ceshine/pluto-shopee-2021-tokenized.jl Secret

## pluto-shopee-2021-tokenized.jl
### A Pluto.jl notebook ###
# v0.14.2

using Markdown
using InteractiveUtils

# ╔═╡ b5ad061a-a297-11eb-0308-91355c5c6d15
using DataFrames, CSV, WordTokenizers, Plots, Random, Statistics

# ╔═╡ 219ff84e-1bb6-4650-b9bc-c1e22a18aff1
md"## Read and Transform the Data

Data source: [Shopee - Price Match Guarantee](https://www.kaggle.com/c/shopee-product-matching/data)
"

# ╔═╡ 6c5e5252-0d4c-4422-ada4-377dbf87e9db
df = DataFrame(CSV.File("../data/train.csv"; select=[:title, :label_group]))

# ╔═╡ e342798e-e079-4cea-a0f3-482ac0f37c2d
md"Make all characters lower cased:"

# ╔═╡ ee90d5e9-e978-4ad2-a82b-ae4fc0a1abe7
df.title = map(x -> lowercase(x), df.title);

# ╔═╡ 2f6784e0-a1e8-4f77-b8b9-34f23797cc17
md"Group by label_group:"

# ╔═╡ 3d04e1cf-dbbf-479d-81dc-fd5d5e6a9b8c
groups = groupby(df, "label_group")

# ╔═╡ 0e66e630-4ed6-4b7d-938b-ed1a1c14daad
md"## Tokenizer demo

List of tokenize functions demonstrated here (from the `WordTokenizers` package):

1. `punctuation_space_tokenize`
2. `penn_tokenize`
3. `nltk_word_tokenize`
4. `poormans_tokenize`
"

# ╔═╡ 1598b104-0e27-45cc-8306-e8b68159362b
punctuation_space_tokenize("sprei lady rose 180x200 king terlaris keroppi")

# ╔═╡ e8032f5f-10b8-42f9-b072-07440d06f6f3
penn_tokenize("sprei lady rose 180x200 king terlaris keroppi")

# ╔═╡ 5bbc4ba9-9ed0-4c22-b934-7ac470ba6e12
punctuation_space_tokenize("sprei lady rose 180x200 king terlaris keroppi")

# ╔═╡ 3e14d636-ed5c-4a4a-9ecf-0a9bd463966d
nltk_word_tokenize("sprei lady rose 180x200 king terlaris keroppi")

# ╔═╡ c83c778f-baa8-4d2b-a8a9-70411eb64867
punctuation_space_tokenize("double tape 3m vhb 12 mm x 4,5 m original / double foam tape")

# ╔═╡ abd84ebb-a419-4bfa-856f-3a328251c506
poormans_tokenize("double tape 3m vhb 12 mm x 4,5 m original / double foam tape")

# ╔═╡ 1b64053e-375b-44a7-a037-d783f1e0c3a5
penn_tokenize("double tape 3m vhb 12 mm x 4,5 m original / double foam tape")

# ╔═╡ 6277476d-64a3-484f-bdf4-df1187d97412
nltk_word_tokenize("double tape 3m vhb 12 mm x 4,5 m original / double foam tape")

# ╔═╡ ecb1c961-6669-413b-b514-49aaa9743422
md"## Tokenize and Count"

# ╔═╡ d1e10f59-71e3-4326-87c3-50dc1dc9a729
function tokenize_and_count(groups::GroupedDataFrame, tokenize_func::Function)
	df_combo = DataFrame(
		text_1 = String[], text_2 = String[], label_group = Int64[],
		n_1 = Int64[], n_2 = Int64[], intersect = Int64[], union = Int64[]
	)
	for group in groups
		tokens = map(x -> Set(tokenize_func(x)), group[!, :title])
		for i in 1:(nrow(group)-1)
			for j in (i+1):nrow(group)
				push!(
					df_combo,
					(
						group[i, :title], group[j, :title], group[i, :label_group],
						length(tokens[i]), length(tokens[j]),
						length(∩(tokens[i], tokens[j])),
						length(∪(tokens[i], tokens[j]))
					)
				)
			end
		end
	end
	df_combo.jaccard = df_combo.intersect ./ (
		df_combo.n_1 + df_combo.n_2 - df_combo.intersect)
	df_combo.overlap = df_combo.intersect ./ min.(
		df_combo.n_1, df_combo.n_2)
	return df_combo
end

