tabeltekstilo

tabeltekstilo is a multi-purpose tool for manipulating text in tabular data format.

introduction

text in tabular data format is text formatted as a table (usually stored as a spreadsheet file), where each row of the table contains one word of the text. one column contains the actual word as it appears in the text, while other columns may contain more information about the word, like the page and line number where it appears, a cleaned-up form (with uniform casing and no punctuation), its lemma, its grammatical category,…

from text in that format, tabeltekstilo can generate:

• an alphabetical dictionary by aggregating columns
• an alphabetical index (like the ones that appear at the end of books)
• an xml file

see the examples section below for concrete examples.

features

general features:

• alphabetical sorting using the unicode collation algorithm
• right-to-left text support

dictionary features:

• dictionary generation
• multiple column aggregation with custom join string

index features:

• multi-level index generation
• multiple values in parent columns support for agglutinated forms
• multiple reference support (for example: page, line)
• grouping of identical references with count
• total count of form occurrences at each parent level
• filtering with regular expressions

xml features:

• nesting of forms under multiple levels of parent columns
• custom attributes on form elements
• custom root element
• optional header with custom copyright and licensing information

usage

tabeltekstilo takes a subcommand, an input filename and an output filename as arguments, as well as some options. input and output files should be in opendocument (.ods) or office open xml (.xlsx) format.

dictionary

the minimal usage is:

tabeltekstilo dictionary --form-col form --agg-col agg input.ods output.ods

where form is title of the column (in input.ods) that contains the form that will appear in the dictionary and agg is the title of the column (in input.ods) that contains the values to aggregate next to the form.

to display a full description of the usage syntax:

tabeltekstilo dictionary --help

index

the minimal usage is:

tabeltekstilo index --ref-col ref --form-col form input.ods output.ods

where ref is the title of the column (in input.ods) that contains the reference to use in the index (the page number, for example) and form is title of the column (in input.ods) that contains the form that will appear in the index.

to display a full description of the usage syntax:

tabeltekstilo index --help

xml

the minimal usage is:

tabeltekstilo xml input.ods output.xml

where input.ods is the input file and output.xml the output file to generate.

to display a full description of the usage syntax:

tabeltekstilo xml --help

examples

dictionary

let’s take the following example text:

le reste des avions vola vers l’est. nous avions du retard. c’est ce qu’il reste des vers à propos des vers.

it must first be converted to this format as input.ods:

word	form	lemma	type
le	le	le (la)	det_art
reste	reste	reste	noun
des	des	de+le (la)	prep+det_art
avions	avions	avion	noun
vola	vola	voler	verb
vers	vers	vers	prep
l’	l’	le (la)	det_art
est.	est	est	noun
nous	nous	nous	pro_per
avions	avions	avoir	verb
du	du	de+le (la)	prep+det_art
retard.	retard	retard	noun
c’	c’	ce	pro_dem
est	est	être	verb
ce	ce	ce	pro_dem
qu’	qu’	que	conjs
il	il	il	pro_per
reste	reste	rester	verb
des	des	de+le (la)	prep+det_art
vers	vers	vers	noun
à	à	à	prep
propos	propos	propos	noun
des	des	de+le (la)	prep+det_art
vers.	vers	ver	noun

now, let’s generate the dictionary by calling:

tabeltekstilo dictionary --form-col form --agg-col lemma --agg-col type input.ods output.ods

this will generate the following table as output.ods:

	form	lemma	type
0	à	à	prep
1	avions	avion; avoir	noun; verb
2	c’	ce	pro_dem
3	ce	ce	pro_dem
4	des	de+le (la)	prep+det_art
5	du	de+le (la)	prep+det_art
6	est	est; être	noun; verb
7	il	il	pro_per
8	l’	le (la)	det_art
9	le	le (la)	det_art
10	nous	nous	pro_per
11	propos	propos	noun
12	qu’	que	conjs
13	reste	reste; rester	noun; verb
14	retard	retard	noun
15	vers	ver; vers; vers	noun; noun; prep
16	vola	voler	verb

index

let’s take the following example text, and say that it appears on line 1 and 2 of page 42:

la suno brilas hodiaŭ. hieraŭ estis malvarme, sed hodiaŭ estas varme.
ni estas bonŝancaj!

it must first be converted to this format as input.ods:

page	line	word	form	lemma
42	1	la	la	la
42	1	suno	suno	suno
42	1	brilas	brilas	brili
42	1	hodiaŭ.	hodiaŭ	hodiaŭ
42	1	hieraŭ	hieraŭ	hieraŭ
42	1	estis	estis	esti
42	1	malvarme,	malvarme	varma
42	1	sed	sed	sed
42	1	hodiaŭ	hodiaŭ	hodiaŭ
42	1	estas	estas	esti
42	1	varme.	varme	varma
42	2	ni	ni	ni
42	2	estas	estas	esti
42	2	bonŝancaj!	bonŝancaj	bona+ŝanco

now, let’s generate the index by calling:

tabeltekstilo index --ref-col page --ref-col line --parent-col lemma --form-col form --split-char + input.ods output.ods

this will generate the following table as output.ods:

	lemma_count	lemma	form_count	form	refs
0	1	bona	1	bonŝancaj	42, 2
1	1	brili	1	brilas	42, 1
2	3	esti	2	estas	42, 1; 42, 2
3			1	estis	42, 1
4	1	hieraŭ	1	hieraŭ	42, 1
5	2	hodiaŭ	2	hodiaŭ	42, 1 (2)
6	1	la	1	la	42, 1
7	1	ni	1	ni	42, 2
8	1	ŝanco	1	bonŝancaj	42, 2
9	1	sed	1	sed	42, 1
10	1	suno	1	suno	42, 1
11	2	varma	1	malvarme	42, 1
12			1	varme	42, 1

note that “bonŝancaj” appears twice in the index, once under the form “bona” and once under the form “ŝanco”. this is because the lemma column contained two values, separated by the defined split character.

