Spaces:
Runtime error
Runtime error
orionweller
commited on
Commit
•
bf3cbf2
1
Parent(s):
b742bcd
init commit
Browse files- app.py +277 -0
- diffs.jsonl +0 -0
- packages.txt +1 -0
- requirements.txt +7 -0
app.py
ADDED
@@ -0,0 +1,277 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import streamlit as st
|
2 |
+
import os
|
3 |
+
import pathlib
|
4 |
+
import pandas as pd
|
5 |
+
from collections import defaultdict
|
6 |
+
import json
|
7 |
+
import ast
|
8 |
+
import copy
|
9 |
+
import re
|
10 |
+
import tqdm
|
11 |
+
|
12 |
+
|
13 |
+
import pandas as pd
|
14 |
+
from collections import Counter
|
15 |
+
import string
|
16 |
+
import os
|
17 |
+
import streamlit as st
|
18 |
+
import difflib
|
19 |
+
from html import escape
|
20 |
+
|
21 |
+
|
22 |
+
def generate_diff_html_word_level(text1, text2):
|
23 |
+
"""
|
24 |
+
Generates word-level difference between text1 and text2 as HTML, correctly handling spaces.
|
25 |
+
"""
|
26 |
+
# Splitting texts into words
|
27 |
+
words1 = text1.split()
|
28 |
+
words2 = text2.split()
|
29 |
+
|
30 |
+
diff = []
|
31 |
+
matcher = difflib.SequenceMatcher(None, words1, words2)
|
32 |
+
|
33 |
+
for opcode in matcher.get_opcodes():
|
34 |
+
tag, i1, i2, j1, j2 = opcode
|
35 |
+
if tag == 'replace':
|
36 |
+
diff.append('<del style="background-color: #fbb6ce;">' + escape(' '.join(words1[i1:i2])) + '</del>')
|
37 |
+
diff.append('<ins style="background-color: #b7e4c7;">' + escape(' '.join(words2[j1:j2])) + '</ins>')
|
38 |
+
elif tag == 'delete':
|
39 |
+
diff.append('<del style="background-color: #fbb6ce;">' + escape(' '.join(words1[i1:i2])) + '</del>')
|
40 |
+
elif tag == 'insert':
|
41 |
+
diff.append('<ins style="background-color: #b7e4c7;">' + escape(' '.join(words2[j1:j2])) + '</ins>')
|
42 |
+
elif tag == 'equal':
|
43 |
+
diff.append(escape(' '.join(words1[i1:i2])))
|
44 |
+
|
45 |
+
# Construct final HTML string
|
46 |
+
final_html = ' '.join(diff).replace('</del> <ins', '</del> <ins')
|
47 |
+
return f'<pre style="white-space: pre-wrap;">{final_html}</pre>'
|
48 |
+
|
49 |
+
|
50 |
+
|
51 |
+
os.environ["PROTOCOL_BUFFERS_PYTHON_IMPLEMENTATION"] = "python"
|
52 |
+
st.set_page_config(layout="wide")
|
53 |
+
|
54 |
+
query_input = None
|
55 |
+
|
56 |
+
@st.cache_data
|
57 |
+
def convert_df(df):
|
58 |
+
# IMPORTANT: Cache the conversion to prevent computation on every rerun
|
59 |
+
return df.to_csv(path_or_buf=None, index=False, quotechar='"').encode('utf-8')
|
60 |
+
|
61 |
+
|
62 |
+
def get_current_data():
|
63 |
+
cur_query_data = []
|
64 |
+
for id_num, checkbox in current_checkboxes:
|
65 |
+
if checkbox:
|
66 |
+
qid, pid = id_num.split("-----")
|
67 |
+
cur_query_data.append({
|
68 |
+
"qid": qid,
|
69 |
+
"pid": pid,
|
70 |
+
"is_relevant": 0
|
71 |
+
})
|
72 |
+
return convert_df(pd.DataFrame(cur_query_data))
|
73 |
+
|
74 |
+
@st.cache_data
|
75 |
+
def escape_markdown(text):
|
76 |
+
# List of characters to escape
|
77 |
+
# Adding backslash to the list of special characters to escape itself as well
|
78 |
+
text = text.replace("``", "\"")
|
79 |
+
text = text.replace("$", "\$")
|
80 |
+
special_chars = ['\\', '`', '*', '_', '{', '}', '[', ']', '(', ')', '#', '+', '-', '.', '!', '|', "$"]
|
81 |
+
|
82 |
+
# Escaping each special character
|
83 |
+
escaped_text = "".join(f"\\{char}" if char in special_chars else char for char in text)
|
84 |
+
|
85 |
+
return escaped_text
|
86 |
+
|
87 |
+
|
88 |
+
|
89 |
+
if 'cur_instance_num' not in st.session_state:
|
90 |
+
st.session_state.cur_instance_num = -1
|
91 |
+
|
92 |
+
|
93 |
+
with st.sidebar:
|
94 |
+
st.title("Options")
|
95 |
+
|
96 |
+
|
97 |
+
@st.cache_data
|
98 |
+
def load_data():
|
99 |
+
data = []
|
100 |
+
with open("diffs.jsonl", "r") as f:
|
101 |
+
for line in f:
|
102 |
+
data.append(json.loads(line))
|
103 |
+
df = pd.DataFrame(data)
|
104 |
+
return df
|
105 |
+
|
106 |
+
|
107 |
+
df = load_data()
|
108 |
+
df["id_combined"] = df.apply(lambda x: str(x["qid"]), axis=1)
|
109 |
+
# aggregate the df so that each qid is a row and the rest is a list of instances
|
110 |
+
df = df.groupby("id_combined").agg(lambda x: x.tolist()).reset_index()
|
111 |
+
|
112 |
+
# print details about the diff type
|
113 |
+
# diff_types = df["diff_type"].value_counts()
|
114 |
+
# st.write(diff_types)
|
115 |
+
|
116 |
+
# # print how many titles differ by looping through each instances
|
117 |
+
# diff_titles = 0
|
118 |
+
# for index, row in df.iterrows():
|
119 |
+
# if row["old"] is not None and row["new"] is not None and row["old"][0]["title"] != row["new"][0]["title"]:
|
120 |
+
# diff_titles += 1
|
121 |
+
# st.write(f" Number of titles that differ: {diff_titles}")
|
122 |
+
|
123 |
+
|
124 |
+
original_map = {item["id_combined"]: item for item in df.to_dict(orient="records")}
|
125 |
+
|
126 |
+
|
127 |
+
col1, col2 = st.columns([1, 3], gap="large")
|
128 |
+
|
129 |
+
with st.sidebar:
|
130 |
+
st.success("All files uploaded")
|
131 |
+
|
132 |
+
with col1:
|
133 |
+
# breakpoint()
|
134 |
+
ids = df["id_combined"].tolist()
|
135 |
+
set_of_cols = set(ids)
|
136 |
+
container_for_nav = st.container()
|
137 |
+
name_of_columns = sorted([item for item in set_of_cols])
|
138 |
+
instances_to_use = name_of_columns
|
139 |
+
st.title("Instances")
|
140 |
+
|
141 |
+
def sync_from_drop():
|
142 |
+
if st.session_state.selectbox_instance == "Overview":
|
143 |
+
st.session_state.number_of_col = -1
|
144 |
+
st.session_state.cur_instance_num = -1
|
145 |
+
else:
|
146 |
+
index_of_obj = name_of_columns.index(st.session_state.selectbox_instance)
|
147 |
+
# print("Index of obj: ", index_of_obj, type(index_of_obj))
|
148 |
+
st.session_state.number_of_col = index_of_obj
|
149 |
+
st.session_state.cur_instance_num = index_of_obj
|
150 |
+
|
151 |
+
def sync_from_number():
|
152 |
+
st.session_state.cur_instance_num = st.session_state.number_of_col
|
153 |
+
# print("Session state number of col: ", st.session_state.number_of_col, type(st.session_state.number_of_col))
|
154 |
+
if st.session_state.number_of_col == -1:
|
155 |
+
st.session_state.selectbox_instance = "Overview"
