AhmadMustafa commited on
Commit
e8e9f36
·
1 Parent(s): dd57ac1

update: si prompt

Browse files
Files changed (1) hide show
  1. app.py +13 -4
app.py CHANGED
@@ -346,22 +346,28 @@ Your task is to analyze this street interview transcript and identify the final/
346
  - The LAST occurrence is always considered the best version
347
  - This includes cases where parts of an answer are scattered throughout the transcript
348
  - Even slight variations of the same answer should be tracked
 
349
 
350
  2. Introduction handling:
351
  - Question 1 is ALWAYS the speaker's introduction/self-introduction
352
- - If someone introduces themselves multiple times, use the last timestamp
353
  - Include all variations of how they state their name/background
 
354
 
355
  3. Question sequence:
356
  - After the introduction, list questions in the order they were first asked
357
- - If a question is revisited later, use the later timestamp
358
  - Track partial answers to the same question across the transcript
359
 
360
  Return format:
361
 
362
  [Question Title]
363
  Total takes: [X] (Include ONLY if content appears more than once)
364
- [Best Answer: start_time - end_time]({link_start}://{{origin}}/collab/{{cid}}/{{rsid}}?st={{start_time_in_sec}}&et={{end_time_in_sec}})
 
 
 
 
365
 
366
  URL formatting:
367
  - Convert timestamps to seconds (e.g., 10:13 → 613)
@@ -371,7 +377,10 @@ URL formatting:
371
  Example:
372
  1. Introduction
373
  Total takes: 2
374
- [Best Answer: 10:13 - 10:18]({link_start}://roll.ai/colab/1234aq_12314/51234151?st=613&et=618)"""
 
 
 
375
  else:
376
  prompt = f"""Given the transcript {transcript}, For All the speakers, short list all people, news, events, trends, and source that are discussed by speakers along with the start time of that topic and end time of that topic from the transcript. Rank all topics based on what would make for the best social clips. I need atleast 3 topics per speaker.
377
  You should mention the Speaker Name first, then atleast 3 posts with their timestamps, and so on.
 
346
  - The LAST occurrence is always considered the best version
347
  - This includes cases where parts of an answer are scattered throughout the transcript
348
  - Even slight variations of the same answer should be tracked
349
+ - List timestamps for ALL takes, with the final take highlighted as the best answer
350
 
351
  2. Introduction handling:
352
  - Question 1 is ALWAYS the speaker's introduction/self-introduction
353
+ - If someone introduces themselves multiple times, use the last timestamp as best answer
354
  - Include all variations of how they state their name/background
355
+ - List ALL introduction timestamps chronologically
356
 
357
  3. Question sequence:
358
  - After the introduction, list questions in the order they were first asked
359
+ - If a question or introduction is revisited later at any point, please use the later timestamp
360
  - Track partial answers to the same question across the transcript
361
 
362
  Return format:
363
 
364
  [Question Title]
365
  Total takes: [X] (Include ONLY if content appears more than once)
366
+ Takes:
367
+ 1. [start_time - end_time]
368
+ 2. [start_time - end_time]
369
+ ...
370
+ Best Answer: [start_time - end_time]({link_start}://{{origin}}/collab/{{cid}}/{{rsid}}?st={{start_time_in_sec}}&et={{end_time_in_sec}})
371
 
372
  URL formatting:
373
  - Convert timestamps to seconds (e.g., 10:13 → 613)
 
377
  Example:
378
  1. Introduction
379
  Total takes: 2
380
+ Takes:
381
+ 1. [09:45 - 09:50]
382
+ 2. [10:13 - 10:18]
383
+ Best Answer: [10:13 - 10:18]({link_start}://roll.ai/colab/1234aq_12314/51234151?st=613&et=618)"""
384
  else:
385
  prompt = f"""Given the transcript {transcript}, For All the speakers, short list all people, news, events, trends, and source that are discussed by speakers along with the start time of that topic and end time of that topic from the transcript. Rank all topics based on what would make for the best social clips. I need atleast 3 topics per speaker.
386
  You should mention the Speaker Name first, then atleast 3 posts with their timestamps, and so on.