note that the word “hodiaŭ” appears twice on the same line. this is why its reference has “(2)” appended to it.

filtering

the tabeltekstilo index function allows to filter rows based on column values using regular expressions.

for example, using the same input file as in the previous example, let’s say that only noun lemmas should appear. in this case, they all end with “o”, so this command can be used:

tabeltekstilo index --ref-col page --ref-col line --parent-col lemma --form-col form --split-char + --filter "lemma:.*o" input.ods output.ods

in this example, the argument is quoted to avoid the * character to be interpreted by the shell. this depends on the shell used.

this will generate the following table:

	lemma_count	lemma	form_count	form	refs
0	1	ŝanco	1	bonŝancaj	42, 2
1	1	suno	1	suno	42, 1

note that “bonŝancaj” appears only once in this case, because the lemma “bona” was filtered out.

multiple filter arguments may be used. the format of the filter expressions is col:regex, where col is a column name and regex is a regular expression matching the value (after splitting). any column of the input table can be used, even those not used by the index.

by default, filtering is inclusive, which means that at least one expression should match for the row to be included. this behavior can be reversed with --filter-exclude. in this case, any row matching an expression is excluded; only the rows not matching any of the expressions are included.

for example, still using the same input file, let’s say that forms with less than 4 letters should be excluded. this command can be used:

tabeltekstilo index --ref-col page --ref-col line --parent-col lemma --form-col form --split-char + --filter "form:.{1,3}" --filter-exclude input.ods output.ods

this will generate the following table:

	lemma_count	lemma	form_count	form	refs
0	1	bona	1	bonŝancaj	42, 2
1	1	brili	1	brilas	42, 1
2	3	esti	2	estas	42, 1; 42, 2
3			1	estis	42, 1
4	1	hieraŭ	1	hieraŭ	42, 1
5	2	hodiaŭ	2	hodiaŭ	42, 1 (2)
6	1	ŝanco	1	bonŝancaj	42, 2
7	1	suno	1	suno	42, 1
8	2	varma	1	malvarme	42, 1
9			1	varme	42, 1

tabeltekstilo uses python’s regular expressions. their documentation is here.

xml

let’s take the table from the index command example, but change it slightly by renaming the first 2 columns and putting the word column last :

page p	line l	form	lemma	word
42	1	la	la	la
42	1	suno	suno	suno
42	1	brilas	brili	brilas
42	1	hodiaŭ	hodiaŭ	hodiaŭ.
42	1	hieraŭ	hieraŭ	hieraŭ
42	1	estis	esti	estis
42	1	malvarme	varma	malvarme,
42	1	sed	sed	sed
42	1	hodiaŭ	hodiaŭ	hodiaŭ
42	1	estas	esti	estas
42	1	varme	varma	varme.
42	2	ni	ni	ni
42	2	estas	esti	estas
42	2	bonŝancaj	bona+ŝanco	bonŝancaj!

now, let’s generate the xml file by calling:

tabeltekstilo xml input.ods output.xml

this will generate the following file as output.xml:

<?xml version="1.0" encoding="utf-8"?>
<document>
  <page p="42">
    <line l="1">
      <word form="la" lemma="la">la</word>
      <word form="suno" lemma="suno">suno</word>
      <word form="brilas" lemma="brili">brilas</word>
      <word form="hodiaŭ" lemma="hodiaŭ">hodiaŭ.</word>
      <word form="hieraŭ" lemma="hieraŭ">hieraŭ</word>
      <word form="estis" lemma="esti">estis</word>
      <word form="malvarme" lemma="varma">malvarme,</word>
      <word form="sed" lemma="sed">sed</word>
      <word form="hodiaŭ" lemma="hodiaŭ">hodiaŭ</word>
      <word form="estas" lemma="esti">estas</word>
      <word form="varme" lemma="varma">varme.</word>
    </line>
    <line l="2">
      <word form="ni" lemma="ni">ni</word>
      <word form="estas" lemma="esti">estas</word>
      <word form="bonŝancaj" lemma="bona+ŝanco">bonŝancaj!</word>
    </line>
  </page>
</document>

the parent columns are identified by the fact that they contain a space character which separates the element name from the attribute name.

the last column is used as the deepest element, with all non-parent columns before it used as attributes.

credits

This development was funded by Bastien Kindt for the GREgORI Project.
https://uclouvain.be/fr/instituts-recherche/incal/ciol/gregori-project.html
https://www.v2.gregoriproject.com/
with financial support from
INCAL - Institut des civilisations, arts et lettres
https://uclouvain.be/fr/instituts-recherche/incal
CIOL - Centre d'études orientales - Institut orientaliste de Louvain
https://uclouvain.be/fr/instituts-recherche/incal/ciol