Slides - Slide 1

Document Sample
Slides - Slide 1 Powered By Docstoc
					  SWM43 Research in Computing:
Introduction to Computing Research

                   Anja Belz
      Natural Language Technology Group
                     CMIS
         A.S.Belz@brighton.ac.uk


                                          1
         Purpose of this lecture
• To touch on all aspects of research at least
  briefly, giving pointers to further reading (see
  list of references on last two slides)
• To look in more detail at the two areas that
  are of most practical relevance to you at the
  moment:
  – Planning research
  – Reviewing the literature
• To give practical guidance (including an
  exercise) on how to get started with research
                                                     2
Overview
 Part I – Introduction
    1.   Four golden lessons
    2.   What field are we in anyway?
    3.   Examples of research projects in informatics

 Part II – Preparing the ground
    4.   Planning research
    5.   Reviewing the literature
    6.   Exercise: first steps in planning research




                                                        3
Part I – Introduction
 1. Four golden lessons
 2. What field are we in anyway?
 3. Examples of research projects in
    informatics




                                       4
Four golden lessons (Weinberg, 2003)
 1. No one knows everything, and you don't
    have to.
 2. Go for the messes — that's where the
    action is.
 3. Forgive yourself for wasting time.
 4. Learn something about the history of
    science, or at a minimum the history of your
    own branch of science.
                                               5
What field are we in anyway?
 • Computing science, computer science, information
   processing, information technology, software
   engineering … ?
 • Now increasingly known as: Informatics
 • Old subdivisions:
   – Computer science: study/build artificial systems
   – Artificial intelligence: emulate intelligence using artificial
     systems
   – Cognitive science: study mind from computational
     perspective
   … aren’t really separate anymore!
                                                                      6
Informatics techniques (Bundy)
 Types of techniques:
    •    Theories
    •    Architectures
    •    Information representation
    •    Algorithms
    •    Software engineering processes
 Research aims:
    1.   Extend knowledge about properties of techniques
    2.   Improve existing techniques
    3.   Create new techniques
    4.   Combine techniques to create systems              7
Informatics techniques (Bundy)
 Implications of technique type and research aim:
    – What broad phases your research will go through
    – What kind of evaluation is appropriate
    – What kind of use will be made of your research
 Specific research field also important:
    – Fine-grained divisions: evolutionary robotics, machine
      translation, computer vision, etc.
    – Standard tools and methodologies to apply
    – Terminology, knowledge you can take as given
    – Dissemination media and style
                                                               8
Examples of informatics research
• Academic:
   – Masters and PhD thesis research
   – Research internally funded by universities etc.; some
     incidental, some more formally in projects
   – Publicly funded research projects (UK research councils,
     European Commission, US National Science Foundation)
• Industry: R&D departments, research labs, dedicated
  research companies; limited dissemination
• Private: people working from their home computers, in
  their garages, attics, garden sheds
   – French girl who invented speech recognition technology in
     her parents’ garage
   – Chinese farmer Wu’s robots
   – Participants in DARPA Autonomous Vehicle Challenge        9
Examples of informatics research
 • Some Informatics MSc topics from University of
   Edinburgh (2003-07):
    – A P2P Network Visualiser
    – Automated Probability Assessment in Plausible Crime
      Diagnosis
    – EvoTanks II: Co-evolutionary Development of Game
      Playing Agents
    – Cooperative Multi Agent Systems in Automobiles
    – A Web Service Interface to Astronomical Databases
    – Clustering Tags of Social Bookmarking Sites
 • Hardware Evolution: Automatic design of electronic
   circuits in reconfigurable hardware by artificial evolution
   (PhD, University of Sussex)
                                                                 10
Examples of informatics research
 • Probabilistic Deep Generation (3-year project, funded
   by EPSRC): ―to develop, for the first time, a comprehensive,
   linguistically informed, probabilistic methodology for
   generating language that substantially improves development
   time, reusability and language variation in language generation
   systems, and thereby enhances their commercial viability‖.
   Value (£):211,199, University of Brighton.
 • Research Consortium in Speckled Computing (4-year
   project, funded by EPSRC): ―a radically new concept in
   information technology […] realised by minute autonomous
   specks, each of which encapsulates sensing, programmable
   computation and wireless networking. Computing with
   minute specks will enable linkages between the material and
   digital worlds […] will be fundamental to truly ubiquitous
   computing‖. Value (£):3,721,432, Edinburgh.                 11
Part I – take-home points
 • Take on board Weinberg’s 4 golden lessons
 • There is a huge variety of research in informatics
    – Be aware of the types of technique you’re working
      on, and of what your overall research aim is
    – It’s important to learn about the research
      methodologies, evaluation criteria and other
      conventions in your field of informatics research
 • You can make an important contribution to
   research wherever you are
                                                          12
Part II – Preparing the ground
 4. Planning research
   •   First steps
   •   Background reading
   •   Methodology
   •   Aims, outcomes, deliverables
   •   Requirements
   •   Subdividing and estimating effort: work packages
   •   Writing a research proposal
 5. Reviewing the literature
                                                      13
Planning research – first steps
 • Aim: Come up with a rough sketch of the research you
   want to do
 • Identify an area of informatics that interests you
 • Do some superficial background reading:
    – Wikipedia – but bear in mind it’s not 100% reliable, and don’t
      cite it!
    – Course web pages from leading universities
    – Websites of professional organisations
    – Research project websites
    – Individual researchers’ webpages
 • Decide on a smaller area in which to locate your
   research (but still larger than your project), and
   identify several ways in which you could make an
   original contribution
                                                                14
Planning research – first steps
 • Start identifying key characteristics of your chosen
   area:
    – Important conferences
    – Leading journals
    – Internationally leading researchers
    – Current research projects
    – Mailing lists
    – History: when did it begin? What are key advances, when
      did they happen?
    – How good is current technology?
    – What are the issues research is currently grappling with?
                                                                  15
Planning research – background reading
 • Aim: familiarise yourself with chosen field of research well
   enough to decide on your project and write a short
   outline of it
 • Read survey papers and look up textbooks
 • Start reading (abstracts of) academic papers in
   conference proceedings and journals
 • Start compiling a bibliography (with star ratings)
 • Use tools like Google Scholar to check status of
   publications
 • Continue to collect key characteristics of area
 • Write project outline (a few sentences)
                                                             16
Planning research – methodology
 • Aim: to decide on the technical details of how you’re
   going to carry out your research
 • Won’t be able to specify all of this in advance – some
   of it is necessarily part of doing the research
 • E.g. if building a system: outline of architecture and
   functionality of modules; general approach (e.g.
   symbolic or statistical), even algorithms
 • Good idea to include fall-back options (if A doesn’t
   work I’ll do B)

