Spaces:
Sleeping
Sleeping
oceansweep
commited on
Commit
•
c96320d
1
Parent(s):
cd5e862
Upload Confabulation_check.py
Browse files
App_Function_Libraries/Confabulation_check.py
ADDED
@@ -0,0 +1,81 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# Confabulation_check.py
|
2 |
+
#
|
3 |
+
# This file contains the functions that are used to check the confabulation of the user's input.
|
4 |
+
#
|
5 |
+
#
|
6 |
+
# Imports
|
7 |
+
#
|
8 |
+
# External Imports
|
9 |
+
#
|
10 |
+
# Local Imports
|
11 |
+
#
|
12 |
+
#
|
13 |
+
####################################################################################################
|
14 |
+
#
|
15 |
+
# Functions:
|
16 |
+
from App_Function_Libraries.Chat import chat_api_call
|
17 |
+
from App_Function_Libraries.ms_g_eval import validate_inputs, detailed_api_error
|
18 |
+
|
19 |
+
|
20 |
+
def simplified_geval(transcript: str, summary: str, api_name: str, api_key: str, temp: float = 0.7) -> str:
|
21 |
+
"""
|
22 |
+
Perform a simplified version of G-Eval using a single query to evaluate the summary.
|
23 |
+
|
24 |
+
Args:
|
25 |
+
transcript (str): The original transcript
|
26 |
+
summary (str): The summary to be evaluated
|
27 |
+
api_name (str): The name of the LLM API to use
|
28 |
+
api_key (str): The API key for the chosen LLM
|
29 |
+
temp (float, optional): The temperature parameter for the API call. Defaults to 0.7.
|
30 |
+
|
31 |
+
Returns:
|
32 |
+
str: The evaluation result
|
33 |
+
"""
|
34 |
+
try:
|
35 |
+
validate_inputs(transcript, summary, api_name, api_key)
|
36 |
+
except ValueError as e:
|
37 |
+
return str(e)
|
38 |
+
|
39 |
+
prompt = f"""You are an AI assistant tasked with evaluating the quality of a summary. You will be given an original transcript and a summary of that transcript. Your task is to evaluate the summary based on the following criteria:
|
40 |
+
|
41 |
+
1. Coherence (1-5): How well-structured and organized is the summary?
|
42 |
+
2. Consistency (1-5): How factually aligned is the summary with the original transcript?
|
43 |
+
3. Fluency (1-3): How well-written is the summary in terms of grammar, spelling, and readability?
|
44 |
+
4. Relevance (1-5): How well does the summary capture the important information from the transcript?
|
45 |
+
|
46 |
+
Please provide a score for each criterion and a brief explanation for your scoring. Then, give an overall assessment of the summary's quality.
|
47 |
+
|
48 |
+
Original Transcript:
|
49 |
+
{transcript}
|
50 |
+
|
51 |
+
Summary to Evaluate:
|
52 |
+
{summary}
|
53 |
+
|
54 |
+
Please provide your evaluation in the following format:
|
55 |
+
Coherence: [score] - [brief explanation]
|
56 |
+
Consistency: [score] - [brief explanation]
|
57 |
+
Fluency: [score] - [brief explanation]
|
58 |
+
Relevance: [score] - [brief explanation]
|
59 |
+
|
60 |
+
Overall Assessment: [Your overall assessment of the summary's quality]
|
61 |
+
"""
|
62 |
+
|
63 |
+
try:
|
64 |
+
result = chat_api_call(
|
65 |
+
api_name,
|
66 |
+
api_key,
|
67 |
+
prompt,
|
68 |
+
"",
|
69 |
+
temp=temp,
|
70 |
+
system_message="You are a helpful AI assistant tasked with evaluating summaries."
|
71 |
+
)
|
72 |
+
except Exception as e:
|
73 |
+
return detailed_api_error(api_name, e)
|
74 |
+
|
75 |
+
formatted_result = f"""
|
76 |
+
Confabulation Check Results:
|
77 |
+
|
78 |
+
{result}
|
79 |
+
"""
|
80 |
+
|
81 |
+
return formatted_result
|