TxT360 / curated.py
omkarenator's picture
initial commit
e137e27
raw
history blame
10.3 kB
from fasthtml.common import *
from fasthtml.components import *
from plotly import graph_objects as go
from fh_plotly import plotly2fasthtml
import pandas as pd
import json
from data_viewer import view_data, gen_random_id
from rich import print
import uuid
data_sources = [
"Freelaw",
"Wikipedia",
"PhilPapers",
"Arxiv",
"S2ORC",
"S2ORC Abstract",
"Pubmed",
"USPTO",
"Hackernews",
"Ubuntu IRC",
"StackExchange",
"DM Maths",
"PG19",
"Europarl",
]
def get_data(data_source: str = "Freelaw", doc_id: int = 3, target: str = "foo"):
doc_id = max(0, min(int(doc_id), 9))
if data_source == "Freelaw":
raw_sample_doc = json.load(open("data/curated_samples/freelaw_raw.json"))
extracted_sample_doc = json.load(
open("data/curated_samples/freelaw_extract.json")
)
elif data_source == "Wikipedia":
raw_sample_doc = extracted_sample_doc = json.load(
open("data/curated_samples/wiki.json")
)
elif data_source == "StackExchange":
raw_sample_doc = json.load(open("data/curated_samples/stackexchange_raw.json"))
extracted_sample_doc = json.load(
open("data/curated_samples/stackexchange_extract.json")
)
elif data_source == "PhilPapers":
raw_sample_doc = extracted_sample_doc = json.load(
open("data/curated_samples/philpapers_raw.json")
)
elif data_source == "Arxiv":
raw_sample_doc = json.load(open("data/curated_samples/arxiv_raw.json"))
extracted_sample_doc = json.load(
open("data/curated_samples/arxiv_extract.json")
)
elif data_source == "S2ORC":
raw_sample_doc = extracted_sample_doc = json.load(
open("data/curated_samples/s2orc_raw.json")
)
elif data_source == "S2ORC Abstract":
raw_sample_doc = extracted_sample_doc = json.load(
open("data/curated_samples/s2orc_abstract_raw.json")
)
elif data_source == "Pubmed":
raw_sample_doc = json.load(open("data/curated_samples/pubmed_raw.json"))
extracted_sample_doc = json.load(
open("data/curated_samples/pubmed_extract.json")
)
elif data_source == "DM Maths":
raw_sample_doc = json.load(open("data/curated_samples/dm_maths_raw.json"))
extracted_sample_doc = json.load(
open("data/curated_samples/dm_maths_extract.json")
)
elif data_source == "PG19":
raw_sample_doc = extracted_sample_doc = json.load(
open("data/curated_samples/pg19_raw.json")
)
elif data_source == "Europarl":
raw_sample_doc = extracted_sample_doc = json.load(
open("data/curated_samples/europarl_raw.json")
)
else:
raw_sample_doc = extracted_sample_doc = [{} for _ in range(10)]
raw_json = raw_sample_doc[doc_id]
extracted_json = extracted_sample_doc[doc_id]
return view_data(
raw_json,
extracted_json,
doc_id=doc_id,
data_source=data_source,
data_sources=data_sources,
target=target,
)
def get_chart_28168342():
fig = go.Figure()
filter_names = [
"Download",
"Language",
"Min word count",
"Title Abstract",
"Majority language",
"Paragraph count",
"Frequency",
"Unigram log probability",
"Local dedup",
]
data_sources = [
("Wikipedia", [100, 90, 80, 70, 60, 50, 40, 30, 20]),
("Freelaw", [100, 90, 80, 70, 60, 50, 40, 20, 20]),
("DM Maths", [100, 90, 80, 70, 60, 40, 40, 30, 20]),
("USPTO", [100, 90, 80, 70, 60, 40, 40, 30, 20]),
("PG19", [100, 90, 80, 70, 60, 40, 40, 30, 20]),
("Hackernews", [100, 90, 80, 70, 60, 40, 40, 30, 20]),
("Ubuntu IRC", [100, 90, 80, 70, 60, 40, 40, 30, 20]),
("Europarl", [100, 90, 80, 70, 60, 40, 40, 30, 20]),
("StackExchange", [100, 90, 80, 70, 60, 40, 40, 30, 20]),
("Arxiv", [100, 90, 80, 70, 60, 40, 40, 30, 20]),
("S2ORC", [100, 90, 80, 70, 60, 40, 40, 30, 20]),
("S2ORC Abstract", [100, 90, 80, 70, 60, 40, 40, 30, 20]),
("PubMed Central", [100, 90, 80, 70, 60, 40, 40, 30, 20]),
("PubMed Central Abstract", [100, 90, 80, 70, 60, 40, 40, 30, 20]),
("PhilPapers", [100, 90, 80, 70, 60, 40, 40, 30, 20]),
]
for name, x_values in data_sources:
fig.add_trace(
go.Funnel(
name=name,
orientation="h",
y=filter_names,
x=x_values,
textinfo="value+percent total",
textposition="inside",
)
)
fig.update_layout(height=500, plot_bgcolor="rgba(0,0,0,0)")
return fig
def update(target: str, request):
params = request.query_params
