Feedback on Slides for Project Presentation

A few groups sought feedback on their slides for project presentation in the last week of classes. Comments on their slides follow.

G10 : Oracle SDO

1. 14 slides may be hard to cover in 10-12 minutes.
	Condense the information into about 8-9 slides.
   - Edit master slide to reduce space for title and increase
	space for information.
   - Many slide are quite sparse and can be combined.
	"Motivation", "Indexing and Tuning", "More Information", "Query Results"
	 have only 3-4 lines and should be merged with other slides.
	Two slides on approach can be merged.
	Similarly, example query and query result may be merged
	by reducing discussion of query details. It may be useful	
	to put SQL3/OGIS query in place of query detail for comparison.

2. Coverage of SDO data types and operators should
	include a comparison with OGIS data types and operators.
	List the commonality and differences. Are there OGIS
	data types not supported by SDO? Does SDO support any
	types/operators beyond OGIS?
	Revise two slides on "Facts" to provide the comparison.
	Use a table to condense comparison information to 1 slide.
3. Set of example queries is very small (i.e. 1 query).
	Did you run other queries? Consider putting a list of
	queries (english versions at least in slides, SQL versions
	may be put in writeup).
4. Compare the SQL/SDO queries with SQL3/OGIS queries.
	Which form is more intuitive to 8705 students?
5. THere are no picture or tables in the slides.
	Consider presenting the "approach" steps as a flow-chart
	diagram. Consider using table to compare OGIS/SQL3 with SDO.

Feedback on Project Draft (HW 4)

General Comments : A few comments apply to multiple papers. These comments are summarized below:
1. A. Consider formatting your papers and reports like the papers in the reading list. Group 2's draft has a format close to the desired one. Show page numbers, section numbers, figure numbers, table numbers in your paper. Consider using latex to compose your papers. It will check for many of the stylistics problems and help you focus on technical writing. It is easy to learn and we can get short tutorial from fellow graduate students. If you use HTML or MS Word instead of latex, be sure to use appropriate features to supportcross-reference, citation, numbering of sections/ figures etc.
2. B. Include a list of refernces at the end of papers. Every reference should identify authors, title, publication forum (publisher, conference/journal/book, volume, number, url, ...), date of publication etc. so that readers can locate them easily.
3. C. Cite the sources (entries in your list of references) near the text referring to relevant concepts/contributions. Clearly identify text quotations, figures or tables from sources.
4. D. Use figures and table to highlight key messages in your paper.
5. E. Avoid using white-paper-ish material from commercial websites and marketing literature. Focus on technical material from research conferences, journals, books, etc.
General Comments on Survey Papers
1. F. A common format for survey paper are those used by ACM Computing Survey. Follow similar formats for your papers. Provide a table of content and a list of figures and tables to help readers navigate the paper.
2. G. Use a classification diagram to show the overall grouping of the ideas in the literature. Explain the criteria used for classification. Define the classes and then provide general summary of ideas for papers belonging to each class. Do not provide summary of individual papers. It will also help you identify future work by looking for combinations not explored by any papers yet.
3. H. Recall that presentations on survey papers will include motivation, major problems in the area, key results, open problems, and key sources. Survey paper should clearly identify these components as well. Abstract, introduction and conclusion should highlight the major problems and key results.
General Comments on Reports on Implementation Projects
1. I. Recall that presentations on projects should follow the format of paper analysis. Candidate sections include motivation, problem definition, key issues and alternative ways of resolving those, related work and their limitations, your approach, validation, conclusions (key contributions), and future work (assumptions and potential extensions). Project report should clearly identify these components as well. Abstract, introduction and conclusion should highlight the major contributions.
2. J. It is important to identify the key design decisions, alternative ways of resolving those, chosen approach to resolve those and justificationof the chosen approach. Justification of chosen approach relates to validation and may use validation methodologies such as theoretical / experimental comparison with other approaches, case studies, examples etc.
G1 : Digital Map Accuray Assessment
1. Overall format of the writeup is close to the desired one. Few areas of improvement follow. Many references (e.g. 2, 3, 4 ) are incomplete. See the general comments about the references.
2. Literature survey is too narrow. Include more sources on definition of errors and accuracy. See Wei-Li or me for additional references.
