Split DELAS into DELAS and DELAS-Pr

[ppKrauss]

There are a lot of "pure named entity" as proper nom, that are not real "dictionary words".

Examples: abel,N004+Pr, abelson,N004+Pr, abélson,N004+Pr, abigail,N104+Pr, abília,N104+Pr, abílio,N004+Pr, abraão,N004+Pr, abraham,N004+Pr, abrantes,N306+Pr, abrão,N004+Pr, zico,N004+Pr, zilda,N104+Pr, zimbábue,N304+Pr, zingarelli,N306+Pr, zoroastro,N004+Pr, zucolotto,N306+Pr, zurique,N104+Pr

Many are usual human given names (modern as zico, zilda or classic as zoroastro) or surnames (zingarelli, zucolotto). Other are commom toponyms, as country names, city names (abrantes,zurique), etc.

So, at DELAS-pr must include a column indicating the type of entity where the name is usually used (ex. Italy is a country-name but in Brasil there is also a female name).

There are other sources of names and its use-statistics, see here datasets-br/prenomes or datasets-br/city-codes, for confirmed Brazilian names, and world-cities, etc. for international.

[ppKrauss]

From examples:

graph_id	classifications
N004	ms
N104	:fs
N304	:ms:fs
N306	:ms:mp:fs:fp

---

If all ok... Lets split!

Spliting with grep:

1. check wc -l DELAS.csv (75106)
2. grep -E "(Pr|SIGL)" DELAS.csv > DELAS_Pr.csv
   check 4898
3. prepend... DELAS_Pr.csv
4. grep -vwE "(Pr|SIGL)" DELAS.csv > /tmp/DELAS_noPr.csv; mv /tmp/DELAS_noPr.csv DELAS.csv
   check 70208
5. check 70208+4898=75106?

...

Or by SQL...

• Check with SELECT DISTINCT regexp_replace(graph_id,'^[^\+]+\+','') FROM dataset.vw2_delas WHERE graph_id like '%+%' shows:
  Dem, PosXTra, Art+Ind, Ind, Pr, Pos, Art+Def, Num, Tra, Pes, DemXInd, Int, Rel.
  No SIGL.