danielrosehill commited on
Commit
be11543
·
verified ·
1 Parent(s): e78ddf0

Export 341 sentences from the capture app bank

Browse files
Files changed (4) hide show
  1. LICENSE +1 -0
  2. README.md +83 -0
  3. data/test.jsonl +24 -0
  4. data/train.jsonl +0 -0
LICENSE ADDED
@@ -0,0 +1 @@
 
 
1
+ Mixed, per row: see the `source_licence` column and the README's "Attribution and licence" section. Rows without a `source_licence` are CC-BY-4.0.
README.md ADDED
@@ -0,0 +1,83 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: other
3
+ license_name: mixed-per-row
4
+ license_link: LICENSE
5
+ language:
6
+ - en
7
+ task_categories:
8
+ - automatic-speech-recognition
9
+ - text-generation
10
+ pretty_name: Linux Admin Sentence Bank
11
+ tags:
12
+ - asr
13
+ - fine-tuning
14
+ - read-aloud-sentences
15
+ - read-aloud
16
+ size_categories:
17
+ - n<1K
18
+ configs:
19
+ - config_name: default
20
+ data_files:
21
+ - split: train
22
+ path: data/train.jsonl
23
+ - split: test
24
+ path: data/test.jsonl
25
+ ---
26
+
27
+ # Linux Admin Sentence Bank
28
+
29
+ English sentences about Linux desktop and server administration, built around names a speech model writes as an ordinary English word: Konsole (console), Okular (ocular), Thunar (thunder), LUKS (looks), Btrfs, GRUB, Chrony (crony). Where natural, a sentence also uses the ordinary word, so the model has to tell them apart from context.
30
+
31
+ These are **sentences to read aloud** for fine-tuning a speech-recognition model, not
32
+ recordings. They feed a targeted fine-tune of NVIDIA Parakeet TDT 0.6B v2
33
+ ([current model](https://huggingface.co/danielrosehill/Parakeet-TDT-0.6B-v2-English-Hebrew-Codeswitching)). A capture app shows these sentences one at a time to be read aloud; every take lands in [Read-Aloud-Voice-Sentences](https://huggingface.co/datasets/danielrosehill/Read-Aloud-Voice-Sentences) and is copied into the matching audio subset. The recordings of these sentences are in
34
+ [Linux-Admin-Sentences](https://huggingface.co/datasets/danielrosehill/Linux-Admin-Sentences).
35
+
36
+ This dataset gathers sentences from several places into one file for my own training, and
37
+ it is reproduced here for convenience. Every row says where it came from, who wrote it and
38
+ how it was changed. Credit for upstream sentences belongs to their authors (see
39
+ "Attribution").
40
+
41
+ ## Size
42
+
43
+ 317 train, 24 test (341 total). The test split is **term-disjoint**:
44
+ no target term in `test` appears in `train`, so an evaluation on it measures generalisation,
45
+ not memorisation.
46
+
47
+ Derivation: generated 341.
48
+ Sentence authors: `claude (Claude Code subagent)` 336, `claude-opus-5.5 (Claude Code session)` 5.
49
+
50
+ ## Columns
51
+
52
+ | Column | Meaning |
53
+ |---|---|
54
+ | `id` | sentence id: a hash of the text (same id as the capture app and the audio datasets' `prompt_id`) |
55
+ | `text` | the sentence, exactly as read. Punctuated and cased; numbers in digits for versions, sizes, percentages and money |
56
+ | `target_terms` | the words the sentence exists to teach |
57
+ | `term_origins` | for each target term, who proposed it: `daniel` (from my own repos and notes, or I asked for it), `claude` (an AI agent proposed it), or the public list it came from (`openrouter`, `github-topics`, `cncf`, `pypi`, `stackoverflow`) |
58
+ | `categories` | `tech`, `linux`, `smart_home`, `homophone`, `numeric`, … |
59
+ | `homophone_set` | homophone rows only: the set the context disambiguates |
60
+ | `split` | `train` or `test` |
61
+ | `derivation` | `mined-verbatim` (taken unchanged from a public dataset), `mined-edited` (taken, then lightly fixed for casing, digits or readability), `generated` (written by an LLM for this project, then reviewed by a second model), `borrowed` (taken whole from an earlier sentence set) |
62
+ | `sentence_author` | the upstream dataset, or the model that wrote the sentence |
63
+ | `source_dataset`, `source_licence`, `source_url`, `source_authors`, `source_date` | upstream provenance for mined and borrowed rows. `source_url` links the paper or post; `source_authors` lists arXiv authors (Reddit usernames are deliberately not copied) |
64
+ | `original_text` | the upstream text before editing (`mined-edited` rows) |
65
+ | `review_note` | the reviewer's note when a sentence was fixed |
66
+
67
+ ## Attribution and licence
68
+
69
+ Licensing is per row (`source_licence`). Rows without one were written for this project and
70
+ are CC-BY-4.0.