# ╔═╡ d8cb00a0-d3a6-42b2-b2d7-9fd9631b88e2
md"Compare results from the two tokenizers (nltk and penn):"

# ╔═╡ 8db610b0-f606-42d1-8e06-91f4b9e45a01
df_nltk = tokenize_and_count(groups, nltk_word_tokenize); df_nltk[1:3, :]

# ╔═╡ 3c0f90fe-512a-470d-8c1d-dbdfb062b4f8
df_penn = tokenize_and_count(groups, penn_tokenize); df_penn[1:3, :]

# ╔═╡ 841a2b3b-d353-4e12-906f-c3cfa3c180dd
md"## Histogram"

# ╔═╡ 00c78e55-ce31-41e9-acc6-f5fff37f569f
begin
	histogram(df_penn.jaccard, label="jac")
	title!("Jaccard Distance (Penn)")
end

# ╔═╡ d325f3ef-8bf0-4e39-aff8-3289d4956bd9
begin
	histogram(df_penn.overlap .* 100, label="overlap")
	title!("Overlap(%) (Penn)")
end

# ╔═╡ ea0eee30-213d-468c-a380-9062871e061f
begin
	histogram(df_nltk.jaccard, label="jac")
	title!("Jaccard Distance (NLTK)")
end

# ╔═╡ 5702777b-e3be-4c54-8f5c-f230f9e5d312
begin
	histogram(df_nltk.overlap .* 100, bins=20, label="overlap")
	title!("Overlap(%) (NLTK)")
end

# ╔═╡ 70e7a1d7-75af-4293-b17f-66e371f51bcf
begin
	histogram(
		combine(
			groupby(df_nltk, :label_group),
			:overlap => mean
		).overlap_mean .* 100, bins=20, label="overlap")
	title!("Mean Overlap(%) in Group (NLTK)")
end

# ╔═╡ f724ebaa-5ee1-4de8-bc52-a7105f8b51e1
begin
	histogram(
		combine(
			groupby(df_nltk, :label_group),
			:overlap => (x -> min(x...)) => :min
		).min .* 100, bins=20, label="overlap")
	title!("Min Overlap(%) in Group (NLTK)")
end

# ╔═╡ cb7af49b-6d06-435a-a616-57da1d3ac7c6
begin
	histogram(
		combine(
			groupby(df_nltk, :label_group),
			:overlap => (x -> max(x...)) => :max
		).max .* 100, bins=20, label="overlap")
	title!("Max Overlap(%) in Group (NLTK)")
end

# ╔═╡ b9f20397-d9a8-4fa4-ba48-87b820abb5ad
begin
	histogram(
		combine(
			groupby(df, :label_group),
			nrow
		).nrow, bins=50, label="n")
	title!("Label Group Sizes")
end

# ╔═╡ 7a77465e-aece-4472-b59c-a439c7b8fc1e
md"(# of groups with fewer than ten members, # of groups with ten or more members):"

# ╔═╡ be867d59-86c8-4b94-93c6-94a88c7628c3
begin
	df_group_size = combine(
		groupby(df, :label_group),
		nrow
	)
	(
		nrow(filter(:nrow => <(10), df_group_size)),
		nrow(filter(:nrow => >=(10), df_group_size))
	)
end

# ╔═╡ ade6a594-15b4-4c97-a76b-60ce3da85be3
md"Larger groups:"

# ╔═╡ 486ebf45-f33b-4ce9-85e6-c88338a14577
filter(
	:nrow => ==(max(df_group_size.nrow...)),
	df_group_size
)

# ╔═╡ c7cd4170-64d2-47c0-af5f-99486dee8c72
md"Pick one group as an example:"

# ╔═╡ 844405b4-ff74-4d7f-9c91-1873f3a88e9e
filter(:label_group => ==(3627744656), df)

# ╔═╡ 4fbf3f5b-8842-476f-a940-1f9ae8aed69f
md"## Samples"

# ╔═╡ 8047da13-9795-47ed-ae4d-4447aef085b4
function sample(df::DataFrame, cond, n::Int64)
	df_tmp = filter(cond, df)
	(nrow(df_tmp), df_tmp[shuffle(1:nrow(df_tmp))[1:n], :])
end

