analýza označuje nesprávné pojmy

[marketaadamcova]

Popis problému

Popište:

• vytvořila slovník procházením souboru 'dataset1_filtered' , poté spustila analýzu
• analýza označila pojmy ale u některých (odhadem tak 30%). je přiřazen úplně jiný pojem.
• na první pohled se zdá, že je pojem označen správně (pojem se ve slovníku vyskytuje), ale po klinutí na pojem se zobrazí úplně jiný přiřazený pojem

• https://kbss.felk.cvut.cz/termit-csat/#/
• [slovník komponent a závad - Markéta DP]

Závažnost problému

• popište, jak závažný pro práci s TermItem pro Vás problém je:

2. je to vážný problém,

[blcham]

@ledsoft I wanted to download the annotated document but there is not button to download it. Didn't we have this functionality?

@marketaadamcova On our last meeting about text analysis on skype we discussed following:

• your input text should normal sentences or "all smaller case", not "all capital case" like u have
• your terms should be written small case (pls read every ? where you enter data)

Thus I suggest you to try following experiment with term "sidewall panel":

• make following term:

     label: "sidewall panel"
     synonym: --------------- (leave this empty, you don't need to put plurals there)

• make new document that will have only 2 lines but all lower case, e.g.

5382503 during sidewall panel replacement were found several nutplates loosened /missing. shm - pls perform nutplate replacement

these sidewall panels were found worn out of limit at fwd cargo compartment: 453a1410-7, 453a1410-9.

Additional notes:

• I wanted to experiment on small document because text analysis is very fast in such case
• let me know how it works

[ledsoft]

@blcham Indeed we did. But it was removed by @psiotwo when refactoring documents/files processing. I was pushing for it being supported again, but it was never a priority. Anyway, now that I am the only developer of TermIt, I suppose I can implement it back finally.

[marketaadamcova]

@blcham
result of analysis of original document

result of analysis of lowercase text

any tip how to convert capital case text in xls document to lower case ?

I found another problem
I cannot delete term (i accidentely created term in plural but cannot delete it now)

[blcham]

any tip how to convert capital case text in xls document to lower case ?

see first hit:
https://www.google.com/search?q=excel+convert+capital+to+lowercase&oq=excel+convert+capital+&aqs=chrome.0.0i512j69i57j0i512j0i22i30l4j0i390.6022j0j7&client=ubuntu&sourceid=chrome&ie=UTF-8

Another solution would be to export CSV, change capital->lower case in text editor, put back from CSV to excel.

Text editor that knows how to change capita-->lower case is for example vim :
https://stackoverflow.com/questions/1102859/how-to-convert-all-text-to-lowercase-in-vim#:~:text=If%20you%20want%20to%20convert,the%20character%20under%20the%20cursor.

[blcham]

I cannot delete term (i accidentely created term in plural but cannot delete it now)

@ledsoft what is best way to solve it ? Should I delete it using graphdb or there is a better way ?

[marketaadamcova]

@ledsoft
similar problem with deleting whole vocabulary

[marketaadamcova]

any tip how to convert capital case text in xls document to lower case ?

see first hit: https://www.google.com/search?q=excel+convert+capital+to+lowercase&oq=excel+convert+capital+&aqs=chrome.0.0i512j69i57j0i512j0i22i30l4j0i390.6022j0j7&client=ubuntu&sourceid=chrome&ie=UTF-8

Another solution would be to export CSV, change capital->lower case in text editor, put back from CSV to excel.

Text editor that knows how to change capita-->lower case is for example vim : https://stackoverflow.com/questions/1102859/how-to-convert-all-text-to-lowercase-in-vim#:~:text=If%20you%20want%20to%20convert,the%20character%20under%20the%20cursor.

solved, thanks

[ledsoft]

I cannot delete term (i accidentely created term in plural but cannot delete it now)

@ledsoft what is best way to solve it ? Should I delete it using graphdb or there is a better way ?

@blcham Delete occurrences and assignments of the specified term from the repository. Then it should be possible to remove it via TermIt.

[ledsoft]

@blcham As for the vocabulary - as long as it is not empty, it is not possible to remove it. Either remove all its terms and then remove it. Or, perhaps a quicker solution, remove the whole vocabulary context from the repository (as long as there aren't any references to it from elsewhere).

[blcham]

@marketaadamcova let me know what do you want me to remove, thanks.

[ahmadjana]

@blcham
analysis indicates incorrect concepts or terms (technical perspective): I think because of the word score.
it returns the terms with a higher score.
Ker used to determine how important the word in the documents is by giving a score.
the same for issue #240 and issue #241

[blcham]

@ahmadjana I am not sure exactly what you mean here ... what means incorrect concepts or terms?

We have many concrete examples where it fails. Could you explain it explicitly and in greater detail in those examples why it does not work?