1 UNCERTAINTY OF TEXTUAL INFORMATION: LINGUISTICAL PRESUPPOSITION FOR INFORMATION SYSTEM Omar LAROUK ELICO-SII – ENS SIB université de Lyon, France. - презентация

1 1 UNCERTAINTY OF TEXTUAL INFORMATION: LINGUISTICAL PRESUPPOSITION FOR INFORMATION SYSTEM Omar LAROUK ELICO-SII – ENS SIB université de Lyon, France

2 FRAMEWORK PRESENTATION - IMPORTANT PRODUCTION OF TEXTS -IMPORTANT RETRIEVAL OF TEXTS BY USERS INFORMATION AIM OF PROJECT :LOOK FOR A SYSTEM WHO: -TO AUTOMATE MANUAL PROCESS OF INDEXING -TO STRUCTURE THE INFORMATION -TO PERMIT TO OBTAIN INFORMATION

3 5 FRAMEWORK PRESENTATION (2) METHOD : Elaboration of semi-structured DATABASES from Elaboration of semi-structured DATABASES from informations extracted from the textual data Use of linguistic techniques for processing documents (textual data). Search strategies that allow the user to exploit documents.

4 6 DESIGN OF IRS STRUCTURED or NOT ? TWO DIFFERENTS METHODS : - A priori determination of concerned universe (classification, thesaurus, IA) - Not a priori structured universe ( statistical and linguistic approaches)

5 7 Information retrieval on the database INDEXIND IS A EXTRACTION PROCESS Integration of a document in a collection Indentification, in a document, of expressions describing the content of this document. Information Retrieval on the database is a linguistic problem ( problem of ambiguous terms)

6 Documentary Information system Documentary automatization is blocked by the problem of indexing and interrogation Use natural language as textual data for automatic indexing because language integrates the contextual and temporal factor through connectors, verbs, adverbs, etc.. We use the LINGUISTICAL APPROACH because they are lot off incertainty of texts

7 Information Retrieval and presupposition the state of information with a statement carrying a presupposition can be seen as a combination of two things: –i) Check if presupposition holds in the current context, –ii) The update of the current state of the information of data bases in link with the informative contents of the presupposed statement.

8 Information Retrieval and presupposition We think that presupposition is a presupposition of the speaker in a context at first of indexation of the document (during the creation) as if the designer pronounce a contextual sentence. The absence of words presupposed in the answer given by the server (search engines) can be interpreted as the lack at the truth of this sentence in this context of indexation.

9 definition of the presupposition For example : (Russell 1920, Gazdar,1979) /John says that the king of France is bald/ The sentence (utterance): John says that the king of France is bald has two potential presuppositions: –- There is someone identified the noun phrase : « John ». –- There is a « king of France ». Of these two, only the presupposition that there is someone identified as…( John is an actual presupposition)

10 definition of the presupposition /my friend will pass by this evening/ = [ps]=[presupposed] " I have a friend = [posed] "my friend will pass by this evening is the sentence is the presupposition = is the sentence [presupposed] is the sentence [posed]

11 presupposition (test of negation and interrogation) The recognition criterion of the DUCROT presupposition is therefore (1980): A presuppose B if A B and if (¬A ) B or (?A) B The presupposition is the non refutable information included in the sentence. The posed is the "explicit" information

12 Interrogation test for the presupposition When we are looking for document on the web or in database, the interrogation and negation tests are effective. For example we have (¬x) and (?x) : "will my friend pass by this evening?" "my friend will not pass by this evening"

] if [ ]=" title="Table of presupposition (extension of Strawson model) : /The president goes on presiding the ceremony/ = [ps] "The president had presided the ceremony" =[posed] "The president still presides the ceremony presupposes (with =, = ) [ ---ps---> ] if [ ]=" class="link_thumb"> 13 Table of presupposition (extension of Strawson model) : /The president goes on presiding the ceremony/ = [ps] "The president had presided the ceremony" =[posed] "The president still presides the ceremony presupposes (with =, = ) [ ---ps---> ] if [ ]=[true] and [ ]=[True] so is the presupposition If the presupposition is false then the sentence is neither true nor false. There is therefore a blank in the truth table. Hence the application of a logic with more than two values (trivalent logic). ] if [ ]="> ] if [ ]=[true] and [ ]=[True] so is the presupposition If the presupposition is false then the sentence is neither true nor false. There is therefore a blank in the truth table. Hence the application of a logic with more than two values (trivalent logic)."> ] if [ ]=" title="Table of presupposition (extension of Strawson model) : /The president goes on presiding the ceremony/ = [ps] "The president had presided the ceremony" =[posed] "The president still presides the ceremony presupposes (with =, = ) [ ---ps---> ] if [ ]=">

14 PRESUPPOSITION LOGIC or MULTIVALENT LOGIC In multivalent logic, the implication (A B) does not mean the same as the disjunction (not-A or B) so that the notion of tautology must be redefined : are these formulæ always taking the value "True", never taking the value "True" or taking a value in a sub set of the set of truth values [1,0]? Let us consider logic with three values denoted [False, undetermined, True]=[0,u,1]. This trivalent logic presents the following truth value tables for the connector ( AND, ).

