[
https://issues.apache.org/jira/browse/TIKA-2790?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16856102#comment-16856102
]
Tim Allison commented on TIKA-2790:
-----------------------------------
The above was with a single sample. I wanted to gauge variance. So I ran 100
iterations of randomly selected sentences:
||Language||MaxWrongConfAvg||MaxWrongConfStdev||LengthAtMaxWrongConf||LengthAtMaxWrongStdev||
|afr|0.02|0.01|23.5|20.9|
|ara|0.02|0.00|14.3|12.6|
|ast|0.04|0.05|51.5|45.7|
|aze|0.02|0.01|18.8|17.9|
|bak|0.24|0.18|133.5|91.8|
|bel|0.03|0.03|29.2|36.1|
|ben|0.02|0.01|94.4|62.3|
|bos|0.56|0.09|446.0|139.0|
|bre|0.01|0.01|15.9|18.1|
|bul|0.02|0.01|20.3|28.1|
|cat|0.03|0.03|54.0|57.3|
|ceb|0.04|0.05|43.3|53.8|
|ces|0.07|0.13|50.6|65.5|
|che|0.03|0.03|30.0|46.2|
|cmn|0.01|0.00|18.0|11.0|
|cym|0.02|0.01|17.9|17.1|
|dan|0.03|0.05|40.9|44.5|
|deu|0.02|0.02|23.5|24.4|
|ekk|0.30|0.21|189.8|117.3|
|ell|0.03|0.02|76.7|83.3|
|eng|0.02|0.01|24.0|22.0|
|epo|0.01|0.00|13.3|9.5|
|est|0.24|0.21|160.5|119.5|
|eus|0.02|0.01|24.0|23.9|
|fao|0.01|0.01|14.5|12.4|
|fas|0.46|0.14|359.0|136.3|
|fin|0.02|0.01|17.1|17.4|
|fra|0.02|0.01|21.1|20.7|
|fry|0.04|0.08|31.4|45.9|
|gle|0.01|0.00|12.8|11.6|
|glg|0.02|0.02|28.5|28.4|
|gsw|0.03|0.04|35.5|50.4|
|heb|0.04|0.02|115.0|34.2|
|hin|0.02|0.01|15.6|9.1|
|hrv|0.50|0.15|489.4|218.1|
|hun|0.01|0.01|17.8|16.5|
|hye|0.02|0.01|34.0|35.8|
|ind|0.73|0.08|1087.8|348.0|
|isl|0.02|0.01|19.7|18.8|
|ita|0.02|0.02|27.9|40.8|
|jav|0.04|0.05|51.0|55.3|
|jpn|0.02|0.01|19.1|24.3|
|kan|0.01|0.01|40.0|47.6|
|kat|0.03|0.01|63.3|46.2|
|kaz|0.02|0.01|19.7|17.4|
|kir|0.04|0.05|49.0|53.6|
|kor|0.02|0.00|70.0|0.0|
|lat|0.02|0.02|24.8|31.6|
|lav|0.21|0.17|138.2|94.6|
|lim|0.04|0.04|40.4|37.7|
|lit|0.01|0.01|15.4|16.4|
|ltz|0.02|0.01|22.4|19.8|
|lvs|0.18|0.18|121.9|114.7|
|mal|0.02|0.00|50.0|0.0|
|mar|0.02|0.01|18.1|23.2|
|min|0.03|0.03|33.6|39.3|
|mkd|0.03|0.02|27.3|34.9|
|mlt|0.01|0.01|16.6|16.6|
|mon|0.02|0.01|15.2|11.5|
|mri|0.04|0.03|81.1|76.5|
|msa|0.26|0.15|280.0|171.6|
|nan|0.01|0.00|10.7|3.6|
|nds|0.02|0.02|25.1|33.5|
|nep|0.02|0.01|14.6|10.7|
|nld|0.03|0.03|31.6|35.8|
|nno|0.10|0.13|98.8|108.1|
|nob|0.13|0.17|95.3|100.4|
|oci|0.02|0.02|25.9|31.4|
|pan|0.01|0.00|10.0|0.0|
|pes|0.67|0.12|939.5|436.5|
|plt|0.01|0.00|13.1|7.5|
|pnb|0.07|0.07|57.3|42.8|
|pol|0.01|0.00|12.0|6.1|
|por|0.02|0.02|31.8|34.4|
|pus|0.20|0.13|165.8|91.3|
|ron|0.02|0.00|19.1|12.6|
|rus|0.03|0.02|31.5|30.6|
|san|0.03|0.02|32.0|42.6|
|sin|0.03|0.02|53.3|53.1|
|slk|0.02|0.02|17.8|21.5|
|slv|0.01|0.01|17.8|17.9|
|som|0.02|0.01|22.3|28.5|
|spa|0.03|0.05|43.4|61.5|
|sqi|0.01|0.00|16.5|10.1|
|srp|0.02|0.01|15.3|11.6|
|sun|0.57|0.08|711.4|224.6|
|swa|0.02|0.01|21.2|27.0|
|swe|0.02|0.01|19.7|18.9|
|tam|0.02|0.00|23.3|23.1|
|tat|0.25|0.24|152.6|144.0|
|tgk|0.05|0.07|56.2|74.7|
|tgl|0.02|0.01|22.8|34.8|
|tha|0.03|0.01|40.0|25.8|
|tur|0.01|0.00|16.9|14.7|
|ukr|0.02|0.01|13.7|10.1|
|urd|0.10|0.12|80.7|98.5|
|uzb|0.02|0.01|19.9|20.0|
|vie|0.02|0.01|16.7|23.1|
|vol|0.01|0.00|14.7|19.4|
|war|0.02|0.01|22.5|31.2|
|zul|0.02|0.01|22.1|34.0|
> Consider switching lang-detection in tika-eval to open-nlp
> ----------------------------------------------------------
>
> Key: TIKA-2790
> URL: https://issues.apache.org/jira/browse/TIKA-2790
> Project: Tika
> Issue Type: Improvement
> Reporter: Tim Allison
> Priority: Major
> Attachments: fra_mixed_100000_0.0_0.txt, langid_20190509.zip,
> langid_20190510.zip, langid_20190514.zip, langid_20190514_plus_minus_1.zip
>
>
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