# ╔═╡ 6a6061c6-2b25-4395-816a-08844739e120
md"Jaccard index equals 1 (exactly the same):"

# ╔═╡ 964e971c-5e0b-4814-b7c8-ce72ef2be3d3
sample(df_nltk, :jaccard => x -> x == 1, 10)

# ╔═╡ b76a355d-e5c4-4529-89b0-81fbdd61e86f
md"Jaccard index equals 0 (completely different):"

# ╔═╡ 9bf1369d-45f2-420a-95ba-2462cab27631
sample(df_nltk, :jaccard => x -> x == 0, 10)

# ╔═╡ 47b2d64f-9a5a-4015-91c6-9294fd450126
md"Jaccard index between 0 and 0.2 (only slightly similar):"

# ╔═╡ f09c2239-5617-4c60-a774-572342f35e6e
sample(df_nltk, :jaccard => x -> (x > 0) & (x < 0.2), 10)

# ╔═╡ Cell order:
# ╠═b5ad061a-a297-11eb-0308-91355c5c6d15
# ╟─219ff84e-1bb6-4650-b9bc-c1e22a18aff1
# ╠═6c5e5252-0d4c-4422-ada4-377dbf87e9db
# ╟─e342798e-e079-4cea-a0f3-482ac0f37c2d
# ╠═ee90d5e9-e978-4ad2-a82b-ae4fc0a1abe7
# ╟─2f6784e0-a1e8-4f77-b8b9-34f23797cc17
# ╠═3d04e1cf-dbbf-479d-81dc-fd5d5e6a9b8c
# ╠═0e66e630-4ed6-4b7d-938b-ed1a1c14daad
# ╠═1598b104-0e27-45cc-8306-e8b68159362b
# ╠═e8032f5f-10b8-42f9-b072-07440d06f6f3
# ╠═5bbc4ba9-9ed0-4c22-b934-7ac470ba6e12
# ╠═3e14d636-ed5c-4a4a-9ecf-0a9bd463966d
# ╠═c83c778f-baa8-4d2b-a8a9-70411eb64867
# ╠═abd84ebb-a419-4bfa-856f-3a328251c506
# ╠═1b64053e-375b-44a7-a037-d783f1e0c3a5
# ╠═6277476d-64a3-484f-bdf4-df1187d97412
# ╟─ecb1c961-6669-413b-b514-49aaa9743422
# ╠═d1e10f59-71e3-4326-87c3-50dc1dc9a729
# ╠═d8cb00a0-d3a6-42b2-b2d7-9fd9631b88e2
# ╠═8db610b0-f606-42d1-8e06-91f4b9e45a01
# ╠═3c0f90fe-512a-470d-8c1d-dbdfb062b4f8
# ╟─841a2b3b-d353-4e12-906f-c3cfa3c180dd
# ╟─00c78e55-ce31-41e9-acc6-f5fff37f569f
# ╟─d325f3ef-8bf0-4e39-aff8-3289d4956bd9
# ╟─ea0eee30-213d-468c-a380-9062871e061f
# ╟─5702777b-e3be-4c54-8f5c-f230f9e5d312
# ╟─70e7a1d7-75af-4293-b17f-66e371f51bcf
# ╟─f724ebaa-5ee1-4de8-bc52-a7105f8b51e1
# ╟─cb7af49b-6d06-435a-a616-57da1d3ac7c6
# ╟─b9f20397-d9a8-4fa4-ba48-87b820abb5ad
# ╟─7a77465e-aece-4472-b59c-a439c7b8fc1e
# ╟─be867d59-86c8-4b94-93c6-94a88c7628c3
# ╟─ade6a594-15b4-4c97-a76b-60ce3da85be3
# ╟─486ebf45-f33b-4ce9-85e6-c88338a14577
# ╟─c7cd4170-64d2-47c0-af5f-99486dee8c72
# ╠═844405b4-ff74-4d7f-9c91-1873f3a88e9e
# ╟─4fbf3f5b-8842-476f-a940-1f9ae8aed69f
# ╠═8047da13-9795-47ed-ae4d-4447aef085b4
# ╟─6a6061c6-2b25-4395-816a-08844739e120
# ╠═964e971c-5e0b-4814-b7c8-ce72ef2be3d3
# ╟─b76a355d-e5c4-4529-89b0-81fbdd61e86f
# ╠═9bf1369d-45f2-420a-95ba-2462cab27631
# ╟─47b2d64f-9a5a-4015-91c6-9294fd450126
# ╠═f09c2239-5617-4c60-a774-572342f35e6e
	### A Pluto.jl notebook ###
	# v0.14.2

	using Markdown
	using InteractiveUtils

	# ╔═╡ b5ad061a-a297-11eb-0308-91355c5c6d15
	using DataFrames, CSV, WordTokenizers, Plots, Random, Statistics

	# ╔═╡ 219ff84e-1bb6-4650-b9bc-c1e22a18aff1
	md"## Read and Transform the Data

	Data source: [Shopee - Price Match Guarantee](https://www.kaggle.com/c/shopee-product-matching/data)
	"