3. Section 3 and 4 need to identify the key design decisions, alternative ways of resolving those, chosen approach and justification of chosen approach. See General comment J.
G2 : GML-QL
1. Well-written and well-formatted overall.
2. Section2 on GML and OGIS may be placed in an appendix.
3. Contributions of the paper are not highlighted. Is GML-QL a contribution of this project? If so, then it should presented in a section by iteself. Current section 4 has several paragraph on related work. It may be useful to split this section to separate GML_QL from others.
4. Section 5 seems to relate to the validation of GML-QL using examples and case studies. It would be nice to discuss a general comparison of GML-QL with another spatial query language (e.g. SQL3/OGIS) to bring the strengths and limitations.
5. It may also be useful to discuss a few alternate ways of incorporating OGIS types/operators in XML. Briefly explain why the Quilt is chosen over other XML query languages.
6. Does integration of OGIS data types and operators into Quilt raise any new issues for the designers of Quilt?
G3 : Survey of spatial data mining
1. Overall a good start. Paper is relatively comprehensive on clustering methods, association rules.
2. Fill in the sections on classification techniques.
3. Add citations to spatial association rules (Han).
4. Consider adding a section on categorization of the research literature. See general comments G above. Criteria may include patterns (e.g. associations, clusters, classification) as well as data types (e..g point, lines, polygons) and statistical models (e.g. point processes, lattice, geo-statistical or non-stationary), etc. It will also help you identify future work by looking for combinations not explored by any papers yet.
G4 : Conceptual data model for TIGER files
1. Useful preliminary conceptual data models of TIGER data files.
2. Address general comments J.
3. Group received comments on the conceptual data models in office hoours. Some of those follow.
4. Figure 3 may be revised as a inheritance hierarchy for "landmark" entity. All entity shown in figure 3 are subtypes of "landmark" entity. Graphical notation for inheritance / sub-type may be a triangle on an edge.
5. Figure 2 may several "Referenced By" relatioships. Some of these may be eliminated or renamed. Provisional street features and landmarks may be referenced by roads if they have street addresses or information. However, boundaries, hydrgraphy, railroads are not likely to be referenced by roads.
6. Figure 2 may not be connected if spatial relationships are left implicit. If you do not like like that, list spatial relationship explicitly. For example railroads cross roads.
7. Figure 1 needs more work. It appears to list all record types in TIGER files. In that case one should use a relational schema rather than an entity-relationship diagram.
8. It ould be sueful to try to map the records (e.g. relations) in Figure 1 to the entities and relationship in Figure 2. There is a specific translation procedure to convert ER diagrams to relational schema (see chapter 2 , supplementary book). This algorithm may help map the figure 1 and 2.
G5 : Spatial Association Rules
1. General format is okay. A detailed description of spatial association rule mining paper by Han et al.
2. Assuming it is a survey paper, authors should address general comment G. Section 3.1 seems to introduce a few criteria (e.g. image/thematic maps, assumptions, etc.) for categorizing the papers. However actual classification is missing.
3. Set of papers surveyed is rather small. Consider expanding the survey to include few more papers. You may expand the topic to mining of general association rules. There are many concepts (e.g. negative association, data-compression, randomized algorithms) in mining of general association rules which have not been explored in papers on spatial association rules.
4. Last reference is not complete. List publication forum.
G6 : Scientific Visualization
1. Strong literature survey. Has extensive literature search and a good classification of visualization techniques.
2. Design and implementation project should identify design decisions. Address general comment J.
3. Goals of the project (3.2) are well defined. However link to the actual work (section 3.3) is very weak.
4. The Java based tools seems to be useful for only a small set of queries on a specific dataset. It will be useful to explain the full potential of this tool beyond NASA dataset. Can it support roll-up, drill-down of other dataset along different aggregation hierarchy? Should it?
G7 :
G8 : Outlier Detection
1. Contribution seems to be from only member in the group. Contribution from other member needs to be clarified.
G9 : Web based spatial databases
1. Overall well-written paper. Broad survey of the topic with an adequate number of citations.
2. Sections 1, 2 and 3 look disjoint from each other. It will be useful to add a classification scheme (see general comment G) to glue those together.
3. Use a classification to identify areas not covered well by existing research. These may be listed as areas for possible future research.
