Adjustment of goal-driven resolution for natural language processing in TIL

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F61989100%3A27240%2F19%3A10243177" target="_blank" >RIV/61989100:27240/19:10243177 - isvavai.cz</a>

Výsledek na webu

<a href="https://nlp.fi.muni.cz/raslan/raslan19.pdf" target="_blank" >https://nlp.fi.muni.cz/raslan/raslan19.pdf</a>

DOI - Digital Object Identifier

—

Jazyk výsledku

angličtina

Název v původním jazyce

Adjustment of goal-driven resolution for natural language processing in TIL

Popis výsledku v původním jazyce

The paper deals with natural language reasoning and question answering. Having a fine-grained analysis of natural language sentences in the form of TIL (Transparent Intensional Logic) constructions, we apply the General Resolution Method (GRM) with its goal-driven strategy to answer the question (goal) raised on the natural language data. Not only that, we want to answer in an &apos;intelligent&apos; way, so that to provide logical consequences entailed by the data. From this point of view, GRM appears to be one of the most plausible proof techniques. There are two main new results presented here. First, we found out that it is not always possible to apply all the necessary adjustments of the input constructions first, and then to go on in a standard way by applying the algorithm of the transformation of propositional constructions into the Skolem clausal form followed by the GRM goal-driven resolution techniques. There are plenty of features special for the rich natural language semantics that are dealt with by TIL technical rules and these rules must be integrated with the process of the goal-driven resolution technique rather than separated from it. Second, the strategy of generating resolvents from a given knowledge base cannot be strictly goal-driven. Though we start with a given goal/question, it may happen that there is a point at which we have to make a step aside. We have to apply those special TIL technical rules on another clause first, and only then it is possible to go on with the process of resolving clauses with a given goal. Otherwise our inference machine would be heavily underinferring, which is not desirable, of course. We demonstrate these new results by two simple examples. The first one deals with property modifiers and anaphoric references. Anaphoric references are dealt with by our substitution method, and the second example demonstrates reasoning with factive verbs like &apos;knowing&apos; together with definite descriptions and anaphoric references again. Since the definite description occurs de re here, we substitute a pointer to the individual referred to for the respective anaphoric pronoun.

Název v anglickém jazyce

Adjustment of goal-driven resolution for natural language processing in TIL

Popis výsledku anglicky

The paper deals with natural language reasoning and question answering. Having a fine-grained analysis of natural language sentences in the form of TIL (Transparent Intensional Logic) constructions, we apply the General Resolution Method (GRM) with its goal-driven strategy to answer the question (goal) raised on the natural language data. Not only that, we want to answer in an &apos;intelligent&apos; way, so that to provide logical consequences entailed by the data. From this point of view, GRM appears to be one of the most plausible proof techniques. There are two main new results presented here. First, we found out that it is not always possible to apply all the necessary adjustments of the input constructions first, and then to go on in a standard way by applying the algorithm of the transformation of propositional constructions into the Skolem clausal form followed by the GRM goal-driven resolution techniques. There are plenty of features special for the rich natural language semantics that are dealt with by TIL technical rules and these rules must be integrated with the process of the goal-driven resolution technique rather than separated from it. Second, the strategy of generating resolvents from a given knowledge base cannot be strictly goal-driven. Though we start with a given goal/question, it may happen that there is a point at which we have to make a step aside. We have to apply those special TIL technical rules on another clause first, and only then it is possible to go on with the process of resolving clauses with a given goal. Otherwise our inference machine would be heavily underinferring, which is not desirable, of course. We demonstrate these new results by two simple examples. The first one deals with property modifiers and anaphoric references. Anaphoric references are dealt with by our substitution method, and the second example demonstrates reasoning with factive verbs like &apos;knowing&apos; together with definite descriptions and anaphoric references again. Since the definite description occurs de re here, we substitute a pointer to the individual referred to for the respective anaphoric pronoun.

Druh

D - Stať ve sborníku

CEP obor

—

OECD FORD obor

10201 - Computer sciences, information science, bioinformathics (hardware development to be 2.2, social aspect to be 5.8)

Projekt

Výsledek vznikl pri realizaci vícero projektů. Více informací v záložce Projekty.

Návaznosti

P - Projekt vyzkumu a vyvoje financovany z verejnych zdroju (s odkazem do CEP)<br>S - Specificky vyzkum na vysokych skolach

Rok uplatnění

2019

Kód důvěrnosti údajů

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Název statě ve sborníku

Thirteenth workshop on Recent Advances in Slavonic Natural Language Processing, RASLAN 2019 : Karlova Studánka, Czech Republic, December 6-8, 2019 : proceedings

ISBN

978-80-263-1530-8

ISSN

2336-4289

e-ISSN

—

Počet stran výsledku

12

Strana od-do

71-82

Název nakladatele

Tribun EU

Místo vydání

Brno

Místo konání akce

Karlova Studánka

Datum konání akce

6. 12. 2019

Typ akce podle státní příslušnosti

EUR - Evropská akce

Kód UT WoS článku

000604899800009