                                                       17
Planning research – aims, outcomes, deliverables

 • Aim: clarify the purpose of your research to the point where you
   can write it down in detail
 • Aims: overall goals you hope to achieve with your research
 • Outcomes: specific results you plan to achieve with your
   research
    – Will knowledge be increased? How?
    – Will new resources be produced? Which ones?
    – Will new techniques be created? Existing ones improved?
 • Deliverables: the specific documents, software and other
   resources you commit to producing, with deadlines
    – Technical reports, manuals, webpages, etc.
    – Software specifications (modules, interfaces), tools, systems, etc.
    – Data collections (database of images, corpora of texts, etc.)
                                                                            18
Planning research – requirements
 • Aim: to determine everything you will need to carry out
   your project, apart from your own time and effort
 • Are you going to carry out experiments involving
   people? How many subjects? Will the university’s
   research ethics allow it?
 • Programming environments, tools, your skills.
 • Equipment, data, licenses, etc.
 • Will any of it cost anything? Where will the money
   come from?
 • If you’re not sure, find out now!
                                                             19
Planning research – work plan
 • Aim: to create a detailed research plan which lists work packages,
   specifies the amount of time required for each and assigns a time
   slot to each
 • Divide tasks into related groups (work packages, WPs); write
   short descriptive summary for each WP
 • Estimate time/effort each WP will take (person days or
   weeks) – always add contingency!
 • Establish partial order of WPs – which WP requires other
   WPs to have been completed?
 • Create a calendar diagram where each WP is assigned a slot –
   don’t forget to allow for other commitments


                                                                   20
Planning research – writing a research proposal

 • Aim: to put the results of your planning work into prose that will
   convince people that your planned research is of quality and
   worthwhile
 • Actually very time-consuming!
 • Sections:
     – Synopsis – ―executive summary‖
     – Introduction – motivate your research, why is it needed?
     – Aims, outcomes, and objectives
     – Related research – compare and contrast with existing work
     – Methodology – describe what you’re going to do, clarifying what is
       new, and where you’re going to use existing resources and ideas
     – Research in wider context – beneficiaries, dissemination, marketability,
       etc.
     – Work plan
     – Bibliography
                                                                             21
Planning research – writing a research proposal

 • Give it to different people to read – for:
    – Grammar/style: does it read well?
    – Clarity: are your aims and plans clear?
    – Quality: is this a good idea?
 • Look at general advice on academic writing




