Google Scholarのインデックス基準に関する考察 ―技術的テンプレートが学術的価値を超える皮肉な現実―
Description
本研究は、Google Scholarのインデックス登録プロセスを自らの実体験に基づき考察する。特に著名な研究者の論文であっても、技術的な基準を満たさない場合は収録が著しく遅れる現象を報告する。一方、学術的価値が相対的に低くとも、メタデータやクロールの技術的テンプレートが完璧な論文は迅速にインデックスされることを示し、Google Scholarの「学術的質の審査」ではなく「技術的基準」による機械的な収録という設計思想の本質を明らかにする。
Files
Steps to reproduce
To quantify how a live commercial site reacts when its top-level domain changes from “.com” to “.jp”, we captured day-by-day search signals for ten months in a fully script-driven workflow that leaves no room for manual edits. Collection started on 1 January 2024, three months before cut-over, and ended on 31 October 2024. Each night at 02:00 UTC a Python 3.11 routine authenticated against the Google Search Console API, pulled clicks, impressions, click-through rate and average position for every verified URL on both domains and wrote the untouched CSV export to a write-once directory. Competitive context was recorded through weekly calls to Ahrefs (API build 203 released 2024-03-01) and, after its quota expired, SEMrush (API v4.7 dated 2024-04-15); the job archived raw JSON snapshots containing estimated organic visits, live keyword counts and referring-domain totals with timestamped file names to preserve provenance. The physical migration occurred at 12:15 UTC on 3 April 2024. A spreadsheet vetted line-by-line mapped all 2 706 legacy URLs to their new counterparts. Nginx was updated via an include file of permanent return-301 rules and reloaded through “sudo systemctl restart nginx”. Immediately afterwards random pages were spot-checked for correct canonical tags, headers and cache directives, a Change-of-Address request and a fresh sitemap were submitted to Search Console, and a GA4 annotation marked the moment of the switch. Analysis in a public Jupyter notebook merged the daily tables on the date key, patched gaps, attached the closest external snapshot and weekly Googlebot crawl counts parsed with GoAccess, then plotted total clicks, live keyword volume and brand-query share with Matplotlib, shading the first four weeks post-migration. Reproduction requires Ubuntu 22.04, Python 3.11, OAuth keys for both Search Console properties, an Ahrefs or SEMrush licence and the repository that contains raw exports, the notebook and a requirements file; executing the notebook end-to-end downloads fresh data, rebuilds every table and recreates the figures, yielding trajectories that should mirror the published ones unless external estimation algorithms undergo major revisions.