Ezi Ozoani
try fix
386b032
raw
history blame
22.4 kB
# Copyright 2021 The HuggingFace Team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import statistics
import json
import pandas as pd
import seaborn as sns
import streamlit as st
from st_aggrid import AgGrid, GridOptionsBuilder
from .dataset_utils import HF_DESC_FIELD, HF_FEATURE_FIELD, HF_LABEL_FIELD
st.set_option('deprecation.showPyplotGlobalUse', False)
json_file_path = "cache_dir/has_cache.json"
with open(json_file_path, "r", encoding="utf-8") as j:
HAS_CACHE = json.loads(j.read())
def sidebar_header():
st.sidebar.markdown(
"""
This demo showcases the [dataset measures as we develop them](https://huggingface.co/blog/data-measurements-tool).
Right now this has a few pre-loaded datasets for which you can:
- view some general statistics about the text vocabulary, lengths, labels
- explore some distributional statistics to assess properties of the language
- view some comparison statistics and overview of the text distribution
The tool is in development, and will keep growing in utility and functionality 🤗🚧
""",
unsafe_allow_html=True,
)
def sidebar_selection(ds_name_to_dict, column_id):
# ds_names = list(ds_name_to_dict.keys())
ds_names = list(_HAS_CACHE.keys())
with st.sidebar.expander(f"Choose dataset and field {column_id}", expanded=True):
# choose a dataset to analyze
ds_name = st.selectbox(
f"Choose dataset to explore{column_id}:",
ds_names,
index=ds_names.index("hate_speech18"),
)
# choose a config to analyze
ds_configs = ds_name_to_dict[ds_name]
if ds_name == "c4":
config_names = ['en','en.noblocklist','realnewslike']
else:
config_names = list(ds_configs.keys())
config_names = list(_HAS_CACHE[ds_name].keys())
config_name = st.selectbox(
f"Choose configuration{column_id}:",
config_names,
index=0,
)
# choose a subset of num_examples
# TODO: Handling for multiple text features
ds_config = ds_configs[config_name]
# text_features = ds_config[HF_FEATURE_FIELD]["string"]
text_features = [tuple(text_field.split('-')) for text_field in _HAS_CACHE[ds_name][config_name]]
# TODO @yacine: Explain what this is doing and why eg tp[0] could = "id"
text_field = st.selectbox(
f"Which text feature from the{column_id} dataset would you like to analyze?",
[("text",)]
if ds_name == "c4"
else [tp for tp in text_features if tp[0] != "id"],
)
# Choose a split and dataset size
# avail_splits = list(ds_config["splits"].keys())
avail_splits = list(_HAS_CACHE[ds_name][config_name]['-'.join(text_field)].keys())
# 12.Nov note: Removing "test" because those should not be examined
# without discussion of pros and cons, which we haven't done yet.
if "test" in avail_splits:
avail_splits.remove("test")
split = st.selectbox(
f"Which split from the{column_id} dataset would you like to analyze?",
avail_splits,
index=0,
)
label_field, label_names = (
ds_name_to_dict[ds_name][config_name][HF_FEATURE_FIELD][HF_LABEL_FIELD][0]
if len(
ds_name_to_dict[ds_name][config_name][HF_FEATURE_FIELD][HF_LABEL_FIELD]
)
> 0
else ((), [])
)
return {
"dset_name": ds_name,
"dset_config": config_name,
"split_name": split,
"text_field": text_field,
"label_field": label_field,
"label_names": label_names,
}
def expander_header(dstats, ds_name_to_dict, column_id):
with st.expander(f"Dataset Description{column_id}"):
st.markdown(
ds_name_to_dict[dstats.dset_name][dstats.dset_config][HF_DESC_FIELD]
)
st.dataframe(dstats.dset_peek)
def expander_general_stats(dstats, column_id):
with st.expander(f"General Text Statistics{column_id}"):
st.caption(
"Use this widget to check whether the terms you see most represented"
" in the dataset make sense for the goals of the dataset."
)
if dstats.total_words == 0:
st.markdown("Eh oh...not finding the file I need. 😭 Probably it will be there soon. 🤞 Check back later!")
else:
st.markdown("There are {0} total words".format(str(dstats.total_words)))
st.markdown(
"There are {0} words after removing closed "
"class words".format(str(dstats.total_open_words))
)
st.markdown(
"The most common "
"[open class words](https://dictionary.apa.org/open-class-words) "
"and their counts are: "
)
st.dataframe(dstats.sorted_top_vocab_df)
st.markdown(
"There are {0} missing values in the dataset.".format(
str(dstats.text_nan_count)
)
)
if dstats.dedup_total > 0:
st.markdown(
"There are {0} duplicate items in the dataset. "
"For more information about the duplicates, "
"click the 'Duplicates' tab below.".format(str(dstats.dedup_total))
)
else:
st.markdown("There are 0 duplicate items in the dataset. ")
### Show the label distribution from the datasets
def expander_label_distribution(fig_labels, column_id):
with st.expander(f"Label Distribution{column_id}", expanded=False):
st.caption(
"Use this widget to see how balanced the labels in your dataset are."
