Upload README.md with huggingface_hub
Browse files
README.md
CHANGED
|
@@ -31,9 +31,9 @@ This project develops a unique dataset from the public archives of the wonderful
|
|
| 31 |
|
| 32 |
There are (3) datasets available:
|
| 33 |
|
| 34 |
-
1. Booknotes Programs:
|
| 35 |
-
2. Booknotes Transcripts: Full conversation transcripts between Brian Lamb and his guests.
|
| 36 |
-
3. Booknotes Related Items: Related or recommended programs for each episode.
|
| 37 |
|
| 38 |
The `transcripts` dataset is the key dataset here, with the other 2 providing additional context and information about each episode. Using the `sequence`, `speaker_role`, and `text` fields, we can create a chat-like dataset (representing very interesting conversations!) for evaluating language models.
|
| 39 |
|
|
@@ -52,25 +52,25 @@ To understand how the 3 tables work together, we provide the source JSON schema
|
|
| 52 |
"air_date": "October 17, 1993",
|
| 53 |
"transcript": [
|
| 54 |
{
|
| 55 |
-
"
|
| 56 |
"speaker_role": "host",
|
| 57 |
"speaker_name": "BRIAN LAMB, HOST:",
|
| 58 |
"text": "Christopher Hitchens, author of For the Sake of Argument, you've got a section in there called \"Rogues' Gallery.\" Was that your idea?"
|
| 59 |
},
|
| 60 |
{
|
| 61 |
-
"
|
| 62 |
"speaker_role": "guest",
|
| 63 |
"speaker_name": "CHRISTOPHER HITCHENS:",
|
| 64 |
"text": "Yes."
|
| 65 |
},
|
| 66 |
{
|
| 67 |
-
"
|
| 68 |
"speaker_role": "host",
|
| 69 |
"speaker_name": "LAMB:",
|
| 70 |
"text": "Why create a Rogues Gallery?"
|
| 71 |
},
|
| 72 |
{
|
| 73 |
-
"
|
| 74 |
"speaker_role": "guest",
|
| 75 |
"speaker_name": "HITCHENS:",
|
| 76 |
"text": "For a lot of people, their first love is what they'll always remember. For me it's always been the first hate, and I think that hatred, though it provides often rather junky energy, is a terrific way of getting you out of bed in the morning and keeping you going. If you don't let it get out of hand, it can be canalized into writing. In this country where people love to be nonjudgmental when they can be, which translates as, on the whole, lenient, there are an awful lot of bubble reputations floating around that one wouldn't be doing one's job if one didn't itch to prick."
|
|
|
|
| 31 |
|
| 32 |
There are (3) datasets available:
|
| 33 |
|
| 34 |
+
1. Booknotes Programs: Information for ~809 episodes, including title, description and guest information.
|
| 35 |
+
2. Booknotes Transcripts: Full conversation transcripts (~200 turns/conversation) between Brian Lamb and his guests.
|
| 36 |
+
3. Booknotes Related Items: Related or recommended programs (~5) for each episode.
|
| 37 |
|
| 38 |
The `transcripts` dataset is the key dataset here, with the other 2 providing additional context and information about each episode. Using the `sequence`, `speaker_role`, and `text` fields, we can create a chat-like dataset (representing very interesting conversations!) for evaluating language models.
|
| 39 |
|
|
|
|
| 52 |
"air_date": "October 17, 1993",
|
| 53 |
"transcript": [
|
| 54 |
{
|
| 55 |
+
"sequence": 0,
|
| 56 |
"speaker_role": "host",
|
| 57 |
"speaker_name": "BRIAN LAMB, HOST:",
|
| 58 |
"text": "Christopher Hitchens, author of For the Sake of Argument, you've got a section in there called \"Rogues' Gallery.\" Was that your idea?"
|
| 59 |
},
|
| 60 |
{
|
| 61 |
+
"sequence": 1,
|
| 62 |
"speaker_role": "guest",
|
| 63 |
"speaker_name": "CHRISTOPHER HITCHENS:",
|
| 64 |
"text": "Yes."
|
| 65 |
},
|
| 66 |
{
|
| 67 |
+
"sequence": 2,
|
| 68 |
"speaker_role": "host",
|
| 69 |
"speaker_name": "LAMB:",
|
| 70 |
"text": "Why create a Rogues Gallery?"
|
| 71 |
},
|
| 72 |
{
|
| 73 |
+
"sequence": 3,
|
| 74 |
"speaker_role": "guest",
|
| 75 |
"speaker_name": "HITCHENS:",
|
| 76 |
"text": "For a lot of people, their first love is what they'll always remember. For me it's always been the first hate, and I think that hatred, though it provides often rather junky energy, is a terrific way of getting you out of bed in the morning and keeping you going. If you don't let it get out of hand, it can be canalized into writing. In this country where people love to be nonjudgmental when they can be, which translates as, on the whole, lenient, there are an awful lot of bubble reputations floating around that one wouldn't be doing one's job if one didn't itch to prick."
|