ROOTS search tool
Browse files- .gitattributes +1 -0
- Makefile +7 -0
- README.md +5 -4
- app.py +236 -0
- roots_search_tool_specs.pdf +3 -0
.gitattributes
CHANGED
@@ -31,3 +31,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
|
|
31 |
*.zip filter=lfs diff=lfs merge=lfs -text
|
32 |
*.zst filter=lfs diff=lfs merge=lfs -text
|
33 |
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
|
|
|
31 |
*.zip filter=lfs diff=lfs merge=lfs -text
|
32 |
*.zst filter=lfs diff=lfs merge=lfs -text
|
33 |
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
34 |
+
*.pdf filter=lfs diff=lfs merge=lfs -text
|
Makefile
ADDED
@@ -0,0 +1,7 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
.PHONY: style
|
2 |
+
|
3 |
+
# Format source code automatically
|
4 |
+
|
5 |
+
style:
|
6 |
+
black --line-length 119 --target-version py36 .
|
7 |
+
isort .
|
README.md
CHANGED
@@ -1,12 +1,13 @@
|
|
1 |
---
|
2 |
-
title: Roots Search
|
3 |
-
emoji:
|
4 |
-
colorFrom:
|
5 |
-
colorTo:
|
6 |
sdk: gradio
|
7 |
sdk_version: 3.7
|
8 |
app_file: app.py
|
9 |
pinned: false
|
|
|
10 |
---
|
11 |
|
12 |
Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
|
|
|
1 |
---
|
2 |
+
title: Roots Search Tool
|
3 |
+
emoji: πΈ π
|
4 |
+
colorFrom: green
|
5 |
+
colorTo: indigo
|
6 |
sdk: gradio
|
7 |
sdk_version: 3.7
|
8 |
app_file: app.py
|
9 |
pinned: false
|
10 |
+
license: apache-2.0
|
11 |
---
|
12 |
|
13 |
Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
|
app.py
ADDED
@@ -0,0 +1,236 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import http.client as http_client
|
2 |
+
import json
|
3 |
+
import logging
|
4 |
+
import os
|
5 |
+
import re
|
6 |
+
import string
|
7 |
+
|
8 |
+
import gradio as gr
|
9 |
+
import requests
|
10 |
+
|
11 |
+
|
12 |
+
def get_docid_html(docid):
|
13 |
+
data_org, dataset, docid = docid.split("/")
|
14 |
+
|
15 |
+
docid_html = """<a
|
16 |
+
class="underline-on-hover"
|
17 |
+
title="I am hovering over the text"
|
18 |
+
style="color:#2D31FA;"
|
19 |
+
href="https://huggingface.co/datasets/bigscience-data/{}"
|
20 |
+
target="_blank">{}</a><span style="color: #7978FF;">/{}</span>""".format(
|
21 |
+
dataset, data_org + "/" + dataset, docid
|
22 |
+
)
|
23 |
+
return docid_html
|
24 |
+
|
25 |
+
|
26 |
+
PII_TAGS = {"KEY", "EMAIL", "USER", "IP_ADDRESS", "ID", "IPv4", "IPv6"}
|
27 |
+
PII_PREFIX = "PI:"
|
28 |
+
|
29 |
+
|
30 |
+
def process_pii(text):
|
31 |
+
for tag in PII_TAGS:
|
32 |
+
text = text.replace(
|
33 |
+
PII_PREFIX + tag,
|
34 |
+
"""<b><mark style="background: Fuchsia; color: Lime;">REDACTED {}</mark></b>""".format(tag),
|
35 |
+
)
|
36 |
+
return text
|
37 |
+
|
38 |
+
|
39 |
+
def process_results(results, highlight_terms):
|
40 |
+
if len(results) == 0:
|
41 |
+
return """<br><p style='font-family: Arial; color:Silver; text-align: center;'>
|
42 |
+
No results retrieved.</p><br><hr>"""
|
43 |
+
|
44 |
+
results_html = ""
|
45 |
+
for result in results:
|
46 |
+
tokens = result["text"].split()
|
47 |
+
tokens_html = []
|
48 |
+
for token in tokens:
|
49 |
+
if token in highlight_terms:
|
50 |
+
tokens_html.append("<b>{}</b>".format(token))
|
51 |
+
else:
|
52 |
+
tokens_html.append(token)
|
53 |
+
tokens_html = " ".join(tokens_html)
|
54 |
+
tokens_html = process_pii(tokens_html)
|
55 |
+
meta_html = (
|
56 |
+
"""
|
57 |
+
<p class='underline-on-hover' style='font-size:12px; font-family: Arial; color:#585858; text-align: left;'>
|
58 |
+
<a href='{}' target='_blank'>{}</a></p>""".format(
|
59 |
+
result["meta"]["url"], result["meta"]["url"]
|
60 |
+
)
|
61 |
+
if "meta" in result and result["meta"] is not None and "url" in result["meta"]
|
62 |
+
else ""
|
63 |
+
)
|
64 |
+
docid_html = get_docid_html(result["docid"])
|
65 |
+
results_html += """{}
|
66 |
+
<p style='font-size:14px; font-family: Arial; color:#7978FF; text-align: left;'>Document ID: {}</p>
|
67 |
+
<p style='font-size:12px; font-family: Arial; color:MediumAquaMarine'>Language: {}</p>
|
68 |
+
<p style='font-family: Arial;'>{}</p>
|
69 |
+
<br>
|
70 |
+
""".format(
|
71 |
+
meta_html, docid_html, result["lang"], tokens_html
|
72 |
+
)
|
73 |
+
return results_html + "<hr>"
|
74 |
+
|
75 |
+
|
76 |
+
def scisearch(query, language, num_results=10):
|
77 |
+
try:
|
78 |
+
post_data = {"query": query, "k": num_results}
|
79 |
+
if language != "detect_language":
|
80 |
+
post_data["lang"] = language
|
81 |
+
|
82 |
+
output = requests.post(
|
83 |
+
os.environ.get("address"),
|
84 |
+
headers={"Content-type": "application/json"},
|
85 |
+
data=json.dumps(post_data),
|
86 |
+
timeout=60,
|
87 |
+
)
|
88 |
+
|
89 |
+
payload = json.loads(output.text)
|
90 |
+
|
91 |
+
if "err" in payload:
|
92 |
+
if payload["err"]["type"] == "unsupported_lang":
|
93 |
+
detected_lang = payload["err"]["meta"]["detected_lang"]
|
94 |
+
return f"""
|
95 |
+
<p style='font-size:18px; font-family: Arial; color:MediumVioletRed; text-align: center;'>
|
96 |
+
Detected language <b>{detected_lang}</b> is not supported.<br>
|
97 |
+
Please choose a language from the dropdown or type another query.
|
98 |
+
</p><br><hr><br>"""
|
99 |
+
|
100 |
+
results = payload["results"]
|
101 |
+
highlight_terms = payload["highlight_terms"]
|
102 |
+
|
103 |
+
if language == "detect_language":
|
104 |
+
return (
|
105 |
+
(
|
106 |
+
f"""<p style='font-family: Arial; color:MediumAquaMarine; text-align: center; line-height: 3em'>
|
107 |
+
Detected language: <b>{results[0]["lang"]}</b></p><br><hr><br>"""
|
108 |
+
if len(results) > 0 and language == "detect_language"
|
109 |
+
else ""
|
110 |
+
)
|
111 |
+
+ process_results(results, highlight_terms)
|
112 |
+
)
|
113 |
+
|
114 |
+
if language == "all":
|
115 |
+
results_html = ""
|
116 |
+
for lang, results_for_lang in results.items():
|
117 |
+
if len(results_for_lang) == 0:
|
118 |
+
results_html += f"""<p style='font-family: Arial; color:Silver; text-align: left; line-height: 3em'>
|
119 |
+
No results for language: <b>{lang}</b><hr></p>"""
|
120 |
+
continue
|
121 |
+
|
122 |
+
collapsible_results = f"""
|
123 |
+
<details>
|
124 |
+
<summary style='font-family: Arial; color:MediumAquaMarine; text-align: left; line-height: 3em'>
|
125 |
+
Results for language: <b>{lang}</b><hr>
|
126 |
+
</summary>
|
127 |
+
{process_results(results_for_lang, highlight_terms)}
|
128 |
+
</details>"""
|
129 |
+
results_html += collapsible_results
|
130 |
+
return results_html
|
131 |
+
|
132 |
+
return process_results(results, highlight_terms)
|
133 |
+
|
134 |
+
except Exception as e:
|
135 |
+
results_html = f"""
|
136 |
+
<p style='font-size:18px; font-family: Arial; color:MediumVioletRed; text-align: center;'>
|
137 |
+
Raised {type(e).__name__}</p>
|
138 |
+
<p style='font-size:14px; font-family: Arial; '>
|
139 |
+
Check if a relevant discussion already exists in the Community tab. If not, please open a discussion.
|
140 |
+
</p>
|
141 |
+
"""
|
142 |
+
|
143 |
+
return results_html
|
144 |
+
|
145 |
+
|
146 |
+
def flag(query, language, num_results, issue_description):
|
147 |
+
try:
|
148 |
+
post_data = {"query": query, "k": num_results, "flag": True, "description": issue_description}
|
149 |
+
if language != "detect_language":
|
150 |
+
post_data["lang"] = language
|
151 |
+
|
152 |
+
output = requests.post(
|
153 |
+
os.environ.get("address"),
|
154 |
+
headers={"Content-type": "application/json"},
|
155 |
+
data=json.dumps(post_data),
|
156 |
+
timeout=120,
|
157 |
+
)
|
158 |
+
|
159 |
+
results = json.loads(output.text)
|
160 |
+
except:
|
161 |
+
print("Error flagging")
|
162 |
+
return ""
|
163 |
+
|
164 |
+
|
165 |
+
description = """# <p style="text-align: center;"> πΈ π ROOTS search tool π πΈ </p>
|
166 |
+
The ROOTS corpus was developed during the [BigScience workshop](https://bigscience.huggingface.co/) for the purpose
|
167 |
+
of training the Multilingual Large Language Model [BLOOM](https://huggingface.co/bigscience/bloom). This tool allows
|
168 |
+
you to search through the ROOTS corpus. We serve a BM25 index for each language or group of languages included in
|
169 |
+
ROOTS. You can read more about the details of the tool design
|
170 |
+
[here](https://huggingface.co/spaces/bigscience-data/scisearch/blob/main/roots_search_tool_specs.pdf). For more
|
171 |
+
information and instructions on how to access the full corpus check [this form](https://forms.gle/qyYswbEL5kA23Wu99)."""
|
172 |
+
|
173 |
+
|
174 |
+
if __name__ == "__main__":
|
175 |
+
demo = gr.Blocks(
|
176 |
+
css=".underline-on-hover:hover { text-decoration: underline; } .flagging { font-size:12px; color:Silver; }"
|
177 |
+
)
|
178 |
+
|
179 |
+
with demo:
|
180 |
+
with gr.Row():
|
181 |
+
gr.Markdown(value=description)
|
182 |
+
with gr.Row():
|
183 |
+
query = gr.Textbox(lines=2, placeholder="Type your query here...", label="Query")
|
184 |
+
with gr.Row():
|
185 |
+
lang = gr.Dropdown(
|
186 |
+
choices=[
|
187 |
+
"ar",
|
188 |
+
"ca",
|
189 |
+
"code",
|
190 |
+
"en",
|
191 |
+
"es",
|
192 |
+
"eu",
|
193 |
+
"fr",
|
194 |
+
"id",
|
195 |
+
"indic",
|
196 |
+
"nigercongo",
|
197 |
+
"pt",
|
198 |
+
"vi",
|
199 |
+
"zh",
|
200 |
+
"detect_language",
|
201 |
+
"all",
|
202 |
+
],
|
203 |
+
value="pt",
|
204 |
+
label="Language",
|
205 |
+
)
|
206 |
+
with gr.Row():
|
207 |
+
k = gr.Slider(1, 100, value=10, step=1, label="Max Results")
|
208 |
+
with gr.Row():
|
209 |
+
submit_btn = gr.Button("Submit")
|
210 |
+
with gr.Row():
|
211 |
+
results = gr.HTML(label="Results")
|
212 |
+
flag_description = """
|
213 |
+
<p class='flagging'>
|
214 |
+
If you choose to flag your search, we will save the query, language and the number of results you requested.
|
215 |
+
Please consider adding any additional context in the box on the right.</p>"""
|
216 |
+
with gr.Column(visible=False) as flagging_form:
|
217 |
+
flag_txt = gr.Textbox(
|
218 |
+
lines=1,
|
219 |
+
placeholder="Type here...",
|
220 |
+
label="""If you choose to flag your search, we will save the query, language and the number of results
|
221 |
+
you requested. Please consider adding relevant additional context below:""",
|
222 |
+
)
|
223 |
+
flag_btn = gr.Button("Flag Results")
|
224 |
+
flag_btn.click(flag, inputs=[query, lang, k, flag_txt], outputs=[flag_txt])
|
225 |
+
|
226 |
+
def submit(query, lang, k):
|
227 |
+
if query == "":
|
228 |
+
return ["", ""]
|
229 |
+
return {
|
230 |
+
results: scisearch(query, lang, k),
|
231 |
+
flagging_form: gr.update(visible=True),
|
232 |
+
}
|
233 |
+
|
234 |
+
submit_btn.click(submit, inputs=[query, lang, k], outputs=[results, flagging_form])
|
235 |
+
|
236 |
+
demo.launch(enable_queue=True, debug=True)
|
roots_search_tool_specs.pdf
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:44cf8caa3819e8abf036178c4d329363d0649def28ec4e0979a0e0b2b94362d9
|
3 |
+
size 2642911
|