)
if fig_labels is not None:
st.plotly_chart(fig_labels, use_container_width=True)
else:
st.markdown("No labels were found in the dataset")
def expander_text_lengths(dstats, column_id):
_TEXT_LENGTH_CAPTION = (
"Use this widget to identify outliers, particularly suspiciously long outliers."
)
with st.expander(f"Text Lengths{column_id}", expanded=False):
st.caption(_TEXT_LENGTH_CAPTION)
st.markdown(
"Below, you can see how the lengths of the text instances in your dataset are distributed."
)
st.markdown(
"Any unexpected peaks or valleys in the distribution may help to identify instances you want to remove or augment."
)
st.markdown(
"### Here is the relative frequency of different text lengths in your dataset:"
)
try:
st.image(dstats.fig_tok_length_png)
except:
st.pyplot(dstats.fig_tok_length, use_container_width=True)
st.markdown(
"The average length of text instances is **"
+ str(dstats.avg_length)
+ " words**, with a standard deviation of **"
+ str(dstats.std_length)
+ "**."
)
# This is quite a large file and is breaking our ability to navigate the app development.
# Just passing if it's not already there for launch v0
if dstats.length_df is not None:
start_id_show_lengths = st.selectbox(
"Show examples of length:",
sorted(dstats.length_df["length"].unique().tolist()),
key=f"select_show_length_{column_id}",
)
st.table(
dstats.length_df[
dstats.length_df["length"] == start_id_show_lengths
].set_index("length")
)
### Third, use a sentence embedding model
def expander_text_embeddings(
text_dset, fig_tree, node_list, embeddings, text_field, column_id
):
with st.expander(f"Text Embedding Clusters{column_id}", expanded=False):
_EMBEDDINGS_CAPTION = """
### Hierarchical Clustering of Text Fields
Taking in the diversity of text represented in a dataset can be
challenging when it is made up of hundreds to thousands of sentences.
Grouping these text items based on a measure of similarity can help
users gain some insights into their distribution.
The following figure shows a hierarchical clustering of the text fields
in the dataset based on a
[Sentence-Transformer](https://hf.co/sentence-transformers/all-mpnet-base-v2)
model. Clusters are merged if any of the embeddings in cluster A has a
dot product with any of the embeddings or with the centroid of cluster B
higher than a threshold (one threshold per level, from 0.5 to 0.95).
To explore the clusters, you can:
- hover over a node to see the 5 most representative examples (deduplicated)
- enter an example in the text box below to see which clusters it is most similar to
- select a cluster by ID to show all of its examples
"""
st.markdown(_EMBEDDINGS_CAPTION)
st.plotly_chart(fig_tree, use_container_width=True)
st.markdown("---\n")
if st.checkbox(
label="Enter text to see nearest clusters",
key=f"search_clusters_{column_id}",
):
compare_example = st.text_area(
label="Enter some text here to see which of the clusters in the dataset it is closest to",
key=f"search_cluster_input_{column_id}",
)
if compare_example != "":
paths_to_leaves = embeddings.cached_clusters.get(
compare_example,
embeddings.find_cluster_beam(compare_example, beam_size=50),
)
clusters_intro = ""
if paths_to_leaves[0][1] < 0.3:
clusters_intro += (
"**Warning: no close clusters found (best score <0.3). **"
)
clusters_intro += "The closest clusters to the text entered aboce are:"
st.markdown(clusters_intro)
for path, score in paths_to_leaves[:5]:
example = text_dset[
node_list[path[-1]]["sorted_examples_centroid"][0][0]
][text_field][:256]
st.write(
f"Cluster {path[-1]:5d} | Score: {score:.3f} \n Example: {example}"
)
show_node_default = paths_to_leaves[0][0][-1]
else:
show_node_default = len(node_list) // 2
else:
show_node_default = len(node_list) // 2
st.markdown("---\n")
if text_dset is None:
st.markdown("Missing source text to show, check back later!")