71
+
72
+ | Upstream | Licence |
73
+ |---|---|
74
+ | (none: every row was written for this project) | CC-BY-4.0 |
75
+
76
+ Mined sentences were selected by `scripts/mine_tech_sentences.py` in the project repository:
77
+ they are 8–30 words long, contain at least one named tech term, and come from newer posts first.
78
+ An AI reviewer then kept or dropped each one and fixed casing and numbers.
79
+
80
+ ## Updating
81
+
82
+ Upstream sources can be re-mined, and new sentences are added as the project grows. Ids are
83
+ content hashes, so an edited sentence gets a new id.
data/test.jsonl ADDED
@@ -0,0 +1,24 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {"id": "p_0366c4f5ba", "text": "I bought fresh mint for the tea and then spent the whole afternoon reinstalling Linux Mint on the kitchen laptop.", "target_terms": ["Linux Mint"], "categories": ["tech", "linux"], "homophone_set": null, "split": "test", "derivation": "generated", "sentence_author": "claude (Claude Code subagent)", "source_dataset": null, "source_licence": null, "source_url": null, "source_authors": [], "source_date": null, "original_text": null, "review_note": null, "term_origins": ["claude"]}
2
+ {"id": "p_0be9192a6f", "text": "Thunar is the file manager on the old laptop running Xfce, and it handles network shares better than I expected.", "target_terms": ["Thunar"], "categories": ["tech", "linux"], "homophone_set": null, "split": "test", "derivation": "generated", "sentence_author": "claude (Claude Code subagent)", "source_dataset": null, "source_licence": null, "source_url": null, "source_authors": [], "source_date": null, "original_text": null, "review_note": null, "term_origins": ["claude"]}
3
+ {"id": "p_11d33f84a1", "text": "We opened a bottle of wine to celebrate, after Wine finally ran the old accounting program without crashing.", "target_terms": ["Wine"], "categories": ["tech", "linux"], "homophone_set": null, "split": "test", "derivation": "generated", "sentence_author": "claude (Claude Code subagent)", "source_dataset": null, "source_licence": null, "source_url": null, "source_authors": [], "source_date": null, "original_text": null, "review_note": null, "term_origins": ["claude"]}
4
+ {"id": "p_17514b3e4d", "text": "My parents' computer runs Linux Mint now, and they haven't called me about Windows updates since.", "target_terms": ["Linux Mint"], "categories": ["tech", "linux"], "homophone_set": null, "split": "test", "derivation": "generated", "sentence_author": "claude (Claude Code subagent)", "source_dataset": null, "source_licence": null, "source_url": null, "source_authors": [], "source_date": null, "original_text": null, "review_note": null, "term_origins": ["claude"]}
5
+ {"id": "p_2af5ec5c88", "text": "Read up on how cgroups version 2 handles CPU weights before I tune the limits on the transcription containers.", "target_terms": ["cgroups"], "categories": ["tech", "linux", "numeric"], "homophone_set": null, "split": "test", "derivation": "generated", "sentence_author": "claude (Claude Code subagent)", "source_dataset": null, "source_licence": null, "source_url": null, "source_authors": [], "source_date": null, "original_text": null, "review_note": null, "term_origins": ["claude"]}
6
+ {"id": "p_2fe8f45d41", "text": "Note to self: bind Spectacle to the print screen key and make it copy rectangular captures straight to the clipboard.", "target_terms": ["Spectacle"], "categories": ["tech", "linux"], "homophone_set": null, "split": "test", "derivation": "generated", "sentence_author": "claude (Claude Code subagent)", "source_dataset": null, "source_licence": null, "source_url": null, "source_authors": [], "source_date": null, "original_text": null, "review_note": null, "term_origins": ["claude"]}
7
+ {"id": "p_3f7e665159", "text": "The thunder last night knocked out the power, and when the laptop came back Thunar had forgotten all my bookmarks.", "target_terms": ["Thunar"], "categories": ["tech", "linux"], "homophone_set": null, "split": "test", "derivation": "generated", "sentence_author": "claude (Claude Code subagent)", "source_dataset": null, "source_licence": null, "source_url": null, "source_authors": [], "source_date": null, "original_text": null, "review_note": null, "term_origins": ["claude"]}
8
