EPSRC Reference: |
EP/P007104/1 |
Title: |
Comparing linguistic representations for scope resolution |
Principal Investigator: |
Read, Dr J |
Other Investigators: |
|
Researcher Co-Investigators: |
|
Project Partners: |
|
Department: |
Sch of Computing |
Organisation: |
Teesside University |
Scheme: |
First Grant - Revised 2009 |
Starts: |
03 January 2017 |
Ends: |
02 July 2018 |
Value (£): |
95,611
|
EPSRC Research Topic Classifications: |
Artificial Intelligence |
Comput./Corpus Linguistics |
|
EPSRC Industrial Sector Classifications: |
No relevance to Underpinning Sectors |
|
|
Related Grants: |
|
Panel History: |
|
Summary on Grant Application Form |
This project operates at the interface between natural language processing (the engineering of computer systems that handle unstructured data represented by human language) and computational linguistics (a field of research that seeks to understand human language through linguistic theories and computational modelling). The fields interact extensively, and their impact is evident in many commonly-used applications --- from Internet search engines through to speech-enabled personal assistants and on to unstructured data mining using text analytics (a global market predicted to reach a value of US$6.5 billion by 2020).
Cutting-edge applications of natural language processing (for instance sentiment analysis, national and corporate intelligence gathering, and text mining) need to deal with various subtle aspects of language such as negation (when a writer inverts the meaning of language), speculation (the degree to which an author expresses certainty), and attribution (when a writer refers to expressions originated by other people). This research will approach the analysis of each of these aspects of language by framing them as instances of two more generic tasks: cue detection and scope resolution. That is, analysis will be a matter of firstly detecting words that are indicative of the phenomenon and secondly determining the parts(s) of the sentence(s) it affects (i.e., what words are negated, speculated, or attributed). To achieve this, the project will explore how best to generate and describe candidate scopes for subsequent analysis by machine learning algorithms.
To generate and characterise candidate scopes the project builds upon research in computational linguistics, by exploring a variety of linguistic representations (i.e., formal descriptions of the structure of sentences and other linguistic phenomena). Using the representations to generate candidates and their features for input into machine learning algorithms, the project will evaluate and compare the contribution of each representation to each instance of the scope resolution task. Furthermore, the project will explore how best to apply information from each representation in parallel, so as to achieve an optimum balance between the reliability of the analyses produced and the computational cost in terms of processing time and memory use.
The resulting technology will be deployed as a package of software, models, and recommendations for each of the scope resolution task instances. The package will also facilitate the research and development of models for novel instances of the cue detection and scope resolution tasks. The package will be freely available for use by both academia and industry.
|
Key Findings |
This information can now be found on Gateway to Research (GtR) http://gtr.rcuk.ac.uk
|
Potential use in non-academic contexts |
This information can now be found on Gateway to Research (GtR) http://gtr.rcuk.ac.uk
|
Impacts |
Description |
This information can now be found on Gateway to Research (GtR) http://gtr.rcuk.ac.uk |
Summary |
|
Date Materialised |
|
|
Sectors submitted by the Researcher |
This information can now be found on Gateway to Research (GtR) http://gtr.rcuk.ac.uk
|
Project URL: |
|
Further Information: |
|
Organisation Website: |
http://www.tees.ac.uk |