LanguageTool Docker

[Telos]

Just wanted to call attention to a resource by one our contributors @Dr_Frankenstein ...

LanguageTool provides a browser spell check feature, as well as some grammar checks, and offers a synonym feature. But rather than share all your post content with that organization, you can run the tool on your NAS.

I found his post, easy to follow. The only modification I made was to drop the network host mode. FWIW, my docker-compose file follows...

version: "3.8"
services:
  languagetool:
    image: erikvl87/languagetool:latest
    container_name: languagetool
    # network_mode: host
    ports:
        - 8010:8010
    environment:
        - langtool_languageModel=/ngrams
        - Java_Xms=512m
        - Java_Xmx=1g
    volumes:
        - /volume1/docker/languagetool/ngrams:/ngrams
restart: unless-stopped

 

[SynoMan]

I'm using this already for some time, and it works perfect. I ditched Grammarly for this. Thanks to @Dr_Frankenstein!

 

[Huggy]

I'm using this already for some time, and it works perfect. I ditched Grammarly for this. Thanks to @Dr_Frankenstein!

Same here

 

[Danabw]

How do you enable it in your browser once it's started? I'm using Chrome.

 

[SynoMan]

Install Chrome extension, click on it and choose More advanced settings (cog wheel). Scroll down to Experimental settings (Only for advanced users). Select Other server and fill in the field.

 

[Telos]

How do you enable it in your browser once it's started?

That's described in the blog post I linked in #1 post.

 

[Danabw]

That's described in the blog post I linked in #1 post.

Thanks...navigation to his website wasn't clear to me, initially I clicked in the link and it took me to his "Latest Activity" page here on this site. After your comment I poked about a little more and found that it was just a hover on the link and the dialog w/the link to his website popped up. Learn something new every day, right?!

 

[Telos]

I clicked in the link and it took me to his "Latest Activity" page

Wrong link. Click the link associated with the word "post".

 

[Danabw]

Wrong link. Click the link associated with the word "post".

Wow...never saw that link...on my screen the "post" link just blends too well w/the text near it so I didn't see it, the link highlighting is too weak. The older my eyes get, the more they toy with me... ;-)

 

[Dr_Frankenstein]

Howdy - just noticed the @!

Glad my tutorial was useful - Need to update my profile on here!

 

[goldcoins]

Hello, just tried setting it up, I received this error in my logs. Any ideas how to fix it? Running 6.2.3

Error: Internal Error: java.lang.RuntimeException: org.apache.lucene.index.CorruptIndexException: checksum failed (hardware problem?) : expected=1cd4a2d4 actual=c26f8237 (resource=BufferedChecksumIndexInput(MMapIndexInp

 

[goldcoins]

Hello, just tried setting it up, I received this error in my logs. Any ideas how to fix it? Running 6.2.3

Error: Internal Error: java.lang.RuntimeException: org.apache.lucene.index.CorruptIndexException: checksum failed (hardware problem?) : expected=1cd4a2d4 actual=c26f8237 (resource=BufferedChecksumIndexInput(MMapIndexInp

Just a quick update, I tried installing version 4.9.1 and that worked fine - no more errors.

 

[Dr_Frankenstein]

Hey, it looks like there is an issue with the latest version

Checksum error since 5.7 · Issue #38 · Erikvl87/docker-languagetool

Hi, Using the container linux/amd64 (5.11.0-49-generic #55-Ubuntu SMP Wed Jan 12 17:36:34 UTC 2022 x86_64 x86_64 x86_64 GNU/Linux) latest output a checksum error related to ngrams (whichever the la...

github.com

You should be able to roll to the version suggested. It seems to only affect people setting up for the first time as running 5.7 OK from an upgrade.

 

[goldcoins]

Thanks for the quick reply - I just tried 5.6 update 3 and that worked fine as suggested.

 

[Telos]

Recently, I've had occasions where the underlined "misspellings" reveal "corrections" that may be in French, or sometimes Italian (despite English as the chosen language).

It happens only briefly, and then reverts to the correct language.

I reached out to the LanguageTool folks, and their reply was to the effect...

It seems you are using your own LanguageTool server.
Make sure to configure “FastText” to improve language detection.
If you google for “LanguageTool FastText” you will find plenty of resources describing how to set it up.

Overall, I found this vague and unhelpful.

Has anyone experienced this?

 

[Dr_Frankenstein]

Recently, I've had occasions where the underlines "misspellings" reveal "corrections" that may be in French, or sometimes Italian (despite English as the chosen language).

It happens only briefly, and then reverts to the correct language.

I reached out to the LanguageTool folks, and their reply was to the effect...

It seems you are using your own LanguageTool server.
Make sure to configure “FastText” to improve language detection.
If you google for “LanguageTool FastText” you will find plenty of resources describing how to set it up.

Overall, I found this vague and unhelpful.

Has anyone experienced this?

Hey, so their instructions were to Google it lol, I have had this same issue with random languages coming up on what to me was clearly English.

So I just did a bit of Googling I think these are the first steps – just trying them now… but calling it a night shortly so will be trying more tomorrow!

Go here

Get started · fastText

## What is fastText?

fasttext.cc

Follow the instructions for building the commandline version, I did this on a PopOS (Ubuntu) laptop which will generate a load of files you just need the 'fasttext' binary its about 428kb (I can share this if you need it)

Then grab the language identification binary from here Language identification · fastText the 126mb version.

Next steps are currently unclear....BBL

 

[Telos]

Here's a recent example...

 

[one-eyed-king]

From what I understand, the fasttext binary and model need to be mapped into the container as volume and configured in the configuration file like this:

fasttextModel=/mount/path/in/container/lid.176.bin
fasttextBinary=/mount/path/in/container/fastText

I could try to build a custom image with the preinstalled fasttext binary based on erikvl87/languagetool and publish it on dockerhub. It should be easy to add environment variables to define which language packs to be downloaded and wether fasttext should be enabled or not.

 

[Telos]

Wondering what the point of this is...

Or, is that not implemented in the Docker image.

 

[one-eyed-king]

I have no idea. Does it allow to install the ngrams? I haven't used languagetool yet, I just stumbled accross Language Tool and though, it can't be that hard to extend the existing image to support fasttest.

if it was not directed to me, just ignore this post