4. Relate the literature survey to the topics covered in the course. For example, GML and semi-join over spatial relations discussed in chapter 5 of supplementary book may relate.
G10 : Oracle SDO
1. Overall a nice explorations of Oracle SDO library supporting spatial data types and operators. A few minor refinement will be make it even better.
2. Supply the answer generated by oracle to the queries listed in spatial analysis section. If possible provides maps of the data inputted in Oracle to help explin the answer.
3. Provide a short summary of "experiences". You may list simple comparison of ORacle facilities with those discussed in the course. You may also list the major hurdles in testing the SDO library as well as some nice things Oracle SDO provides that everyone in the course should know about.
G11 : Direction based queries
1. Good survey of literature on mobile objects, direction based queries and location based services. Identifies useful problems as well. Minor improvements are possible.
2. Identify the contributions clearly. It will help focus the paper.
3. List the new problems raised by the new application environments. Which of the problems are computationally hard? Focus on those.
4. Reorganize the paper into a traditional sequence of sections: application requirements, data models (conceptual, logical, physical), query languages and query processing, etc. This is a common organization for many books on databases.
5. Provide more links across sections. For example, how do the pictograms used in data modeling related to the algorithms for processing direction queries?
6. Provide more links to literature on mobile objects. How can your query processing algorithm take advantage of indexes on mobile objects?
G12 : Routing with turn restrictions
1. Interesting model to combine several different route quality metrics into a single weight for use by Dijktra's algorithm. Writeup is quite mature and can be refined a bit.
2. Paper formatting is informal. See general comment A.
3. Identify the contributions clearly. Is it simply the combination function ? If so, then there should some analysis of the combination function proposed inthe paper. Are there other ways of combining the components instead of linear combination ? Why did we choose linear combination of component weights ?
4. References - Several references are incomplete. See general comment B.
5. There are too few references to papers on routing. Consider looking at the papers cited in the supplementary book chapter 7 in the paper on hierarchical routing in the reading list.
6. Construction of dual graph: how does one transform the source and destination nodes to dual graph ? Do those become edges in dual graph? In that case, dijktra's algorithm can not be applied since it does route from an edge to another edge.
7. Construction of dual graph: A simple analysis of correspondence between Dual graph and original graph is needed to show the correctness of approach. Does the transformation of graph perserve optimality of routes? If yes, then a simple proof is required. If not, then the approach is in jeopardy.
8. Choice of routing algorithm (Dijktra's) needs to be justified. Can one use other algorithms, e.g. A* or hierarchical routing ?
9. Hint: There is a recent paper in ACMGIS 2001 on construction of dual graph to model turn restriction. You may use the proof from there. Drop by in office hours or ask Raju for a copy of the paper.
G13 : XML databases
1. Extensive literature survey. Nice categorization of the papers in the literature.
2. Problem statement section is really a statement of the focus or "Scope".
3. "Broad areas of integration" section proposes three classes for the papers in the literature. It will be useful to add some more discussion of the definition and relationship among these classes. A possibility is to use a process diagram to show the difference between publishing, querying and storage.
4. Section on XML querying is much larger than that on XML storage. Consider reducing details of dividing the section into multiple parts.
5. Format is a bit informal. Follow general comment 1 for formatting of the paper.

Feedback on Paper Analysis Slides

General Comments : A few comments apply to multiple papers. These comments are summarized below:
1. A. Review the format used by slides on Survey paper on spatial databases: accomplishments and research needs. These are available from the class notes page. Follow similar formats for your slides.
2. B. Use phrases instead of complete sentences. Each phrase should be 6 to 8 words. You may omit articles and reword phrases to reduce the number of words. Remember the slides are providing highlights and you would be able to fill in the details in the oral presentation.
3. C. Limit the amount of text on each slide to 6 to 8 phrases. Too much text forces the audience to read the text and prevents them from listening to you. Consider covering parts of the slides with too much text during presentation.
4. D. Put a slide on "outline" listing the major groups of slides. Know the number of slides in each group. Time your presentation to know the amount of time you spend in presenting the slides in each group. This will allow you to complete the presentationon time.
