Threadbourne commited on
Commit
bc07c99
·
verified ·
1 Parent(s): 2092878

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +38 -14
README.md CHANGED
@@ -10,22 +10,46 @@ pinned: false
10
  license: mit
11
  short_description: 'Turn .md / .txt into csv '
12
  ---
13
- This tool converts structured chat transcripts written in markdown or plain text into a canonical turn-level CSV format suitable for analysis with Threadscope: Drift & Hold.
14
 
15
- Supported Format:
16
- This convertor expects transcripts that use explicit speaker markers, for example:
17
- You said: Hello
18
- ChatGPT said: Hi there
19
- Marker strings are configurable in the UI, but the transcript must follow a consistent "marker to content" pattern
20
 
21
- Not supported:
22
- Arbitarily formatted prose
23
- Chat logs without speaker markers
24
- Exported HTML/JSON without preprocessing (in process - update coming soon)
25
 
26
- Output:
27
- The output CSV includes: turn, speaker, chars, words, tokens_est (lightweight est.)
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
28
 
29
- This tool is intended as a preprocessing step for use with Threadscope, which visualizes long-form interaction dynamics such as drift, stability, and perturbation response.
30
- Files are processed in-session only and are not stored
31
  Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
 
10
  license: mit
11
  short_description: 'Turn .md / .txt into csv '
12
  ---
13
+ ## What this tool does
14
 
15
+ This tool converts structured chat transcripts written in Markdown or plain text
16
+ into a canonical turn-level CSV format suitable for analysis with **Threadscope:
17
+ Drift & Hold**.
 
 
18
 
19
+ ## Supported transcript format
 
 
 
20
 
21
+ This converter expects transcripts that use explicit speaker markers, for example:
22
+
23
+ - 'You said: Hello'
24
+ - 'ChatGPT said: Hi there'
25
+
26
+ Marker strings are configurable in the UI, but the transcript must follow a
27
+ consistent “marker → content” pattern.
28
+
29
+ ## Not supported
30
+
31
+ - Arbitrarily formatted prose
32
+ - Chat logs without explicit speaker markers
33
+ - Exported HTML or JSON without preprocessing
34
+
35
+ ## Output
36
+
37
+ The generated CSV includes the following columns:
38
+
39
+ - `turn`
40
+ - `speaker`
41
+ - `chars`
42
+ - `words`
43
+ - `tokens_est` (lightweight estimate)
44
+
45
+ ## How this fits into the Threadbourne lab
46
+
47
+ This tool is intended as a preprocessing step for use with
48
+ **Threadscope: Drift & Hold**, which visualizes long-form interaction dynamics
49
+ such as drift, stability, and perturbation response.
50
+
51
+ ## Privacy
52
+
53
+ Files are processed in-session only and are not stored.
54
 
 
 
55
  Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference