8f27082a54659c8f8f8be9226b7822b2

Review for team 8f27082a54659c8f8f8be9226b7822b2

Legend:
  + Positive
  ~ Neutral
  - Negative

Given a scale with quarter points I would award:
    Textual description:    5.25 / 6
    Code quality:               5.75 / 6
    Results:                        5.25 / 6

The grades given are based on the comments below. The negative and positive points are comments about either something lacking or which impressed me. A lack of comment on something is indicative that I considered it good or reasonably good.

Overall comments:
  + Clear and sufficient textual description.
  + Code comments are clear when present.
  + Good use of the libraries given.
  - Some areas are lacking code comments.
  - A few minor typos in some of the textual description though which do not reduce readability.

    Comments per section:

Section A:

A1:
  - For loops for dataframe operations doesn’t make the most of pandas' vectorization. Using Filter or some other provided method should speed up execution. (Later on you use apply which is good.)

A2:
  + Nice use of the in-place attribute for dataframe operations.
  + Good checking that the index is correct after modifications to it.

A3.1:
  + Use of a boxplot to show outliers is very appropriate.
  + Good explanations as to why the extra outlier shown in the boxplot graph is not considered.
  + Logical explanation of why the outliers considered are to be dropped.
  - The last outlier in the boxplot not being considered is not explained.

A3.2:
  + Printing unique year values give an exhaustive list of everything to be considered.
  - No code comments during this sub-exercise, code is still quite easy to follow but a comment or two on the parser would be appreciated.

A3.3
  - Not immediately clear as to why it is concluded that the papers from 2010 and 2011 are not referenced, a bit more textual description as to what is going on would help.
A3.4
  - The ‘.’ from the paper titles is not taken into account when comparing titles.
Good comments in the scraper.

A4
  - The ‘period’ comment and equivalent code conflict, the code has a +1 while the comment ignores it.
  - The division function checking for 0 is rendered redundant due to the +1.
  - Law not found or guessed.


Section B:

B1
  - Unconvincing argument as to the limitations of the naïve scores. The quality of papers or conferences is not discussed.

B2.1
  + Very good comments in this section.

B2.2
  ~ Perhaps use a bar graph for more readability rather than a long list.

B2.3
  + Once again, the comments in this section are very much sufficient.
  + Complete analysis of naïve vs H5 ranking.

B2.4
  + Great analysis as to the problems with naïve ranking.
  - The textual description with ‘about a 1000 ranks’ is referencing a difference of 2000-3000 ranks which can lead to confusion.

B2.5
  + Nice and clear, the potential limitations of the H5 index were found.
  ~ No code comments but the textual description is enough to know what is being done. However the actual implementation would be easier to follow with comments.

B3
  + The use of H5 per year is justified and the following analysis reflects the results found.

Section C:

C
  - The symbolic versus connectionist plots were asked to be in 2 different graphs.
  + The symbolic versus connectionist plots show the trends to be analysed very well.
  ~ When modifying a plot as the last line of a cell, you can for example write ‘_ = ax.set_title(...)’ to eliminate the automatic print of the return of the function.
  - Hard to follow exactly why authors per paper means the connectionist approach does not weaken.