5. E. Problem statement slide should identify four things: inputs (what information is given), outputs (what solution is to be found ), objectives (how does one judge the quality of solution), and constraints (which assumptions can be made on inputs and environment). Problom definition should be stated in a way that allows solutions other than what authors have proposed. This is fundamental to understand and critique any paper without being biased by the assumptions/biases of the authors. Many research papers will have a section on problem statement identifying these components. Others will list assumptions in scope, experiment design, proofs of theorems and lemmas, proposed algortihms/methods etc. If this information is missing, you should try to come up with best guesses for these components.
6. F. Key concepts should be defined and explained. Consider using examples and figures in explaining the key concepts. Use examples from paper. In addition you may ad new examples. You may refer to figures in the papers by proving figure numbers and page numbers. Figures should be simple with few (e.g. half a dozen) components. Consider using 3 to 5 slides on key concepts since audience should understand these well to appreciate rest of your analysis.
7. G. A validation methodology allows researchers to back up their claims in an objective manner. It allows readers to reproduce the results and conclusions of the authors. This is a fundamental requirement for science and scientific methods. Common methodologies used in computer science research include the following:
  - Software protoypes and implementations for use by a user community. Papers often quote popularity of the system and testimonials from users. Example include first implementations of Unix, Ingres, mosaic, terraserver (paper 21), Geominer (paper 3) etc.
  - Experimental evaluation of design decisions using a prototype of a simulation model (e.g. queuing model, algebraic cost models). Example include papers 8, 14, etc. It is useful to provide some details of the experiment design, e.g. benchmark data/queries, fixed/variable parameters, candidates evaluated and simulation model.
  - Theoretical proofs of correctness, computational complexity, Examples include papers 10, 15 etc. Nature of the proof (e.g. contruction, contridiction, reduction...) may be discussed.
  - Examples, detailed examples, case studies of real usages are used for concepts where more formal validation may be difficult or too early to do. Data modeling papers often use this. Other examples include harvard case studies.
  - Surveys of experts and statistical analysis
8. Assumptions, Rewrite Today: It is good to add the these two slide to distinguish between the conclusions of authors and your conclusions.
G1: Cost Modeling of Queries Using R-trees
1. Motivation slide: missing.
2. Problem Statement: Too much text. In addition the lists of items under "Given", Find", "Objectives" are not correct. For query window should be part of "Given". "Find" should include the cost model and constraint should include that the cost model is an algebraic formular. Consider soming to office hours to discuss it.
3. Contributions: Merge slides on related work and contributions. Provide examples of things previous work could not do or did not do as well.
4. Key Concepts: The cost model use a lot of notation. Include a summary definition of symbols unless you can refer to a table in the original paper for this purpose. It is useful to explain the meaning of the symbols via examples.
5. Validation Methodology: Rename the experimental evaluation slide to be Validation. Add some detail of the experiment design, e.g. benchmark data/queries, fixed/variable parameters, candidates evaluated.
6. Assumptions: Missing slide
7. Rewrite Today: reasonable.
G2: Effect of Buffering on Performance of R-tree
1. Motivation slide: Add a general motivation slide stating the popularity of R-tree family in spatial databases along withe the lack of study of buffer management strategies for spatial indexes. Use first 3 slides on review of R-tree as backup since it may common knowledge for the 8705 audience.
2. Problem Statement: Add "Objectives" subsection to problem statement. Is it to study the effect of buffering on R-tree performance?
3. Contributions Quite good. Consider placing it after problem definition.
4. Key Concepts: The cost model use a lot of notation. It is goo that the slides refer to a table in the original paper to explain the meaning of the symbols. Consider providing a comparison with other cost model for queries on R-tree (see G1's paper).
5. Validation Methodology: Merge conclusions slides with validation and results. Conclusions slide seems to reproduce conclusion of the paper rather than your conclusions.
6. Assumptions, Rewrite Today: It is good to add the these two slide to distinguish between the conclusions of authors and your conclusions.
G3:
1. Motivation slide: Example - use a figure or a table from the paper. Huang Yan has a nice figure for example in a follow up paper. Hui can provide you with the figure as well.
2. Problem Statement: Essentially reproduced from the paper. Try to rephrase it your words. Otherwise simply refer to the paper with page number instead of duplicating it.
