Classified the 23:55 Text Converter Keyword Planner (932 terms): ASR dropped, image-to-text absorbed into N1, Word/HTML-to-text deferred.

Date: 2026-09-08 08:35 Summary: Classified the 23:55 Text Converter Keyword Planner (932 terms): ASR dropped, image-to-text absorbed into N1, Word/HTML-to-text deferred. Visibility: people

[question] 增加关键 planner，Keyword Planner 09-07-2026 at 23-55-12，进行分析

[try to solve]

Context

A new Google Keyword Planner CSV landed in docs/seo/keywords/ocr-scan/ named Keyword Planner 09-07-2026 at 23-55-12.csv. The existing 2026-09-07 OCR+Scan analysis already covered 2643 hardware/OCR rows. This dump is a different seed: “Text Converter / convert to text,” 932 unique keywords, 52 ad groups.

Process

Parsed volumes and ad groups (1 term at 10k–100k: speech to text; 22 at 1k–10k).
Clustered by actual job, not Free/Online/Tool labels: speech/audio/video ASR vs photo OCR vs PDF text layer vs Word file vs HTML/URL vs shells.
Compared catalog: N1 convert-a-jpg-to-text-with-ocr already does photo→TXT; extract-text-from-pdf does digital PDF text layer; HTML-to-Word and images-to-Word emit DOCX, not TXT; markdown-to-html emits HTML.
Wrote 2026-09-08-text-converter-keyword-planner.md as a sibling authority (does not replace N1–N5). Appended pool rows and tracker snapshot. No work-tasks/ created.

Root cause / analysis

Planner expanded a magnet phrase. About 40% of rows are cloud speech-to-text (Google/YouTube/MP3). That is not OCR and cannot be a local WASM product on this site. The useful 1k-band file jobs are jpg/jpeg/png/image to text (same control as N1; H1 already says JPG to text) and pdf to text (A1, not scanned OCR). Word→TXT and HTML/URL→TXT are real missing jobs with different controls, so they stay defer with suggested slugs rather than being stuffed into N1 or screenshot tools. Converter-category heads (ocr converter, online converter) stay doorway drop.

Solution

Absorb: image-family → N1 (FAQ still owes image to text / png to text); pdf to text → extract-text-from-pdf; csv to text partial → csv-json; md to text boundary FAQ on markdown-to-html.
Defer: extract-text-from-a-word-document, convert-html-or-a-web-page-to-plain-text, weak PPT; N2 gains scanned pdf to text evidence only.
Drop: entire ASR cluster, converter shells, text-recovery homonym, handwriting as a product URL.
No long_gap, no weekly build, no new work-tasks.

Notes / boundaries

N1 page copy was not edited in this analysis pass. Neighbor HTML-to-Word must not absorb “webpage to text.” Do not start T4/T5 unless the user names them.

[actions]

Added docs/seo/keywords/ocr-scan/2026-09-08-text-converter-keyword-planner.md
Updated ocr-scan and keywords README, 09-07 cross-link, keyword-daily-pool.tsv, keyword-to-tool-tracker.md