if data_source := params.get(f"data_source_{target}"):
return get_data(
data_source, params.get(f"doc_id_{target}", 3), target)
if doc_id := params.get(f"doc_id_{target}"):
return get_data(
params.get(f"data_source_{target}"), doc_id, target)
def curated(request):
data_preparation_steps = pd.DataFrame(
{
"Method": [
"HTTP/FTP dumps",
"Web crawling",
"Archive snapshot",
"Generated",
"Curated",
],
"Description": [
"Acquiring data from HTTP/FTP dumps",
"Crawling websites to extract data",
"Working with archive dumps",
"Generating synthetic data",
"High quality curated data",
],
"Source": [
"Freelaw | Wikipedia | PhilPapers | Arxiv | S2ORC | Pubmeds",
"USPTO | Hackernews | Ubuntu IRC",
"StackExchange",
"DM Maths",
"PG19 | Europarl",
],
}
)
table_html = data_preparation_steps.to_html(index=False, border=0)
table_div = Div(NotStr(table_html), style="margin: 40px;")
text = P("""This initial stage serves as the foundation for the entire
process. Here, we focus on acquiring and extracting the raw data, which can
come from various sources such as crawling websites, using HTTP/FTP dumps,
or working with archive dumps. For instance, to download and prepare a
dataset, we can specific downloaders based on the data source. Each dataset
might have its own downloader script which can be updated in real time to
handle changes in the data source. Here is a general outline of the data
preparation process: It's worth noting that some pipelines might require
invoking additional functions or scripts to handle specific data sources or
formats. These helper scripts can be located within specific directories
or modules dedicated to the dataset.""")
data_preparation_div = Div(
H3("Data Preparation"),
text,
table_div,
Div(
get_data(target=gen_random_id()),
style="border: 1px solid #ccc; padding: 20px;",
),
)
text = P("""Data preprocessing is a crucial step in the data science
pipeline. It involves cleaning and transforming raw data into a format that
is suitable for analysis. This process includes handling missing values,
normalizing data, encoding categorical variables, and more.""")
preprocessing_steps = pd.DataFrame(
{
"Step": [
"Language Filter",
"Min Word Count",
"Title Abstract",
"Majority Language",
"Paragraph Count",
"Frequency",
"Unigram Log Probability",
],
"Description": [
"Filtering data based on language",
"Setting a minimum word count threshold",
"Extracting information from the title and abstract",
"Identifying the majority language in the dataset",
"Counting the number of paragraphs in each document",
"Calculating the frequency of each word in the dataset",
"Calculating the log probability of each unigram",
],
"Need": [
"To remove documents in unwanted languages",
"To filter out documents with very few words",
"To extract relevant information for analysis",
"To understand the distribution of languages in the dataset",
"To analyze the structure and length of documents",
"To identify important words in the dataset",
"To measure the significance of individual words",
],
"Pros": [
"Improves data quality by removing irrelevant documents",
"Filters out low-quality or incomplete documents",
"Provides additional information for analysis",
"Enables language-specific analysis and insights",
"Helps understand the complexity and content of documents",
"Identifies important terms and topics in the dataset",
"Quantifies the importance of individual words",
],
"Cons": [
"May exclude documents in less common languages",
"May remove documents with valuable information",
"May introduce bias in the analysis",
"May not accurately represent the language distribution",
"May not capture the complexity of document structure",
"May be sensitive to noise and outliers",
"May not capture the semantic meaning of words",
],
}
)
table_html = preprocessing_steps.to_html(index=False, border=0)
table_div = Div(NotStr(table_html), style="margin: 40px;")
data_preprocessing_div = Div(H3("Data Preprocessing"), text, table_div)
return Div(
Section(
H2("Curated Sources"),
plotly2fasthtml(get_chart_28168342()),
data_preparation_div,
data_preprocessing_div,
id="inner-text",
)
)