|
156 |
+
else:
|
157 |
+
st.session_state.selectbox_instance = name_of_columns[st.session_state.number_of_col]
|
158 |
+
|
159 |
+
|
160 |
+
number_of_col = container_for_nav.number_input(min_value=-1, step=1, max_value=len(instances_to_use) - 1, on_change=sync_from_number, label=f"Select instance by index (up to **{len(instances_to_use) - 1}**)", key="number_of_col")
|
161 |
+
selectbox_instance = container_for_nav.selectbox("Select instance by ID", ["Overview"] + name_of_columns, on_change=sync_from_drop, key="selectbox_instance")
|
162 |
+
st.divider()
|
163 |
+
|
164 |
+
|
165 |
+
with col2:
|
166 |
+
# get instance number
|
167 |
+
inst_index = number_of_col
|
168 |
+
|
169 |
+
if inst_index >= 0:
|
170 |
+
inst_num = instances_to_use[inst_index]
|
171 |
+
|
172 |
+
st.markdown("<h1 style='text-align: center; color: black;text-decoration: underline;'>Editor</h1>", unsafe_allow_html=True)
|
173 |
+
|
174 |
+
current_checkboxes = []
|
175 |
+
|
176 |
+
diff_types = original_map[inst_num]['diff_type']
|
177 |
+
num_new = Counter(diff_types)["new"]
|
178 |
+
instances = original_map[inst_num]
|
179 |
+
|
180 |
+
container = st.container()
|
181 |
+
|
182 |
+
container.subheader(f"Combined ID: {inst_num} with {len(instances['pid'])} docs ({num_new} are new)")
|
183 |
+
# container.markdown(f"Diff Type: **{**")
|
184 |
+
|
185 |
+
container.divider()
|
186 |
+
|
187 |
+
container.subheader(f"Query Info")
|
188 |
+
|
189 |
+
container.markdown(f"Query ID: {instances['query_info'][0]['_id']}")
|
190 |
+
container.markdown(f"**Query**:{instances['query_info'][0]['text']}")
|
191 |
+
|
192 |
+
# container.markdown(f"Query Instruction OG: {original_map[inst_num]['query_info']['instruction_og']}")
|
193 |
+
# container.markdown(f"Query Instruction New: {original_map[inst_num]['query_info']['instruction_changed']}")
|
194 |
+
|
195 |
+
|
196 |
+
# container.subheader("Instruction ")
|
197 |
+
processed_diff = generate_diff_html_word_level(instances['query_info'][0]['instruction_og'], instances['query_info'][0]['instruction_changed'])
|
198 |
+
with container.container():
|
199 |
+
st.markdown("**Instruction**: " + processed_diff, unsafe_allow_html=True)
|
200 |
+
|
201 |
+
container.divider()
|
202 |
+
|
203 |
+
for i in range(len(instances["pid"])):
|
204 |
+
|
205 |
+
container.markdown(f"Doc {instances['pid'][i]}: **{diff_types[i] if diff_types[i] == 'new' else 'changed'}**")
|
206 |
+
|
207 |
+
# previous qrel score was either relevant (>0) or non-relevant
|
208 |
+
qrel_score = instances["qrel_score"][i]
|
209 |
+
# if relevant highlight blue, if non-relevant make orange with html tags
|
210 |
+
if qrel_score > 0:
|
211 |
+
container.markdown(f"<h3>Previous: <span style='color: blue;'>Relevant</span></h3>", unsafe_allow_html=True)
|
212 |
+
else:
|
213 |
+
container.markdown(f"<h3>Previous: <span style='color: orange;'>Not Relevant</span></h3>", unsafe_allow_html=True)
|
214 |
+
|
215 |
+
combined_id = str(instances["qid"][i]) + "-----" + instances["pid"][i]
|
216 |
+
container.subheader(f"Title")