3. Contributions - okay.
4. Key Concepts - Do not copy algorithms from the paper. Simply refer to the page number in the paper. Instead focus on a simple description of the algorithm via an example execution trace and comparison with other algorithms (e.g. apriori) everyone might know.
5. Validation Methodology - Rename correctness/compleness slide to this. Comment on the methodology in addition to the results.
6. Assumptions - missing slides.
7. Rewrite Today - missing slides
G4: Modeling Geospatial Application ... UML:
1. Too much text on slides. Reduce to about 6 bullets per slide with each bullet using about 6 - 8 words.
2. Split "Problem Statement" slide into two section: motivation and problem statement. Motivation should list a few applications and needs (e.g. bullet 3). Problem statement needs to identify four components as described in common comments. For example: Given UML, Geomatics standards, Find a design for geospatial repository, ... Think about the criteria to evaluate the design and the assumptions made by the authors.
3. Contributions slide - reduce the length of bullets
4. Key concepts need more explanation. Consider adding examples. Can you use any figures from the paper? Expand to 3 slides. Cut down material on assumptions and other areas.
5. Validation methodoloy - Contents are off target. You are listing Methodology referrences. See common comments for choice of methodologies. Choose one from the list.
6. Assumptions - too many. Too much text. Reduce to one slide
7. Rewrite - reduce text
G5: Concurrency control
1. Motivation slide: Rename "Problem Statement" slide to "Motivation".
2. Problem Statement: missing. Need to state "Given" (e.g. a set of concurrent operation on an index), "Find" (e.g. a concurrency control technique), "Objectives" (e.g. serializability), and "Constraints" (e.g. operations are limited to search and insert).
3. Contributions: missing.
4. Key Concepts: Too many slides. Omit slide 12 since you can show the paper using the camera in the class. Simply refer to the table number and page number.
5. Validation Methodology: okay.
6. Assumptions: missing
7. Rewrite Today: Rename "future work" to "Rewrite Today". Ensure that you have more points than those listed in the last section of the paper.
G6: Pictograms and ERDs Group 6 received comments in the office hours.
G7: Pictograms and ERDs Group 6 received comments in the office hours.
G8: Processing object-oriented direction queries
1. Motivation slide: Rename the first two slides after outline as motivation.
2. Problem Statement: What is "Exercise" in this slides?
3. Contributions: List too long. Pick 2 or 3.
4. Key Concepts: okay.
5. Validation Methodology: Slide too long. Split into multiple slides.
6. Assumptions: okay.
7. Rewrite Today: Ensure that the list is not same as that listed in the last section of the paper.
G9: Hierarchical Routing Algorithms Use a typesetting systems (e.g. Powerpoint, MS Word, latex, ...) to provide professional look and feel to the slides.
1. Motivation slide: Rename "problem addressed" to "Motivation".
2. Problem Statement: Expand S3 for this purpose. Split the list in this slide into OBJECTIVES and CONSTRAINTS. Add GIVEN (e.g. roadmap which is a labelled graph with edge weights, start node, end node, memory size) and FIND (e.g. shortest path between start and end nodes).
3. Contributions
4. Key Concepts: S7 should name the figures and examples in the paper that will be used to explain the concepts.
5. Validation Methodology: Reasonable. Add the experiment design details as described in the common comments.
6. Assumptions: missing
7. Rewrite Today: Reasonable. Shortest the text to 6-7 words per phrase.
G10: SQL/SDA
1. Long sentences should be reduced to phrases with at most 6 to 7 words (e.g. Motivation slide: spatial analysis for decision making, web-oriented data analysis). Otherwise Audience will be reading your long sentences instead of listening to you. You would be there to speak out complete sentences. My old comments said this and you have addressed it well in Key Concepts slides. Try to do that for other slides. Problem statement, Major contributions, Assumptions, Future Work slides have to do this as well.
2. Motivation slide: Add specific examples, e.g. those on slide on Example of a SQL/SDA query. Another example for verbal presentation is choosing land parcels for locating a fast food restaurant. Criteria include close to highways, outside flood plains, close to families with young children, etc.