                                                  22
Writing a literature review
 • Literature review can build on, but goes
   beyond, research planning
 • More in-depth reading/understanding than you
   need for project proposal
 • Typically part of reports of completed
   research (Masters and PhD theses, project
   reports etc.)
 • Or publications in their own right: survey
   articles in journals or as book chapters
                                              23
Writing a literature review
 • Aim: to thoroughly review a given area of research,
   mentioning all important relevant research
 • Two basic forms:
    – Survey/state-of-the-art: balanced overview of given area of
      research; inclusion and space reflect importance of work
      in field; keep opinion to minimum
    – Project-specific review: inclusion and space reflect
      relevance to project; lead up to justification and
      motivation for project; opinion is important part of review
 • Important difference: for surveys, you don’t need to
   understand in detail how techniques work, but if it’s
   relevant to your project you do need to
                                                               24
Examples of survey-type literature reviews
 • Emotional language generation:
   http://www.itri.brighton.ac.uk/~Anja
   .Belz/Publications/ITRI-03-21.pdf
 • Speech technology (1994):
   http://doi.acm.org/10.1145/175247.17
   5252




                                      25
Writing a literature review – reading
 • Identify relevant publications
 • Determine status/importance (e.g. use Google
   Scholar to check on citations) – you can’t read all
   relevant publications
 • Use star-ratings to reflect importance
 • Read at different depths
    ***: read article in depth
    **: skim article
    *: read abstract and conclusions
 • Build up annotated bibliography (title, publication
   details, summary of contents, your comments)
                                                         26
Writing a literature review – writing
 • Don’t just list names and contents of papers – that’s
   an annotated bibliography, not a literature review
 • Turn it into a story – tell the story of the field
 • How does it all fit together?
 • What are the subfields, developments,
   controversies?
 • What is the state of the art?
 • What are the hot topics at the moment (reflected in
   special themes and special sessions at conferences,
   one-off workshops, and special issues of journals)
                                                       27
Writing a literature review – writing
 • Give your review structure:
    – Introduce the field
    – Bring out commonalities and differences between
      approaches
    – Can key results be summarised in a table or a graph?
    – Comment on more/less successful approaches
    – Conclusion:
       • Survey: summarise the state of the art of the field
       • Project-specific: identify area(s) where more research is needed
    – Bibliography
    – Appendix: research groups, data resources, web links
                                                                            28
Part II – take-home points
 • Zoom in on your chosen area of research gradually
   (don’t start with academic papers):
    – Superficial reading of online material
    – Background reading of survey articles, text-books, etc.
    – Literature review of research area of appropriate size:
      academic papers in conference proceedings and journals,
      book chapters, etc.
 • Read at different depths:
    – Read most important papers carefully
    – Skim less important papers
    – Read abstracts and conclusions of least important papers
                                                                 29
Part II – take-home points
 • Plan every aspect of your research thoroughly
   and in detail
 • For a literature review, do not just list
   publications and contents – instead, tell a
   story!




                                               30
Exercise: first steps in research planning
 Aim: write a short review of the area of Machine Translation
   and prepare a brief presentation of it

 Steps:
 1. Look at p. 14, do online research, and fill in as many of
    the categories on p. 15 as you can;
 2. Write your findings up as a 1-page report;
 3. Prepare a presentation of your findings, about 5
    minutes in length;
 4. Deliver the presentation on Friday morning.

 Work in lab (W622) this afternoon, tomorrow morning
  and on Thursday; finish report and presentation for
  Friday.
                                                                31
References
• Stephen Weinberg’s Four Golden Lessons:
  http://www.nature.com/nature/journal/v426/n6965/full
  /426389a.html
• Types of Research in Computing Science, Software
  Engineering and Artificial Intelligence by Aaron
  Sloman:
  http://www.cs.bham.ac.uk/research/projects/cogaff/mi
  sc/cs-research.html
• Alan Bundy’s Researcher’s Bible:
  http://homepages.inf.ed.ac.uk/bundy/how-
  tos/resbible.html
• CMU’s Advice on Research and Writing:
  http://www.cs.cmu.edu/afs/cs.cmu.edu/user/mleone/web
  /how-to.html
                                                     32
References
• A Computer Scientist's Guide to Writing and Publishing
  Technical Articles by Paul Martin:
  http://www.mcs.vuw.ac.nz/comp/Publications/archive/C
  S-TR-95/CS-TR-95-4.pdf
• Why you can’t cite Wikipedia in my class by Neil Waters
  (2007): http://doi.acm.org/10.1145/1284621.1284635
• Robert Dale’s tips for presentations:
  http://www.nltg.bton.ac.uk/teaching/SWM43/dale-
  presentations.pdf
• Robert Dale’s advice on time management:
  http://www.nltg.bton.ac.uk/teaching/SWM43/dale-time-
  management.pdf

                                                           33
References
• Cooper, H. (1998). Synthesizing Research: A Guide for
  Literature Reviews. Main points summarised here:
  http://library.ucsc.edu/ref/howto/literaturereview.h
  tml




                                                          34

				
DOCUMENT INFO
Shared By:
Categories:
Stats:
views:5
posted:8/1/2010
language:English
pages:34