else:
show_node = st.selectbox(
f"Choose a leaf node to explore in the{column_id} dataset:",
range(len(node_list)),
index=show_node_default,
)
node = node_list[show_node]
start_id = st.slider(
f"Show closest sentences in cluster to the centroid{column_id} starting at index:",
0,
len(node["sorted_examples_centroid"]) - 5,
value=0,
step=5,
)
for sid, sim in node["sorted_examples_centroid"][start_id : start_id + 5]:
# only show the first 4 lines and the first 10000 characters
show_text = text_dset[sid][text_field][:10000]
show_text = "\n".join(show_text.split("\n")[:4])
st.text(f"{sim:.3f} \t {show_text}")
### Then, show duplicates
def expander_text_duplicates(dstats, column_id):
# TODO: Saving/loading figure
with st.expander(f"Text Duplicates{column_id}", expanded=False):
st.caption(
"Use this widget to identify text strings that appear more than once."
)
st.markdown(
"A model's training and testing may be negatively affected by unwarranted duplicates ([Lee et al., 2021](https://arxiv.org/abs/2107.06499))."
)
st.markdown("------")
st.write(
"### Here is the list of all the duplicated items and their counts in your dataset:"
)
if dstats.dup_counts_df is None or dstats.dup_counts_df.empty:
st.write("There are no duplicates in this dataset! 🥳")
else:
st.dataframe(dstats.dup_counts_df.reset_index(drop=True))
def expander_npmi_description(min_vocab):
_NPMI_CAPTION = (
"Use this widget to identify problematic biases and stereotypes in your data."
)
_NPMI_CAPTION1 = """
nPMI scores for a word help to identify potentially
problematic associations, ranked by how close the association is."""
_NPMI_CAPTION2 = """
nPMI bias scores for paired words help to identify how word
associations are skewed between the selected selected words
([Aka et al., 2021](https://arxiv.org/abs/2103.03417)).
"""
st.caption(_NPMI_CAPTION)
st.markdown(_NPMI_CAPTION1)
st.markdown(_NPMI_CAPTION2)
st.markdown(" ")
st.markdown(
"You can select from gender and sexual orientation "
"identity terms that appear in the dataset at least %s "
"times." % min_vocab
)
st.markdown(
"The resulting ranked words are those that co-occur with both "
"identity terms. "
)
st.markdown(
"The more *positive* the score, the more associated the word is with the first identity term. "
"The more *negative* the score, the more associated the word is with the second identity term."
)
### Finally, show Zipf stuff
def expander_zipf(z, zipf_fig, column_id):
with st.expander(
f"Vocabulary Distribution{column_id}: Zipf's Law Fit", expanded=False
):
try:
_ZIPF_CAPTION = """This shows how close the observed language is to an ideal
natural language distribution following [Zipf's law](https://en.wikipedia.org/wiki/Zipf%27s_law),
calculated by minimizing the [Kolmogorov-Smirnov (KS) statistic](https://en.wikipedia.org/wiki/Kolmogorov%E2%80%93Smirnov_test)."""
powerlaw_eq = r"""p(x) \propto x^{- \alpha}"""
zipf_summary = (
"The optimal alpha based on this dataset is: **"
+ str(round(z.alpha, 2))
+ "**, with a KS distance of: **"
+ str(round(z.distance, 2))
)
zipf_summary += (
"**. This was fit with a minimum rank value of: **"
+ str(int(z.xmin))
+ "**, which is the optimal rank *beyond which* the scaling regime of the power law fits best."
)
alpha_warning = "Your alpha value is a bit on the high side, which means that the distribution over words in this dataset is a bit unnatural. This could be due to non-language items throughout the dataset."
xmin_warning = "The minimum rank for this fit is a bit on the high side, which means that the frequencies of your most common words aren't distributed as would be expected by Zipf's law."
fit_results_table = pd.DataFrame.from_dict(
{
r"Alpha:": [str("%.2f" % z.alpha)],
"KS distance:": [str("%.2f" % z.distance)],
"Min rank:": [str("%s" % int(z.xmin))],
},
columns=["Results"],
orient="index",
)
fit_results_table.index.name = column_id
st.caption(
"Use this widget for the counts of different words in your dataset, measuring the difference between the observed count and the expected count under Zipf's law."
)
st.markdown(_ZIPF_CAPTION)
st.write(
"""
A Zipfian distribution follows the power law: $p(x) \propto x^{-α}$
with an ideal α value of 1."""
)
st.markdown(
"In general, an alpha greater than 2 or a minimum rank greater than 10 (take with a grain of salt) means that your distribution is relativaly _unnatural_ for natural language. This can be a sign of mixed artefacts in the dataset, such as HTML markup."
)
st.markdown(
"Below, you can see the counts of each word in your dataset vs. the expected number of counts following a Zipfian distribution."
)
st.markdown("-----")
st.write("### Here is your dataset's Zipf results:")
st.dataframe(fit_results_table)
st.write(zipf_summary)
# TODO: Nice UI version of the content in the comments.
# st.markdown("\nThe KS test p-value is < %.2f" % z.ks_test.pvalue)
# if z.ks_test.pvalue < 0.01:
# st.markdown(
# "\n Great news! Your data fits a powerlaw with a minimum KS " "distance of %.4f" % z.distance)
# else:
# st.markdown("\n Sadly, your data does not fit a powerlaw. =(")
# st.markdown("Checking the goodness of fit of our observed distribution")
# st.markdown("to the hypothesized power law distribution")
# st.markdown("using a Kolmogorov–Smirnov (KS) test.")
st.plotly_chart(zipf_fig, use_container_width=True)
if z.alpha > 2:
st.markdown(alpha_warning)
if z.xmin > 5:
st.markdown(xmin_warning)
except:
st.write("Under construction! 😱 🚧")
### Finally finally finally, show nPMI stuff.
def npmi_widget(npmi_stats, min_vocab, column_id):
"""
Part of the main app, but uses a user interaction so pulled out as its own f'n.
:param use_cache:
:param column_id:
:param npmi_stats:
:param min_vocab:
:return:
"""
with st.expander(f"Word Association{column_id}: nPMI", expanded=False):
try:
if len(npmi_stats.available_terms) > 0:
expander_npmi_description(min_vocab)
st.markdown("-----")
term1 = st.selectbox(
f"What is the first term you want to select?{column_id}",
npmi_stats.available_terms,
)
term2 = st.selectbox(
f"What is the second term you want to select?{column_id}",
reversed(npmi_stats.available_terms),
)
# We calculate/grab nPMI data based on a canonical (alphabetic)
# subgroup ordering.
subgroup_pair = sorted([term1, term2])
try:
joint_npmi_df = npmi_stats.load_or_prepare_joint_npmi(subgroup_pair)
npmi_show(joint_npmi_df)
except KeyError:
st.markdown(
"**WARNING!** The nPMI for these terms has not been pre-computed, please re-run caching."
)
else:
st.markdown(
"No words found co-occurring with both of the selected identity terms."
)
except:
st.write("Under construction! 😱 🚧")
def npmi_show(paired_results):
if paired_results.empty:
st.markdown("No words that co-occur enough times for results! Or there's a 🐛. Or we're still computing this one. 🤷")
else:
s = pd.DataFrame(paired_results.sort_values(by="npmi-bias", ascending=True))
# s.columns=pd.MultiIndex.from_arrays([['npmi','npmi','npmi','count', 'count'],['bias','man','straight','man','straight']])
s.index.name = "word"
npmi_cols = s.filter(like="npmi").columns
count_cols = s.filter(like="count").columns
if s.shape[0] > 10000:
bias_thres = max(abs(s["npmi-bias"][5000]), abs(s["npmi-bias"][-5000]))
print(f"filtering with bias threshold: {bias_thres}")
s_filtered = s[s["npmi-bias"].abs() > bias_thres]
else:
s_filtered = s
# TODO: This is very different look than the duplicates table above. Should probably standardize.
cm = sns.palplot(sns.diverging_palette(270, 36, s=99, l=48, n=16))
out_df = (
s_filtered.style.background_gradient(subset=npmi_cols, cmap=cm)
.format(subset=npmi_cols, formatter="{:,.3f}")
.format(subset=count_cols, formatter=int)
.set_properties(
subset=count_cols, **{"width": "10em", "text-align": "center"}
)
.set_properties(**{"align": "center"})
.set_caption(
"nPMI scores and co-occurence counts between the selected identity terms and the words they both co-occur with"
)
) # s = pd.read_excel("output.xlsx", index_col="word")
st.write("### Here is your dataset's nPMI results:")
st.dataframe(out_df)
### Dumping unused functions here for now
### Second, show the distribution of text perplexities
def expander_text_perplexities(text_label_df, sorted_sents_loss, fig_loss):
with st.expander("Show text perplexities A", expanded=False):
st.markdown("### Text perplexities A")
st.plotly_chart(fig_loss, use_container_width=True)
start_id_show_loss = st.slider(
"Show highest perplexity sentences in A starting at index:",
0,
text_label_df.shape[0] - 5,
value=0,
step=5,
)
for lss, sent in sorted_sents_loss[start_id_show_loss : start_id_show_loss + 5]:
st.text(f"{lss:.3f} {sent}")