+ {"id": "p_4a7b91d438", "text": "The VM itself is fine; it's the cgroups limit on that one container that caps it at 2 GB of memory.", "target_terms": ["cgroups"], "categories": ["tech", "linux", "numeric"], "homophone_set": null, "split": "test", "derivation": "generated", "sentence_author": "claude (Claude Code subagent)", "source_dataset": null, "source_licence": null, "source_url": null, "source_authors": [], "source_date": null, "original_text": null, "review_note": null, "term_origins": ["claude"]}
9
+ {"id": "p_5cd97637ad", "text": "Please write a short setup checklist for a fresh Linux Mint install on an old ThinkPad.", "target_terms": ["Linux Mint"], "categories": ["tech", "linux"], "homophone_set": null, "split": "test", "derivation": "generated", "sentence_author": "claude (Claude Code subagent)", "source_dataset": null, "source_licence": null, "source_url": null, "source_authors": [], "source_date": null, "original_text": null, "review_note": null, "term_origins": ["claude"]}
10
+ {"id": "p_643c042852", "text": "Check whether the container is hitting its memory limit in cgroups, because the process keeps getting throttled without any obvious error in the logs.", "target_terms": ["cgroups"], "categories": ["tech", "linux"], "homophone_set": null, "split": "test", "derivation": "generated", "sentence_author": "claude (Claude Code subagent)", "source_dataset": null, "source_licence": null, "source_url": null, "source_authors": [], "source_date": null, "original_text": null, "review_note": null, "term_origins": ["claude"]}
11
+ {"id": "p_67ef9d4c66", "text": "I mostly run quick commands in Yakuake and keep the long sessions in a proper terminal window.", "target_terms": ["Yakuake"], "categories": ["tech", "linux"], "homophone_set": null, "split": "test", "derivation": "generated", "sentence_author": "claude (Claude Code subagent)", "source_dataset": null, "source_licence": null, "source_url": null, "source_authors": [], "source_date": null, "original_text": null, "review_note": null, "term_origins": ["claude"]}
12
+ {"id": "p_6ab8262078", "text": "Spectacle isn't capturing the second monitor anymore, so check whether it's a Wayland problem or a KWin bug.", "target_terms": ["Spectacle"], "categories": ["tech", "linux"], "homophone_set": null, "split": "test", "derivation": "generated", "sentence_author": "claude (Claude Code subagent)", "source_dataset": null, "source_licence": null, "source_url": null, "source_authors": [], "source_date": null, "original_text": null, "review_note": "factual: Spectacle captures via KWin, not the screen-sharing portal", "term_origins": ["claude"]}
13
+ {"id": "p_740ad3f913", "text": "Add Yakuake to the autostart list so it's ready the moment I log in on Kubuntu.", "target_terms": ["Yakuake"], "categories": ["tech", "linux"], "homophone_set": null, "split": "test", "derivation": "generated", "sentence_author": "claude (Claude Code subagent)", "source_dataset": null, "source_licence": null, "source_url": null, "source_authors": [], "source_date": null, "original_text": null, "review_note": null, "term_origins": ["claude"]}
14
+ {"id": "p_74ffb7c0e2", "text": "The kids built a cardboard starship in the living room while I was trying to make Starship load faster.", "target_terms": ["Starship"], "categories": ["tech", "linux"], "homophone_set": null, "split": "test", "derivation": "generated", "sentence_author": "claude (Claude Code subagent)", "source_dataset": null, "source_licence": null, "source_url": null, "source_authors": [], "source_date": null, "original_text": null, "review_note": null, "term_origins": ["claude"]}
15
+ {"id": "p_7661992503", "text": "Set up Fail2Ban on the VPS so that anyone who fails the SSH login five times gets banned for an hour.", "target_terms": ["Fail2Ban"], "categories": ["tech", "linux", "numeric"], "homophone_set": null, "split": "test", "derivation": "generated", "sentence_author": "claude (Claude Code subagent)", "source_dataset": null, "source_licence": null, "source_url": null, "source_authors": [], "source_date": null, "original_text": null, "review_note": null, "term_origins": ["claude"]}
16
+ {"id": "p_79652e68c7", "text": "Can you find out why Thunar shows a blank thumbnail for every video file on the Synology share?", "target_terms": ["Thunar"], "categories": ["tech", "linux"], "homophone_set": null, "split": "test", "derivation": "generated", "sentence_author": "claude (Claude Code subagent)", "source_dataset": null, "source_licence": null, "source_url": null, "source_authors": [], "source_date": null, "original_text": null, "review_note": null, "term_origins": ["claude"]}