|
217 |
+
|
218 |
+
if instances['old'][i] is not None and instances['new'][i] is not None and instances['old'][i][0]['title'] == instances['new'][i][0]['title']:
|
219 |
+
container.markdown(f"{instances['old'][i][0]['title']}")
|
220 |
+
elif instances['old'][i] is None and instances['new'][i] is None:
|
221 |
+
container.markdown("")
|
222 |
+
else:
|
223 |
+
if instances['old'][i] is not None:
|
224 |
+
container.markdown(f"{instances['old'][i][0]['title']}")
|
225 |
+
elif instances['new'][i] is not None:
|
226 |
+
container.markdown(f"{instances['new'][i][0]['title']}")
|
227 |
+
else:
|
228 |
+
assert False
|
229 |
+
|
230 |
+
|
231 |
+
if instances['old'][i] is not None and instances['new'][i] is not None:
|
232 |
+
container.subheader("Title Diff")
|
233 |
+
processed_diff = generate_diff_html_word_level(instances['old'][i][0]['title'], instances['new'][i][0]['title'])
|
234 |
+
with container.container():
|
235 |
+
st.markdown(processed_diff, unsafe_allow_html=True)
|
236 |
+
# if both are none, say that:
|
237 |
+
|
238 |
+
on = container.toggle('Show original text', key="toggle" + combined_id)
|
239 |
+
if on:
|
240 |
+
container.subheader(f"Original Text")
|
241 |
+
|
242 |
+
if instances['old'][i] is not None:
|
243 |
+
original_input = container.markdown(instances['old'][i][0]['text'])
|
244 |
+
else:
|
245 |
+
original_input = None
|
246 |
+
container.markdown("")
|
247 |
+
|
248 |
+
container.subheader(f"New Text")
|
249 |
+
# generated_input = container.markdown(instances['new'][i][0]['text'])
|
250 |
+
|
251 |
+
# # Diff
|
252 |
+
# if original_input is not None and generated_input is not None:
|
253 |
+
# container.subheader("Diff")
|
254 |
+
# processed_diff = generate_diff_html_word_level(instances['old'][i][0]['text'], instances['new'][i][0]['text'])
|
255 |
+
# with container.container():
|
256 |
+
# st.markdown(processed_diff, unsafe_allow_html=True)
|
257 |
+
|
258 |
+
# container.subheader("Full Doc")
|
259 |
+
container.markdown(instances['full_doc'][i])
|
260 |
+
|
261 |
+
current_checkboxes.append((combined_id, container.checkbox(f'{combined_id} is Non-Relevant', key=combined_id)))
|
262 |
+
|
263 |
+
|
264 |
+
container.divider()
|
265 |
+
|
266 |
+
# download the editable text and venue name
|
267 |
+
if st.checkbox("Download data as CSV"):
|
268 |
+
st.download_button(
|
269 |
+
label="Download data as CSV",
|
270 |
+
data=get_current_data(),
|
271 |
+
file_name=f'annotation_query_{inst_num}_double_check.csv',
|
272 |
+
mime='text/csv',
|
273 |
+
)
|
274 |
+
|
275 |
+
# none checked
|
276 |
+
elif inst_index < 0:
|
277 |
+
st.title("Overview")
|
diffs.jsonl
ADDED
The diff for this file is too large to render.
See raw diff
|
|
packages.txt
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
default-jre
|
requirements.txt
ADDED
@@ -0,0 +1,7 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
pandas==2.0.3
|
2 |
+
streamlit==1.33.0
|
3 |
+
plotly==5.15.0
|
4 |
+
protobuf==3.20.0
|
5 |
+
beautifulsoup4==4.12.2
|
6 |
+
nltk==3.7
|
7 |
+
tqdm==4.66.2
|