	# ╔═╡ 6c5e5252-0d4c-4422-ada4-377dbf87e9db
	df = DataFrame(CSV.File("../data/train.csv"; select=[:title, :label_group]))

	# ╔═╡ e342798e-e079-4cea-a0f3-482ac0f37c2d
	md"Make all characters lower cased:"

	# ╔═╡ ee90d5e9-e978-4ad2-a82b-ae4fc0a1abe7
	df.title = map(x -> lowercase(x), df.title);

	# ╔═╡ 2f6784e0-a1e8-4f77-b8b9-34f23797cc17
	md"Group by label_group:"

	# ╔═╡ 3d04e1cf-dbbf-479d-81dc-fd5d5e6a9b8c
	groups = groupby(df, "label_group")

	# ╔═╡ 0e66e630-4ed6-4b7d-938b-ed1a1c14daad
	md"## Tokenizer demo

	List of tokenize functions demonstrated here (from the `WordTokenizers` package):

	1. `punctuation_space_tokenize`
	2. `penn_tokenize`
	3. `nltk_word_tokenize`
	4. `poormans_tokenize`
	"

	# ╔═╡ 1598b104-0e27-45cc-8306-e8b68159362b
	punctuation_space_tokenize("sprei lady rose 180x200 king terlaris keroppi")

	# ╔═╡ e8032f5f-10b8-42f9-b072-07440d06f6f3
	penn_tokenize("sprei lady rose 180x200 king terlaris keroppi")

	# ╔═╡ 5bbc4ba9-9ed0-4c22-b934-7ac470ba6e12
	punctuation_space_tokenize("sprei lady rose 180x200 king terlaris keroppi")

	# ╔═╡ 3e14d636-ed5c-4a4a-9ecf-0a9bd463966d
	nltk_word_tokenize("sprei lady rose 180x200 king terlaris keroppi")

	# ╔═╡ c83c778f-baa8-4d2b-a8a9-70411eb64867
	punctuation_space_tokenize("double tape 3m vhb 12 mm x 4,5 m original / double foam tape")

	# ╔═╡ abd84ebb-a419-4bfa-856f-3a328251c506
	poormans_tokenize("double tape 3m vhb 12 mm x 4,5 m original / double foam tape")

	# ╔═╡ 1b64053e-375b-44a7-a037-d783f1e0c3a5
	penn_tokenize("double tape 3m vhb 12 mm x 4,5 m original / double foam tape")

	# ╔═╡ 6277476d-64a3-484f-bdf4-df1187d97412
	nltk_word_tokenize("double tape 3m vhb 12 mm x 4,5 m original / double foam tape")

	# ╔═╡ ecb1c961-6669-413b-b514-49aaa9743422
	md"## Tokenize and Count"

	# ╔═╡ d1e10f59-71e3-4326-87c3-50dc1dc9a729
	function tokenize_and_count(groups::GroupedDataFrame, tokenize_func::Function)
	df_combo = DataFrame(
	text_1 = String[], text_2 = String[], label_group = Int64[],
	n_1 = Int64[], n_2 = Int64[], intersect = Int64[], union = Int64[]
	)
	for group in groups
	tokens = map(x -> Set(tokenize_func(x)), group[!, :title])
	for i in 1:(nrow(group)-1)
	for j in (i+1):nrow(group)
	push!(
	df_combo,
	(
	group[i, :title], group[j, :title], group[i, :label_group],
	length(tokens[i]), length(tokens[j]),
	length(∩(tokens[i], tokens[j])),
	length(∪(tokens[i], tokens[j]))
	)
	)
	end
	end
	end
	df_combo.jaccard = df_combo.intersect ./ (
	df_combo.n_1 + df_combo.n_2 - df_combo.intersect)
	df_combo.overlap = df_combo.intersect ./ min.(
	df_combo.n_1, df_combo.n_2)
	return df_combo
	end

