Data Science and Society End Term Summary

Data Science & Society
End Term Summary


End Term Design 2

Example Questions 2

Week 5 4
Assignment: Statistics in R (part 2) 4

Week 6 6
Lecture: Natural Language Processing 6
Lecture: NLP Embeddings 11
Assignment: NLP part 1 24
Literature: Computational Linguistics and NLP (Clark et al., 2013) 27

Week 7 40
Lecture: Automated Machine Learning 40
Lecture: Cloud Computing & Spot Pricing 51
ADD MONDAY Assignment: NLP Part 2 56
Literature: AutoML (Hutter et al., 2019) 58

Week 9 62
Lecture: Trends in Data Science & Society 62

,End Term Design
50 questions workshop assignments
Mostly NLP
R/Spark
AutoML
50 questions on literature and lectures
25 questions on NLP
15 questions on AutoML
10 questions on cloud computing



Example Questions
Complete this script by selecting the appropriate command at ____
A. Match
B. Fit (correct)
C. Solve
D. Map

How do contemporary neutral networks generally implement language models?
A. They treat each different word (or at least each different lemma) as a distinct atomic category
B. They use a single hidden layer at each position in a sequence
C. They exploit similarities between words by training feature-based representations of them (correct)
D. They consist of two components, one which models the similarities between words and one which
models the individual probabilities

What does this figure visualise?
A. Linguistic Regression Model
B. Projective Dependency Grammar
C. Social Network Analysis
D. Multi-layered Perception (correct)




A. CFG
B. PCFG (correct)
C. CNF
D. PPDG

Which of the following phrase structure statements (between quotes) is correct?
A. “They” is an NP (noun phrase) (correct)
B. “The garden with flowers” is a PP (prepositional phrase)
C. “Below sea level fish thrive.” is a VP (verb phrase)
2

, D. “Come on!” is a GP (GP is not a thing)

What is Parsing in NLP?
A. The algorithms to automatically...
B. The process of automatically analyzing a given sentence to determine underlying syntactic structures
(correct)
C. ..
D. ..

What is k-means?
A. Clustering Algorithm (correct)
B. Meaning Abstraction Algorithm
C. Document embedding algorithm
D. …

Which Python package did we use to retrieve PubMed abstracts?
A. ..
B. ..
C. Biopython (correct)
D. ..




3

, Week 5
Assignment: Statistics in R (part 2)
In this tutorial, you will learn several things about the popular statistical program R http://www.r-project.com
and about how to perform some of the machine learning possibilities of R and Spark.

Code


install.packages("tidyverse")



Explanation

There are many ways to do data analysis in R. An especially easy way to do your “data
wrangling” is by using the so-called Tidyverse. Since dplyr is not only a function, but a
grammar of data manipulation, the same grammar can work regardless of whether you are
using R directly, or use R to provide Spark with commands.


Code


library(sparklyr)

spark_install(version = "2.1.0")

sc <- spark_connect(master = "local")




Explanation

To use R and Spark, we will make use of a package called sparklyr. If Spark is running, it is
possible to connect to the Spark instance using the function spark_connect(). However, if you
use a local installation, you can install Spark through R, and connect to it immediately


Code


library(nycflights13)

flights = na.omit(flights)

flights_tbl <- copy_to(sc, flights, "flights", overwrite = TRUE)



Explanation


4