Language and Computation Day 2010
University of Essex
Time: tbc – Friday Oct 8, 2010
Location: 3.320
Programme
Here is the provisional programme:
[Coffee will be available from 9:30]
- 09:45 Kakia Chatsiou Welcome; Recent activities of the LAC group. ( slides )
- 10:00 Dyaa Albakour (Essex) AutoAdapt @ TREC 2010 ( slides )
- 10:30 Massimo Poesio tbc.
- 11:00 [Coffee]
- 11:15 Deirdre Lungley & Jon Chamberlain (Essex) Towards Adaptive Interactive Search
- 11:45 Chris Fox (Essex) Judgemental Imperatives ( slides )
- 12:15 [Lunch]
- 13:30 Sonja Eisenbeiss (Essex) Acquisition of English Possessive Constructions
- 14:00 Mahmoud El-Haj (Essex) Arabic Multidocument Text Summarisation
- 14:30 Kakia Chatsiou (Essex) Essex Gateway to Arabic Resources Project ( slides )
- 15:00 [Coffee]
- 15:15 Udo Kruschwitz (Essex) (+ many others) Moving towards Adaptive Search in Digital Libraries
- 15:45 Richard Sutcliffe (Limerick), Udo Kruschwitz (Essex) & Kieran White (Limerick) Named Entity Recognition in Intranet Documents and Query Logs
[The Programme will be followed by an evening of ribald celebration and wild excess, or perhaps refined conversation and genteel socializing, in Wivenhoe.]
Abstracts/Slides
tbc (Essex) - Welcome and Recent Activities of the LAC group
Abstract: This talk will give an introduction to the group and an overview of what the group has been up to since the last LAC day.
Dyaa Albakour (Essex) - AutoAdapt @ TREC 2010
Abstract: This talk presents the contribution of the Autoadapt project in the session track of the Text REtrieval Conference (TREC) 2010. I will give a description of the task introduced in this year's track and the experiments we run here at Essex to solve this task.
Massimo Poesio () - tbc
Abstract: Coming soon!
Deirdre Lungley & Jon Chamberlain (Essex) - Towards Adaptive Interactive Search
Abstract: The AutoAdapt project here at Essex is exploring various domain model learning frameworks. Deirdre opens this talk by detailing her lattice-based framework and recent lessons learnt. Jon continues with an outline of the new adaptive interactive search interface for the Essex intranet.
Chris Fox (Essex) - Judgemental Imperatives
Abstract: Coming soon!
Sonja Eisenbeiss (Essex) - Acquisition of English Possessive Constructions
Abstract: In this talk, I will present results from a study on English-speaking children's acquisition of possessive constructions like the Mary's cat or engine of the old red car. Based on an analysis of data from the Brown (1973) corpus of American child language, I will argue that children adapt to the constraints on the use of -s and of early on.
Mahmoud El-Haj (Essex) - Arabic Multidocument Text Summarisation
Abstract: Multi-document summarisation produces a single summary of a set of related documents that belong to the same category. The analysis in this area is mostly done on the sentence or document level. This talk will focus on the progress of my research towards Arabic single and multi-document text summarisation. I will present some results of experiments done to other languages including English and Hebrew in addition to that I will show the results of summarising automatically-translated articles (from English to Arabic) using Google Translate.
Kakia Chatsiou (Essex) - Essex Gateway to Arabic Resources Project
Abstract: The talk will report on the Essex Gateway to Arabic Linguistic Resources, a collection of resources, bibliography and tools on Arabic Language and Linguistics. The project has been funded by a Departmental Research Innovation Scheme Award to Sonja Eisenbeiss and Louisa Sadler and its aim is to build a network of open-access digital collections and resources on Arabic Language and Linguistics at first, then expanding to Semitic Languages and Linguistics. We hope that these resources will unlock access to a wide range of our departmental scholarly output and serve as a point of reference for researchers/visitors interested in Arabic and Semitic Languages.
Udo Kruschwitz (Essex) - Moving towards Adaptive Search in Digital Libraries
Abstract:Search applications have become very popular over the last decade, one of the main drivers being the advent of the Web. Nevertheless, searching on the Web is very different to searching on smaller, often more structured collections such as digital libraries, local Web sites and intranets. One way of helping the searcher locating the right information for a specific information need is by providing well-structured domain knowledge to assist query modification and navigation. There are two challenges: acquiring the domain knowledge and adapting it automatically to the specific interests of the user community. We will outline how a domain model can be automatically acquired using search engine query logs and continuously be updated using methods resembling ant colony behaviour.
Richard Sutcliffe (Limerick), Udo Kruschwitz (Essex) & Kieran White (Limerick) - Named Entity Recognition in Intranet Documents and Query Logs
Abstract: The recognition of Named Entities - person names, company names, places, times, dates, and so on - has long been recognised as being of great importance in text processing. We have observed that NEs commonly occur in search engine logs such as that listing queries submitted to the University of Essex web search engine UKSearch. In this case, however, NEs are often specific to the Essex intranet domain, for example room numbers, committee names or research group titles. We call these Specific NEs or SNEs. Firstly, we have conducted a manual study of SNEs within a sample of queries to give us breakdown statistics of different SNE types, and to produce a provisional itinerary of them. Secondly, we have undertaken a pilot study in which SNEs within queries were recognised following training using the maximum entropy tagger within the NLPTools package. The key technique here was to obtain contexts for the candidate SNE (which is often he sole element within a query) by harvesting occurrences of them from web pages within the intranet.
Thirdly, we have crawled the entire Essex intranet and hence have carried out further SNE analysis on this.