	# ╔═╡ d8cb00a0-d3a6-42b2-b2d7-9fd9631b88e2
	md"Compare results from the two tokenizers (nltk and penn):"

	# ╔═╡ 8db610b0-f606-42d1-8e06-91f4b9e45a01
	df_nltk = tokenize_and_count(groups, nltk_word_tokenize); df_nltk[1:3, :]

	# ╔═╡ 3c0f90fe-512a-470d-8c1d-dbdfb062b4f8
	df_penn = tokenize_and_count(groups, penn_tokenize); df_penn[1:3, :]

	# ╔═╡ 841a2b3b-d353-4e12-906f-c3cfa3c180dd
	md"## Histogram"

	# ╔═╡ 00c78e55-ce31-41e9-acc6-f5fff37f569f
	begin
	histogram(df_penn.jaccard, label="jac")
	title!("Jaccard Distance (Penn)")
	end

	# ╔═╡ d325f3ef-8bf0-4e39-aff8-3289d4956bd9
	begin
	histogram(df_penn.overlap .* 100, label="overlap")
	title!("Overlap(%) (Penn)")
	end

	# ╔═╡ ea0eee30-213d-468c-a380-9062871e061f
	begin
	histogram(df_nltk.jaccard, label="jac")
	title!("Jaccard Distance (NLTK)")
	end

	# ╔═╡ 5702777b-e3be-4c54-8f5c-f230f9e5d312
	begin
	histogram(df_nltk.overlap .* 100, bins=20, label="overlap")
	title!("Overlap(%) (NLTK)")
	end

	# ╔═╡ 70e7a1d7-75af-4293-b17f-66e371f51bcf
	begin
	histogram(
	combine(
	groupby(df_nltk, :label_group),
	:overlap => mean
	).overlap_mean .* 100, bins=20, label="overlap")
	title!("Mean Overlap(%) in Group (NLTK)")
	end

	# ╔═╡ f724ebaa-5ee1-4de8-bc52-a7105f8b51e1
	begin
	histogram(
	combine(
	groupby(df_nltk, :label_group),
	:overlap => (x -> min(x...)) => :min
	).min .* 100, bins=20, label="overlap")
	title!("Min Overlap(%) in Group (NLTK)")
	end

	# ╔═╡ cb7af49b-6d06-435a-a616-57da1d3ac7c6
	begin
	histogram(
	combine(
	groupby(df_nltk, :label_group),
	:overlap => (x -> max(x...)) => :max
	).max .* 100, bins=20, label="overlap")
	title!("Max Overlap(%) in Group (NLTK)")
	end

	# ╔═╡ b9f20397-d9a8-4fa4-ba48-87b820abb5ad
	begin
	histogram(
	combine(
	groupby(df, :label_group),
	nrow
	).nrow, bins=50, label="n")
	title!("Label Group Sizes")
	end

	# ╔═╡ 7a77465e-aece-4472-b59c-a439c7b8fc1e
	md"(# of groups with fewer than ten members, # of groups with ten or more members):"

	# ╔═╡ be867d59-86c8-4b94-93c6-94a88c7628c3
	begin
	df_group_size = combine(
	groupby(df, :label_group),
	nrow
	)
	(
	nrow(filter(:nrow => <(10), df_group_size)),
	nrow(filter(:nrow => >=(10), df_group_size))
	)
	end

	# ╔═╡ ade6a594-15b4-4c97-a76b-60ce3da85be3
	md"Larger groups:"

	# ╔═╡ 486ebf45-f33b-4ce9-85e6-c88338a14577
	filter(
	:nrow => ==(max(df_group_size.nrow...)),
	df_group_size
	)

	# ╔═╡ c7cd4170-64d2-47c0-af5f-99486dee8c72
	md"Pick one group as an example:"

	# ╔═╡ 844405b4-ff74-4d7f-9c91-1873f3a88e9e
	filter(:label_group => ==(3627744656), df)

	# ╔═╡ 4fbf3f5b-8842-476f-a940-1f9ae8aed69f
	md"## Samples"

	# ╔═╡ 8047da13-9795-47ed-ae4d-4447aef085b4
	function sample(df::DataFrame, cond, n::Int64)
	df_tmp = filter(cond, df)
	(nrow(df_tmp), df_tmp[shuffle(1:nrow(df_tmp))[1:n], :])
	end

