rrobbes
commented
8 years ago Title of chapter
Natural Language Processing is no free lunch either
URL to the chapter
https://github.com/ds4se/chapters/blob/master/wagnerst/text-mining.md
Message?
What is the chapter's clear and approachable take away message?
Natural language processing, while appealing, will not yield satisfying results unless specific steps are followed.
Accessible?
- Is the chapters written for a generalist audience (no excessive use of technical terminology) with a minimum of diagrams and references? How can it be made more accessible to generalist? *
Since the chapter's focus is not on very technical content, it is very accessible. The only technical terminology used is very basic (stemming, clustering, clones, etc), which can be pretty much taken for granted.
The only thing (and I'll come to that later), is that some concepts are touched upon quite fast, and could be expanded upon by some examples.
Size?
Is the chapter the right length?
The chapter is approximately the right length, so no issue there. Perhaps it's a bit shorter than the limit.
Should anything missing be added?
While I like the four good practices to follow, I think they are perhaps too high level. I wonder if it would be a good idea to make them more specific. E.g., in the case of stemming, the recommendation boils down to "use stemming". To make it more concrete is there a particular stemming algorithm that should be used (or a list of say 3)?
Likewise, I really miss a concrete example of how clustering can help checking for the level of abstraction. Should this clustering be done based on word frequencies, on something else? This is important since the chapter explicitly comes back to this recommendation in the conclusion.
The same applies to the manual analysis of the data: more concrete examples would make the recommendation much more useful, and would help readers getting started more easily. If anything, it seems to me this recommendation could take more space in the chapter, should space be needed, as it is the most insightful (or so it seems to me).
Basically the chapter gives good pointers to these topics, but in some cases it's not clear what to expect from the reference based on just reading the chapter text, which makes it less likely that people will read the reference.
Can anything superfluous be removed (e.g. by deleting some section that does not work so well or by using less jargon, less formulae, lees diagrams, less references).?
Not really. Perhaps the introduction could be streamlined a bit. The recommendation to use stemming is perhaps really basic, so in the event that more space is needed, it can be shortened.
What are the aspects of the chapter that authors SHOULD change?
As said above, expanding on some of the good practices by means of more concrete recommendations or examples.
Gotta Mantra?
We encouraged (but did not require) the chapter title to be a mantra or something cute/catchy, i.e., some slogan reflecting best practice for data science for SE? If you have suggestion for a better title, please put them here.
The title is catchy enough. Although since the main recommendation to me is the manual analysis, perhaps putting it in the title could help.
Best Points
What are the best points of the chapter that the authors should NOT change?
The good practices presented in the chapter are useful, especially the last one. And the initial example drives the point that "just applying" NLP out of the box gives something far from the initial goal that was envisioned by the researchers.
Minor comment:
Perhaps an additional good reference would be Dave Binkley's paper at ICPC 2014 on LDA: David Binkley, Daniel Heinz, Dawn J. Lawrie, Justin Overfelt: Understanding LDA in source code analysis. ICPC 2014: 26-36
It has good discussions of the parameters of LDA and their impact on the results.
There's a typo in the second reference: "... In In Christian Bird, ... " => "... In Christian Bird, ...
timm
tzimmermsr
wagnerst
After review, relabel to 'reviewTwo'. After second review, relabel to 'EditorsComment'.