17
+ {"id": "p_a31173ceec", "text": "Please add a Starship module that shows the active Python virtual environment and the Kubernetes context.", "target_terms": ["Starship"], "categories": ["tech", "linux"], "homophone_set": null, "split": "test", "derivation": "generated", "sentence_author": "claude (Claude Code subagent)", "source_dataset": null, "source_licence": null, "source_url": null, "source_authors": [], "source_date": null, "original_text": null, "review_note": null, "term_origins": ["claude"]}
18
+ {"id": "p_a31aa55537", "text": "Starship is now my prompt in every shell, so the Git branch shows up the same way everywhere.", "target_terms": ["Starship"], "categories": ["tech", "linux"], "homophone_set": null, "split": "test", "derivation": "generated", "sentence_author": "claude (Claude Code subagent)", "source_dataset": null, "source_licence": null, "source_url": null, "source_authors": [], "source_date": null, "original_text": null, "review_note": null, "term_origins": ["claude"]}
19
+ {"id": "p_c57ad45556", "text": "Yakuake drops down on the wrong monitor every time I press the hotkey, so see if there's a setting to pin it.", "target_terms": ["Yakuake"], "categories": ["tech", "linux"], "homophone_set": null, "split": "test", "derivation": "generated", "sentence_author": "claude (Claude Code subagent)", "source_dataset": null, "source_licence": null, "source_url": null, "source_authors": [], "source_date": null, "original_text": null, "review_note": null, "term_origins": ["claude"]}
20
+ {"id": "p_c99cfe5e91", "text": "I'm running the label printer software through Wine, since there's no Linux version of it.", "target_terms": ["Wine"], "categories": ["tech", "linux"], "homophone_set": null, "split": "test", "derivation": "generated", "sentence_author": "claude (Claude Code subagent)", "source_dataset": null, "source_licence": null, "source_url": null, "source_authors": [], "source_date": null, "original_text": null, "review_note": null, "term_origins": ["claude"]}
21
+ {"id": "p_d1d2803deb", "text": "Can you check whether this Windows tool works under Wine, or whether I need a full virtual machine for it?", "target_terms": ["Wine"], "categories": ["tech", "linux"], "homophone_set": null, "split": "test", "derivation": "generated", "sentence_author": "claude (Claude Code subagent)", "source_dataset": null, "source_licence": null, "source_url": null, "source_authors": [], "source_date": null, "original_text": null, "review_note": null, "term_origins": ["claude"]}
22
+ {"id": "p_d854aa7602", "text": "Fail2Ban banned my own IP again this morning, so I had to log in through the Proxmox console to unblock myself.", "target_terms": ["Fail2Ban"], "categories": ["tech", "linux", "numeric"], "homophone_set": null, "split": "test", "derivation": "generated", "sentence_author": "claude (Claude Code subagent)", "source_dataset": null, "source_licence": null, "source_url": null, "source_authors": [], "source_date": null, "original_text": null, "review_note": null, "term_origins": ["claude"]}
23
+ {"id": "p_e05208fdc2", "text": "The fireworks on the livestream were quite a spectacle, but Spectacle kept saving black frames when I tried to record them.", "target_terms": ["Spectacle"], "categories": ["tech", "linux"], "homophone_set": null, "split": "test", "derivation": "generated", "sentence_author": "claude (Claude Code subagent)", "source_dataset": null, "source_licence": null, "source_url": null, "source_authors": [], "source_date": null, "original_text": null, "review_note": "factual: a screen recorder can't record real fireworks; made it a livestream", "term_origins": ["claude"]}
24
+ {"id": "p_e3b334bb87", "text": "Remember to whitelist the Tailscale range in Fail2Ban before tightening the retry limit any further.", "target_terms": ["Fail2Ban"], "categories": ["tech", "linux", "numeric"], "homophone_set": null, "split": "test", "derivation": "generated", "sentence_author": "claude (Claude Code subagent)", "source_dataset": null, "source_licence": null, "source_url": null, "source_authors": [], "source_date": null, "original_text": null, "review_note": null, "term_origins": ["claude"]}
data/train.jsonl ADDED
The diff for this file is too large to render. See raw diff