add tags to old articles

articles/neocities-cache/page.mmd

@@ -1,7 +1,7 @@
 Title:  Cached Neocities Uploads
 Brief:  Making uploading of directories to Neocities less painful.
 Date:   1707585916
-Tags:   Programming, Bash
+Tags:   Programming, Bash, Script
 CSS:    /style.css
 
 Quick and dirty Bash-based sha256sum checksum solution to create stamps for later checking and rejection.

articles/tiny-elf/page.mmd

@@ -1,7 +1,7 @@
 Title:  Slim Summer Elf
 Brief:  Making of minimal x86 (Linux) ELF executable.
 Date:   1684666702
-Tags:   Programming, Linux, C
+Tags:   Programming, Linux, C, Bash, Linker, Low-level
 CSS:    /style.css
 
 Code below was composed for [4mb-jam](https://itch.io/jam/4mb-jam-2023) which I didn't finish.

articles/toponym-extractor/page.mmd

@@ -0,0 +1,36 @@
+Title:  Geonames Toponym Extractor Utility
+Brief:  Simple script for extracting ASCII toponym fields from geonames datasets
+Date:   1713683410
+Tags:   Python, Script, Programming
+CSS:    /style.css
+
+[Link to code](https://codeberg.org/veclavtalica/geonames-extractor)
+
+Small script I used for extracting data for machine learning endeavors.
+
+Usage:
+```
+dataset feature_class [feature_code] [--dirty] [--filter=mask]
+```
+
+From this invokation ...
+```
+./extractor.py datasets/UA.txt P PPL --filter=0123456789\"\'-\` > UA-prep.txt
+```
+
+... it produces a newline separated list of relevant toponyms of particular kind, such as:
+```
+Katerynivka
+Vaniushkyne
+Svistuny
+Sopych
+Shilova Balka
+```
+
+`--filter=` option is there so that aplhabet size could be reduced for learning purposes,
+as there are usually quite a lot of symbols that are only found few times,
+which produces poor balancing.
+
+`--dirty` option reduces cases such as `Maydan (Ispas)` and `CHAYKA-Transmitter, Ring Mast 4` to `Maydan` and `CHAYKA-Transmitter`.
+
+Duplicates are also removed.