FRIDA

Revision as of 02:31, 31 March 2014

The project FRIDA (Français, Rumantsch, Italiano and Deutsch for Analysis) is devoted to the creation of NL-UNL (analysis) dictionaries for the official languages of Switzerland.

Goal

The project FRIDA has two main goals:

1. To provide several word-to-concept monolingual databases (i.e., encoding or reader's dictionaries). These dictionaries are expected to be used in UNLization, i.e., in generating UNL graphs out of natural language documents, especially through IAN.
2. To find concepts that are not enclosed in the WordNet3.0 and should be incorporated to the UNL Dictionary.

Repository

The whole FRIDA contains, for each language, 30,000 lemmas, and is divided into 6 different repositories according to the frequency of use of lemmas.

• FRIDA-A1 contains the list of the 5,000 most frequent lemmas of the language (including articles, prepositions, conjunctions, auxiliary verbs, etc.);
• FRIDA-A2 contains the list of the following 5,000 most frequent lemmas of the language (including articles, prepositions, conjunctions, auxiliary verbs, etc.);

And so on, up to FRIDA-C2, according to the table below.

Participants

• Anne-Sophie Gloor (French, Université de Genève)
• Anton Maria Prati (Italian, Università Ca' Foscari Venezia)
• Frank Brockmeier (German, Université de Lausanne)
• Katrin Renkwitz (German, Rheinische-Friedrich-Wilhelms-Universität Bonn)
• Myriam Hilout (French, Université Paris XIII)
• Monica Gallo (Italian, Université de Genève)

Instructions

Goal

In the project FRIDA, your main goal is:

• to map the natural language entry to all its possible senses (which are represented by different UW's)
• to provide all the lexical and morphological information concerning the entry (by filling in the corresponding form)

Homonyms

Most natural language entries are semantically ambiguous and may have many different senses. If these different senses do not affect the morphological behavior of the entry, there is no problem and you may provide them all within the same entry. However, sometimes a change in the sense leads to a change in the morphological features of the word (the entry is feminine with the meaning X, and masculine with the meaning Y, for instance; or it's a common noun with the meaning X and a proper noun with the meaning Y). In this case, you have first to assign, to the entry, all possible senses (i.e., you should link the entry to all possible UW's, without caring about the morphological behavior). Next, you have to SPLIT the entry (using the button at the end of the form) and GROUP those that have the same morphological behavior. The entries within the same group may have different senses, but will share the same lexical features. Do not forget to address all groups.

Empty, wrong and incomplete entries

You will find, in the project FRIDA, three different types of entries:

• Entries that are empty, i.e., which have not been mapped yet to any UW and do not contain any information. In this case, you have to link it to as many UW's as possible, and provide all the necessary information.
• Entries that are CORRECT but INCOMPLETE, i.e., which have not been mapped to all possible UW's or that have not been associated yet to inflectional paradigms or subcategorization frames. In this case, you should add the missing mappings and information.
• Entries that are INCORRECT, i.e., which have been associated to wrong UW's or to the wrong paradigms and frames. In this case, you should correct the information and provide any additional mappings, if necessary.

Lexical Category

Whenever the lexical category for a given lemma is provided, check whether it is correct. If it is not correct, decline the entry and report the problem by clicking over the yellow triangle at the right of the main entry. If the lexical category is not provided, select the most likely category.

Lemma

Do not change the lemma. If it is not correct (i.e., if it is misspelled or cannot be considered to be a lexical unit), decline the entry and report the problem by clicking over the yellow triangle at the right of the main entry.

UW

One of your tasks is to map natural language entries to UW's, i.e., to link it to all possible senses that the entry may have. Provide as many UW's as necessary to each lemma, but do not include very rare or unusual cases. And check the order: the most likely senses must appear first.

The list of lemmas contain only citation forms, i.e., the word as it normally appears in ordinary monolingual dictionaries. You must remember that ordinary dictionaries do not bring inflections. If you find a inflected form as a lemma (such as "glasses", for instance), note that this is not an inflection, but a new lemma/entry^[1]. You should never associate inflected forms to the UW's corresponding to the non-marked terms (i.e., do not link "glasses" to the senses of "glass", but only to the specific senses of "glasses"). The same will happen in gender: do not link the feminine to a masculine ("africana", in Italian, or "religieuse", in French, are not the feminine forms for "africano" or "religieux" - they are different entries with specific meanings^[2]). In case you cannot find any specific meaning for the inflected form, please decline the entry.

If, by change, the sense that you are searching for does not exist yet in the UNL Dictionary, you must propose a new UW, by filling in the form that appears when your search for a given UW does not provide any output.

Base Form

The Base Form must be the same as the lemma. Do not change it.

Inflection

Select AND TEST the inflectional paradigm that generates the inflections of the base form. Any errors here will be propagated to the dictionary, so be careful. In case of no applicable inflection paradigm, select the option IRREGULAR or NON-EXISTENT, depending on the case.

Subcategorization

Subcategorization is only required when the word REQUIRES a complement or a specifier (indirect transitive verbs that select an specific preposition, for instance). In this case, you have to inform the corresponding subcategorization frame. In case of no applicable subcategorization frame, select the option IRREGULAR or NON-EXISTENT, depending on the case.

Register

You should inform the register only if the entry does not belong to the standard language, i.e., if it is not used in written form (colloquialism, such as "wanna"), or if it not used anymore (archaism, such as "thou"), or if it is used only in technical domain (jargon, such as "Canis lupus familiaris"), etc.

Region

You should inform the country only if the entry is used only in a specific country, and not among the whole community of the language.

Frequency

Leave 0 (this information is extracted later from corpus)

Priority

Leave 0 (this information is extracted later from corpus)

Other features

This field is used only when the UW requires some attributes. This is very rare and should be avoided.

Comments

You may leave your doubts or problems here, in order for us to discuss them later, but they must be provided in English, because the discussion will involve the whole group.

Notes

1. ↑ Although originally associated to "glass", "glasses" is today a completely independent word
2. ↑ "africana" is the set of books, documents, artifacts, artistic works, etc., reflecting or concerned with African history, life, or culture; and "religieuse" is a type of pastry.