15 Multivalent logic Implication (A B ) B 1 0 u ? A ? u ? ? ? The question mark (?) indicates that the truth tables remain to be completed for this trivalent logic. We suggests using four intermediate values : unknown, absurd, undetermined and unstable.

16 Table of presupposition ( extension de model of Strawson ) : presuppose if ( implies ) and if true so is the presupposition LinesPP1P P1 P1P1 L L L31u ? L L L60u ? L7u1 ?1 L8u0 ?0 L9uu ? Fig. 1: Truth value table of the presupposition

17 Intensional logic and symmetric connectors Let us analysis another form of coordination : /the flag is black and white/ By applying the calculation to example the following solutions are obtained : /the flag is black/ /the flag is white/ /the flag is black and white/ We can say that for example the inferences and are unacceptable in the discourse. The only solution is that which combines the two predicates "black*white".This is the combinatory concept of speech.

18 Multivalent logic and symmetric connector If we applied the calculus of distribution for the following example, the distributiviness give : = /the red and black flag/ ° = /the red flag/ ° = /the black flag/ =>P3>/the red *black flag/ We are going to applied this logic to connectors with the following criteria :

19 Presupposition and symmetric connector Presupposition is a domain where multivalent logic aids analysis. It is necessary to demonstrate situations in which the dichotomy true/false turns out to be insufficient. We shall see that the evaluation of a proposition will be conditioned by the relational links (presupposition). – /the red and black flag / ° /the red flag / ° /the black flag / /the red*black flag/ presupposes if [ > ]=true and =True

20 Presupposition and symmetric connector /the guard saw the fire AND (he) gave the alert/ A Connector B Here the connector "AND" is not a distributive connector but a combining connector which can be assimilated to a combining operator of predicates "AND-asymmetric"But, the connector "AND-asymmetric" presents aspects linked to the general context of the sentence and transmits the facts of presupposition. We shall see how no-classical logic can explain this dynamic phenomenon.

21 Implicit information : Calculation of the presupposed elements At the man/system interface based on knowledge, interrogation of any system is not optimal due to the forgotten information when the user formulates his query. So, the presupposition is the hidden part of the information. This implicit information can be reconstituted (and therefore presupposed) by the designer of database and the user. However, this transmission of information remains very complex in the framework of a man/machine interface.

22 CONCLUSION (1) The whole problem of performance is to optimise the existing links between the configuration of the system and the user needs, and better understand the relationships between the notion of presupposition and queries in natural language.

23 CONCLUSION (2) The aim of an IRS for the user is to retrieve information stored in a database, but also to interact, hence the necessity to take this final objective into account i.e. consulting the base using a query in natural language. However, the processing of this query comes within the scope of automatic language processing. Natural Language possesses a "natural logic" which links it to presupposition logic.

24 Diapo1 : Nous pensons que la présupposition est une présupposition de l'orateur{*speaker*} dans un contexte d'abord d'indexation du document (pendant la création) comme si le designer prononce une phrase{*sentence*} contextuelle. L'absence de mots présupposés dans la réponse donnée par le serveur (des moteurs de recherche) peut être interprétée comme le manque dans la vérité de cette phrase{*sentence*} dans ce contexte d'indexation. Diap2-IRS-pp :On peut voir l'état d'information avec une déclaration portant une présupposition comme une combinaison de deux choses : –i) Vérifier si la présupposition se tient dans le contexte actuel, _ ii) la mise à jour de l'état actuel de l'information de données se base dans la liaison avec le contenu informatif de la déclaration présupposée. Diap3 : À l'interface d'homme / système basée sur la connaissance, l'interrogation de n'importe quel système n'est pas optimale due à l'information oubliée quand l'utilisateur formule sa question. Ainsi, la présupposition est la partie cachée de l'information. Cette information implicite peut être reconstituée (et présupposée donc ) par le designer de base de données et l'utilisateur. Cependant, cette transmission de restes de l'information très complexes dans la structure d'une interface d'homme / machine. Diap4: Le but d'un FISC pour l'utilisateur est de récupérer l'information stockée dans une base de données, mais aussi agir réciproquement, de là la nécessité pour prendre cet objectif final en considération c'est-à-dire la consultation de la base employant une question en langage naturel. Cependant, le traitement de cette question vient dans les limites du traitement de langue automatique. Le langage naturel possède "une logique naturelle" qui le lie avec la logique de présupposition.