3. Problem statement slide is completely off target. Move the Inputs to motivation slide. What is needed here are Given, Find, Objectives, Constraints stated in a way that allows solutions (SQL/OGIS) other than what authors have proposed (e.g. SQL/SDA). This is fundamental to understand and critique any paper without being biased by the assumptions/biases of the authors. Examples of Given are spatial analysis computations and data. Examples of Find is a set of spatial data types/operators and statement syntax/grammar to represent spatial analysis and data. Examples of Constraints are relational databases, SQL, ... Examples of objective are expressability, ease of use, well-defined syntax and semantics, etc.
4. Key concept slides are okay. There are a few long sentences here and there and should be replaced by short phrases.
5. Future WOrk - Rename as Rewrite today. It is not clear if this slide has any ideas from you in addition to those listed by authors. I was hoping that you would compare this work with SQL3/OGIS and comment on specific areas missing in SQL/SDA relative to SQL3/OGIS. It is also useful to provide specific examples for "complex mathematical operations" and carefully discuss how far should SQL go towards that. Clearly implementing entire GIS package in SQL is going too far.
6. ----------
7. Text on the each slide uses long senetences. Use shorter phrases as discussed in general comments.
8. Motivation slide: missing.
9. Problem Statement: Decompose into "Given", "Find", "Objective", and "Constraints".
10. Contributions: okay. Although you may reconsider some of those given the issues in rewrite today.
11. Key Concepts: Skip client-server architecture. List the spatial data types and operations and compare those to OGIS. If there are interesting issues in implementation, e.g. dividing work between GIS and RDBMS , then consider talking about thsoe.
12. Validation Methodology: missing.
13. Assumptions: missing.
14. Rewrite Today: Rename "Related Work in the Future" to "Rewrite Today". Include a brief comparison of SQL92 (SQL2) and SQL99 (SQL3) since this paper seems to use an older standard. Should one use OGIS instead of their data types and operators?
G11:
1. Motivation slide: Problem statement slide may be renamed as Motivation. May bring other slide (e.g. interesting facts) from the end in this area.
2. Problem Statement: missing. Need a slide identifying "Given", "Find", "Objectives", "Constraints".
3. Contributions: good. May mention that Terraserver is among most popular public spatial datawarehouse of satellite imagery on the web. It represented one of the largest dataset managed by MS SQL server.
4. Key Concepts: Need more details. Use 4 - 5 slides referring to figures in the paper to explain the concepts. You are teaching while presenting the key concepts.
5. Validation Methodology: "Implementation" is a better choice than "Experimental". for the methodology. They did not conduct experiments to evaluate alternative methods for resolving a design choice. May bring other slides (e.g. Encountered problems, Tuning Tips) from the end in this section.
6. Assumptions: It may useful to compare the spatial datatypes and spatial operations supperted by Terraserver with those in OGIS model (e.g. chapter 3 of supplementary book).
7. Rewrite Today: Missing. For example, should one use an object relational database supporting OGIS?
G12: The link does (http://www.me.umn.edu/sreeram/8705) not seem to work. Error message - http 404 Not Found. Please send the correct url at your earliest convenience.
1. Motivation slide:
2. Problem Statement:
3. Contributions
4. Key Concepts
5. Validation Methodology:
6. Assumptions
7. Rewrite Today
G13: Comments coming soon.
1. Motivation slide: missing. Identify applications. Why EM in SQL ? In addition use parts of Conclusions and "Why EM in SQL" slides up here.
2. Problem Statement: A good start. It is a bit verbose. Reduce to 7-8 phrase total, each with 5-7 words for the entire slide. Input/Given should include a specification of EM algorithm. Find/Output is to the point. Objectives listed are off-target. Include efficiency, or I/O efficiency as objective. Delete first bullet under objectives and move the second bullet under object to Constraints in an abbreviated form.
3. Contributions: out of order. Bring it forward.
4. Key Concepts: Good though verbose descriptions of clustering, and EM algorithm. Reduce amount of text of the slides as well as the number of slides since these are not the contributions of this paper. Needs to provide examples and figure to explain the key concepts quickly. Move the slide on why EM in SQL to motivation. Also need to provide more details of "Specifics" slide which relates to the contribution of this paper. Expand this material to provide brief explanation for "Optimizations" explaining why those make sense. You may choose to explain a few optimization or groups of optimizations.
5. Validation Methodology: missing.
6. Assumptions: missing
7. Rewrite Today; missing