphtrail2\hyphmax0}\sa120\rtlch\af5\afs24\lang255\ltrch\dbch\af4\langfe255\hich\f0\fs24\lang2057\loch\f0\fs24\lang2057{\rtlch \ltrch\loch\f0\fs24\lang2057\i0\b0\ltrch\hich\f3\loch\f3 * Develop techniques for better managing large bodies of text (summarisation, semantic analysis and other methods necessary for practical use).} \par }