	# ╔═╡ 6a6061c6-2b25-4395-816a-08844739e120
	md"Jaccard index equals 1 (exactly the same):"

	# ╔═╡ 964e971c-5e0b-4814-b7c8-ce72ef2be3d3
	sample(df_nltk, :jaccard => x -> x == 1, 10)

	# ╔═╡ b76a355d-e5c4-4529-89b0-81fbdd61e86f
	md"Jaccard index equals 0 (completely different):"

	# ╔═╡ 9bf1369d-45f2-420a-95ba-2462cab27631
	sample(df_nltk, :jaccard => x -> x == 0, 10)

	# ╔═╡ 47b2d64f-9a5a-4015-91c6-9294fd450126
	md"Jaccard index between 0 and 0.2 (only slightly similar):"

	# ╔═╡ f09c2239-5617-4c60-a774-572342f35e6e
	sample(df_nltk, :jaccard => x -> (x > 0) & (x < 0.2), 10)

	# ╔═╡ Cell order:
	# ╠═b5ad061a-a297-11eb-0308-91355c5c6d15
	# ╟─219ff84e-1bb6-4650-b9bc-c1e22a18aff1
	# ╠═6c5e5252-0d4c-4422-ada4-377dbf87e9db
	# ╟─e342798e-e079-4cea-a0f3-482ac0f37c2d
	# ╠═ee90d5e9-e978-4ad2-a82b-ae4fc0a1abe7
	# ╟─2f6784e0-a1e8-4f77-b8b9-34f23797cc17
	# ╠═3d04e1cf-dbbf-479d-81dc-fd5d5e6a9b8c
	# ╠═0e66e630-4ed6-4b7d-938b-ed1a1c14daad
	# ╠═1598b104-0e27-45cc-8306-e8b68159362b
	# ╠═e8032f5f-10b8-42f9-b072-07440d06f6f3
	# ╠═5bbc4ba9-9ed0-4c22-b934-7ac470ba6e12
	# ╠═3e14d636-ed5c-4a4a-9ecf-0a9bd463966d
	# ╠═c83c778f-baa8-4d2b-a8a9-70411eb64867
	# ╠═abd84ebb-a419-4bfa-856f-3a328251c506
	# ╠═1b64053e-375b-44a7-a037-d783f1e0c3a5
	# ╠═6277476d-64a3-484f-bdf4-df1187d97412
	# ╟─ecb1c961-6669-413b-b514-49aaa9743422
	# ╠═d1e10f59-71e3-4326-87c3-50dc1dc9a729
	# ╠═d8cb00a0-d3a6-42b2-b2d7-9fd9631b88e2
	# ╠═8db610b0-f606-42d1-8e06-91f4b9e45a01
	# ╠═3c0f90fe-512a-470d-8c1d-dbdfb062b4f8
	# ╟─841a2b3b-d353-4e12-906f-c3cfa3c180dd
	# ╟─00c78e55-ce31-41e9-acc6-f5fff37f569f
	# ╟─d325f3ef-8bf0-4e39-aff8-3289d4956bd9
	# ╟─ea0eee30-213d-468c-a380-9062871e061f
	# ╟─5702777b-e3be-4c54-8f5c-f230f9e5d312
	# ╟─70e7a1d7-75af-4293-b17f-66e371f51bcf
	# ╟─f724ebaa-5ee1-4de8-bc52-a7105f8b51e1
	# ╟─cb7af49b-6d06-435a-a616-57da1d3ac7c6
	# ╟─b9f20397-d9a8-4fa4-ba48-87b820abb5ad
	# ╟─7a77465e-aece-4472-b59c-a439c7b8fc1e
	# ╟─be867d59-86c8-4b94-93c6-94a88c7628c3
	# ╟─ade6a594-15b4-4c97-a76b-60ce3da85be3
	# ╟─486ebf45-f33b-4ce9-85e6-c88338a14577
	# ╟─c7cd4170-64d2-47c0-af5f-99486dee8c72
	# ╠═844405b4-ff74-4d7f-9c91-1873f3a88e9e
	# ╟─4fbf3f5b-8842-476f-a940-1f9ae8aed69f
	# ╠═8047da13-9795-47ed-ae4d-4447aef085b4
	# ╟─6a6061c6-2b25-4395-816a-08844739e120
	# ╠═964e971c-5e0b-4814-b7c8-ce72ef2be3d3
	# ╟─b76a355d-e5c4-4529-89b0-81fbdd61e86f
	# ╠═9bf1369d-45f2-420a-95ba-2462cab27631
	# ╟─47b2d64f-9a5a-4015-91c6-9294fd450126
	# ╠═f09c2239-5617-4c60-a774-572342f35e6e