The lexicon lglex is a syntatic lexicon of French verbs and predicative nouns generated by the tool LGExtract (M. Constant and E. Tolone, 2008) from the Lexicon-Grammar tables under LGPL-LR license. It is available both in text and XML format.

________________________________________________________________________________
Description of the content of the lexicon in XML format (element <syn-lexicon>):

Each entry is delimited by the tag <entry> with an attribute id which is the entry identifier generated by concatenating its grammatical category, its class (or table) it comes from and the number of the entry in the table. Information is gathered in four different tags:
- <lexical-info>: lemma and lexical information
- <arguments>: arguments and their nature
- <all-construction>: accepted constructions
- <example>: an illustrative example


1) <lexical-info> contains lexical information corresponding to the entry and has an attribute cat indicating its category ("verb" or "noun"):

For verbs:
- <lexical-value> includes <lemma> the value of which is the lemma of the entry, that can be completed with preverbal pronouns like se, y, en, le, la, les and/or the negation ne ... pas when they are obligatory in this entry (tags <ppvse>, <ppvy>, <ppven>, <ppvle>, <ppvla>, <ppvles>, <neg> with the attribute value "true")
- <aux> indicates the possible auxiliaries accepted by this entry. It  contains a list including the tags <avoir> or <être> with value "true".

For nouns:
- <lexical-value> is composed of element <complete> the value of which is the whole noun (it can be multiword) and the elements <noun1>, <adj1>, <det2>, <noun2> containing the values of its different components. The morphologically derived adjective (resp. verb) might also be indicated in tag <adjassoc> (resp. <vbassoc>).
- <Vsup> describes the support verbs associated with the deverbal noun. It contains <cat> which has always the value "verb" and a list of <value> representing the possible lexical values of the support verb part of basic construction of the entry.
- <Vsup2> is constructed in the same manner and refers to a second list of lexical values for the support verb Vsup2 as it can appear in some converse constructions (e.g. "N1 Vsup2 Det N de (E+la part de) N0").
- <det-modif-list> includes a list of <determiner-modifier> which indicates the distribution of determiners with possible modifiers for the noun. <determiner-modifier> has a tag <det> which indicates the possible determiners separated by '+' (the value <E> stands for the absence of determiner). Tag <modifier> indicates whether the noun accept a modifier with the defined determiners.

For verbs and nouns:
-<prepositions> contains a sequence of <preposition> with an attribute id corresponding to the number of the argument it introduces in the elementary construction (0 for subject, 1 for the first argument, 2 for for the second one, ...). In <preposition>, <prep> represents the different lexical values of the preposition.
For instance, the deverbal noun "allergie" (allergy) in table an01 uses support verb avoir (have) and enters the elementary sentence "N0 Vsup Det N Prép N1". The preposition associated with argument 1 is specified as follows:

<prepositions>
      <preposition id="1">
        <prep value="à" />
      </preposition>
</prepositions>

In the first complément ("Prép N1"), the preposition has the value "à" (e.g. Léa a une certaine allergie à la poussière).

- <locs> defines the locative preposition distributions. It contains a sequence of <loc> with an attribute id refering to the number of the associated argument. Element <loc> also has a list of <prep> representing the possible lexical values of the locative prepositions.


2) <arguments> describes the distribution of the different arguments (subject and complements) of the entry. It includes a set of <constituent> having an attribute pos that indicates the number of the associated argument in the elementary sentence that the entry enters. A constituent is a list of <component>, each of them having the following elements:
- <cat> which has an attribute specifying its syntactic nature: "NP" for noun phrase, "inf" for infinitive (V-inf W), "comp" for a complementizer phrase (Qu P), "leFaitComp" for the noun phrase le fait que P, "siPOuSiP" for the specific complementizer phrase si P ou si P and "adj" for an adjective.

- Various semantic features <hum> (human), <nothum> (non-human), <npr> (proper noun), <abst> (abstract) et <conc> (concrete) with the value "true" when they are verified.

Other semantic features are possible for verbs: <source>, <destination>, <benef> (beneficiary), <mesure> (measure), <prix> (price); and for nouns: <coll> (collective noun) <plur> (obligatory plural).

There also exit several other optional features:
- <mood> with an attribute indicating the mood of the complementizer phrase ("ind" for indicative and "subj" for subjonctive).
- <contr> with an attribute indicating the number of the argument that controls the infinitive. In the previous example, the distribution of the argument at position 1 (i.e. the first complement "Prép N1") is described as follows:

  <arguments>
     ...
    <constituent pos="1">
      <component>
        <cat value="inf" />
        <contr value="0" />
      </component>
      <component>
        <cat value="ceComp" />
        <mood value="ind" />
      </component>
      <component>
        <cat value="ceComp" />
        <mood value="subj" />
      </component>
      <component>
        <cat value="NP" />
        <nothum value="true" />
      </component>
    </constituent>
    ...
  </arguments>

The complement "Prép N1" can be:
- an infinitive controlled by argument 0, i.e. the subject N0: e.g. Léa a une allergie à travailler (Lea is allergic to work). 
- a complementizer phrase in the indicative (Léa a une allergie à ce que nous voyageons)
- a complementizer phrase in the subjective (Léa a une allergie à ce qu'il fasse beau) 
- a noun phrase (see previous example).

For verbs ONLY:
- <origin> contains the list <orig> which indicates the complete name of the columns in the table, that has been used to define the distribution
- <introd-prep> contains a sequence of <prep> providing the prepositions introducing the argument and indicating their lexical values.
- <introd-loc> contains a sequence of <loc> providing the locative prepositions that introduce the argument and indicating their lexical values.


