dlab-berkeley / data-science-social-justice-2022 Goto Github PK

Is this still true or left over from Tom's organizational system? Maybe take out if it's deprecated?

"Our default working directory is wherever we launched the notebook - in our case the "Week 1" folder. We want to access the "Data" folder, which is two levels "up", inside of the main "DIGHUM160" directory."

(on local machine)

under the section incorporating time, this line results in an error: sns.set_theme()

this is error:
AttributeError: module 'seaborn' has no attribute 'set_theme'

my addition in bold:

One of the most famous datasets that allows us to start thinking about social justice is the Titanic dataset. It contains information of all the passengers aboard the RMS Titanic, which unfortunately was shipwrecked. This dataset can be used to predict whether a given passenger survived or not. Specifically, the titanic dataset allows us to explore the impact of a number of demographic attributes, such as socioeconomic status, gender, and age on the likelihood of surviving the wreck.

(ran on local machine, not datahub)

Under the section Phrase Modeling with gensim, this line results in an error: bigram = Phrases(tokens, min_count=2, threshold=3, delimiter='_')

the error is the following:
TypeError: sequence item 0: expected str instance, bytes found

• Use these word vectors to reflect on implicit binaries and normativities in your data;

Consider returning to implicit binaries/ normativities in the word vector section and contextualizing the results in light of these terms.

AttributeError: module 'seaborn' has no attribute 'set_theme'

line 1 sns.set_theme does not work. is this just an issue on my machine?

Under the section Similar Words, the kernel dies bc there's not enough RAM after running the previous lines followed by aita_tokens.similar('partner')

Under preprocessing all data, this line kills the kernel and makes it restart:
lemmas = [line for line in preprocess(df)]

for jupyterhub / datahub need under preprocessing all data, we need to pip install gensim otherwise ModuleNotFoundError: No module named 'gensim'

need to add the following lines to import pandas and spacy for the Binder version
!pip install pandas
!pip install spacy

In the example career/family target set, it looks like there part of another target set (math) that got combined in with them.

Are we expecting the students to perhaps run these notebooks on Google Colab? If so, we should include the option to commented out with the link to the data

calculate_biased_words(model, target1, target2, 4)

Since this is a custom helper function defined elsewhere, clarify the requirements for each of the inputs

Error in the spacy.load() line when running in Datahub:

OSError: [E050] Can't find model 'en_core_web_sm'. It doesn't seem to be a Python package or a valid path to a data directory.

When I was opening the Datahub link in the main branch, it exited with this error:

error: pathspec 'notebooks/module02/01_preprocessing.ipynb' did not match any file(s) known to git

Under implementing TF-IDF, this line results in an error: pd.DataFrame(X.toarray(), columns=vectorizer.get_feature_names_out())

this is the error message:
AttributeError: 'CountVectorizer' object has no attribute 'get_feature_names_out'

Error in following lines:
tsne = TSNE(init='pca', learning_rate='auto')
tsne_vectors = tsne.fit_transform(model.wv.vectors)

UFuncTypeError: ufunc 'multiply' did not contain a loop with signature matching types (dtype('<U4'), dtype('float32')) -> None

"We will only focus on for loops in this section, because while loops are less used."
When I was studying for technical interviews, I had to do a deep dive on while loops. I wonder if it may be useful to let students know that while loops are used in particular contexts, but are outside of the scope of the workshop?

I think the reflection part at the end of the notebook should be highlighted first under the Police Shootings Database. We need to carefully come up with a plan for how to approach that discussion and state our politics up front -- like we ,as scholars, as activists, and as an institution, support Black Lives Matter and recognize that police shootings disproportionately harm Black folks, and that cherry-picked data can be used to erase that fact.

At the very end of the notebook, "The program is complaining that it *"can't multiply sequence by non-int of type 'str'". What this means is that the `` operator is not defined for two strings."

Is the star (*) meant to bold something?

model.wv.most_similar(positive=['woman', 'king'], negative='man')

Key Error. Pretty sure the words in the example are not in the vocabulary of the trained model.

In the introduction text, it says we will use comments from the subreddit, but from what I can tell it looks like we are using the same post data from previous notebooks, rather than comments.

pd.read_csv('aita_sub_top_sm_lemmas.csv') error -- not found

FileNotFoundError: [Errno 2] File aita_sub_top_sm_lemmas.csv does not exist: 'aita_sub_top_sm_lemmas.csv'

I'm currently testing that all of the code runs in the datahub and with 8 GB of ram.

The kernel dies from too much memory being used when you run this line:
from sklearn.metrics.pairwise import cosine_similarity
similarities = cosine_similarity(tfidf)
similarities.shape

Although the biggest memory drain is a few cells before with the .to_dense() function

Word2Vec.load('aita.emb')

Looks like the '.emb' file is not available in Datahub/Binder