Publishing Partner: Cambridge University Press CUP Extra Publisher Login

New from Cambridge University Press!


Revitalizing Endangered Languages

Edited by Justyna Olko & Julia Sallabank

Revitalizing Endangered Languages "This guidebook provides ideas and strategies, as well as some background, to help with the effective revitalization of endangered languages. It covers a broad scope of themes including effective planning, benefits, wellbeing, economic aspects, attitudes and ideologies."

E-mail this page 1

We Have a New Site!

With the help of your donations we have been making good progress on designing and launching our new website! Check it out at!
***We are still in our beta stages for the new site--if you have any feedback, be sure to let us know at***

Dissertation Information

Title: Approaches to Automatic Biographical Sentence Classification: An empirical study Add Dissertation
Author: Mike Conway Update Dissertation
Email: click here to access email
Institution: University of Sheffield, Department of Computer Science
Completed in: 2007
Linguistic Subfield(s): Computational Linguistics;
Director(s): Robert Gaizauskas

Abstract: This thesis addresses the problem of the reliable identification of
biographical sentences, an important subtask in several natural language
processing application areas (for example, biographical multiple document
summarisation, biographical information extraction, and so on). The
biographical sentence classification task is placed within the framework of
genre classification, rather than traditional topic based text classification.

Before exploring methods for doing this task computationally, we need to
establish whether, and with what degree of success, humans can identify
biographical sentences without the aid of discourse or document structure.
To this end, a biographical annotation scheme and corpus was developed, and
assessed using a human study. The human study showed that participants were
able to identify biographical sentences with a good level of agreement.

The main body of the thesis presents a series of experiments designed to
find the best sentence representations for the automatic identification of
biographical sentences from a range of alternatives. In contrast to
previous work, which has centred on the use of single terms (that is,
unigrams) for biographical sentence representations, the current work
derives unigram, bigram and trigram features from a large corpus of
biographical text (including the British Dictionary of National Biography).
In addition to the use of corpus derived n-grams, a novel characteristic of
the current approach is the use of biographically relevant syntactic
features, identified both intuitively and through empirical methods.

The experimental work shows that a combination of n-gram features derived
from the Dictionary of National Biography and biographically orientated
syntactic features yield a performance that surpasses that gained using
n-gram features alone. Additionally, in accordance with the view of
biographical sentence classification as a genre classification task,
stylistic features (for example, topic neutral function words) are shown to
be important for recognising biographical sentences.