3) <all-constructions> list the different constructions that are accepted by the entry:
- <absolute-constructions> includes a list of absolute <construction> which values are the titles of the column entirely specifying the accepted construction with all its constituents.
For instance, in the construction "N0 V N1 Prép N2", N0 represents the subject, V indicates the verbal entry, N1 is the first complement and Prép N2 is the second one introduced by the preposition Prép.
The construction title is preceded by the string "o::" or "true::" 
if the column associated with the construction has been coded "+" in the table ("o") or if it is a constant "+" in the table of classes ("true"). 
In this last case, the property is verified by all the entries of the table. Such properties are the definitional properties of each classes, including the base construction.
- <relative-constructions> contains the sequence of all relative <construction> which values are the column title of all columns specifying the name of the transformations that are applied on the base construction (ex: "[passif par]").

For nouns ONLY:
- <reductions> describing reductions of the base sentence construction into another construction of another category.
Each reduction is described with an element <type> indicating the syntactic category of the reduction (e.g. noun phrase) and an element <construction> the value of which is the title of the column specifying the reduction construction (e.g. "le N entre N0 et N1").


4) <example> illustrates the entry (solely for verbs)
The value of <example> is an example of sentence with the entry.


______________________________________________________________________________________________
An entry described in text format includes the same information and is represented as follows:

ID=category_tableNumber_entryNumber
lexical-info=[...]
args=(...)
all-constructions=[absolute=(...),
                   relative=(...)]
example=[...]

_____________________________________________________________
An example in the XML format (verb "candidater" in table 33):

<entry id="V_33_24">
  <lexical-info cat="verb">
    <lexical-value>
      <lemma value="candidater" />
    </lexical-value>
    <aux />
    <locs />
    <prepositions />
  </lexical-info>
  <arguments>
    <constituent pos="0">
      <component>
        <cat value="NP" />
        <hum value="true" />
        <origin>
          <orig value="N0 =: Nhum" />
        </origin>
        <introd-prep />
        <introd-loc />
      </component>
    </constituent>
    <constituent pos="1">
      <component>
        <cat value="NP" />
        <nothum value="true" />
        <origin>
          <orig value="N1 =: N-hum" />
        </origin>
        <introd-prep />
        <introd-loc />
      </component>
    </constituent>
  </arguments>
  <all-constructions>
    <absolute-constructions>
      <construction value="true::N0 V à N1" />
      <construction value="true::N1 se V en N2" />
    </absolute-constructions>
    <relative-constructions>
      <construction value="Ppv =: le" />
      <construction value="[extrap]" />
      <construction value="Ppv =: y" />
    </relative-constructions>
  </all-constructions>
  <example>
    <example value="Max a candidaté à un poste" />
  </example>
</entry>

____________________________
Same example in text format:

ID=V_33_24
lexical-info=[cat="verb",verb=[lemma="candidater"],
              aux-list=(),
              prepositions=(),
              locs=()]
args=(const=[pos="0",dist=(comp=[cat="NP",hum="true",origin=(orig="N0 =: Nhum"), introd-prep=(),introd-loc=()])],
      const=[pos="1",dist=(comp=[cat="NP",nothum="true",origin=(orig="N1 =: N-hum"), introd-prep=(),introd-loc=()])])
all-constructions=[absolute=(construction="true::N0 V à N1",construction="true::N1 se V en N2"),
                   relative=(construction="Ppv =: le",construction="[extrap]",construction="Ppv =: y")]
example=[example="Max a candidaté à un poste"]


_______________________________________________________________________________________
Another example in xml format (noun "bise" from table fnan, with support verb "faire"):

<entry id="N_fnan_18">
  <lexical-info cat="noun">
    <lexical-value>
      <complete value="bise" />
      <noun1 value="bise" />
    </lexical-value>
    <Vsup>
      <cat value="verb" />
      <value value="faire" />
    </Vsup>
    <det-modif-list>
      <determiner-modifier>
        <det value="un+une" />
        <modif value="false" />
      </determiner-modifier>
      <determiner-modifier>
        <det value="un+une" />
        <modif value="true" />
      </determiner-modifier>
      <determiner-modifier>
        <det value="la" />
        <modif value="false" />
      </determiner-modifier>
    </det-modif-list>
    <prepositions />
  </lexical-info>
  <arguments>
    <constituent pos="0">
      <component>
        <hum value="true" />
        <cat value="NP" />
      </component>
    </constituent>
    <constituent pos="1">
      <component>
        <hum value="true" />
        <cat value="NP" />
      </component>
    </constituent>
  </arguments>
  <all-constructions>
    <absolute-constructions>
      <construction value="true::N0 Vsup Det N à N1" />
      <construction value="true::N0 Vsup Det N" />
    </absolute-constructions>
    <relative-constructions />
    <reductions />
  </all-constructions>
</entry>

____________________________
Same example in text format:

ID=N_fnan_18
lexical-info=[cat="noun",noun=[notperm=[complete="bise"],noun1="bise"]],
              Vsup=[cat="verb",list=(value="faire")],
              detN=[list-det-modif=(det-modif=[det="un+une",modif="false"],det-modif=[det="un+une",modif="true"],det-modif=[det="la",modif="false"]),
              prepositions=()]
args=(const=[pos="0",dist=(comp=[hum="true",cat="NP"])],
      const=[pos="1",dist=(comp=[hum="true",cat="NP"])])
all-constructions=[absolute=(construction="true::N0 Vsup Det N à N1",construction="true::N0 Vsup Det N"),
                   relative=(),
                   reductions=()]

__________
Reference:

Matthieu Constant and Elsa Tolone. A generic tool to generate a lexicon for NLP from Lexicon-Grammar tables. In Proceedings of the 27th Lexicon-grammar conference, L'Aquila, Italy, 2008.
