Hierarchical Meaning Representation and Analysis of Natural Language Documents

HIERARCHICAL MEANING REPRESENTATION AND ANALYSIS OF NATURAL LANGUAGE DOCUMENTS Toyo-aki NISHIDA and Shuji DOSHITA Department of Information Science Faculty of Engineering, Kyoto University Yoshida-honmachi, Sakyo-ku, Kyoto, 606, Japan Abstract This paper attempts to systematize natural language analysis process by (I) use of a partitioned semantic network formalism as the meaning representation and (2) stepwise translation based on Montague Grammar. The meaning representation is obtained in two steps. The first step translates natural language into logical expression. The second step interprets logical expression to generate network structure. We have implemented set of programs which performs the stepwise translation. Experiments are in progress for machine translation and question answering. i. Introduction Conventional AI systems dealing with natural languages paid much efforts on the problem, how to translate natural language input into the internal knowledge structure such as micro PLANNER statements[14], semantic networks[6], frames[l], etc. Most of these systems directly translate input sentences into task oriented internal structure. The architecture of these systems will be much simplified if systematic meaning representation and analysis method based on a formal theory is incorpolated. This paper proposes a stepwise translation system based on Montague Grammar (MG for short)[3]. Partitioned semantic network[6] is employed as a meaning representatin. Input sentence is firstly translated into logical expression and then semantic network is generated by interpreting it. Semantic network is the output of the natural language analyzer. This will be further compiled into task oriented representations to be used by a task oriented problem solver. This paper concentrates on the natural language analyzer. The following is a summary of our approach: NATURAL LANGUAGE ANALYZER & LE .&. PSN ---. task oriented representations where NL: natural language, LE: logical expression, PSN: partitioned semantic network, T: translation mapping, I: interpretation mapping. We have developed natural language analyzers for English and Japanese respectively. This paper describes the one for English. The experiments are in progress with these systems. The applicability of the proposed approach is discussed briefly. 2. Overview of the approach This section gives the readers an overview of the system by illustrating an example. Before illustration we shall present the formalisms of LE and PSN. LE -- lo$ical expression The notion of LE is based on Cresswell's %-categorial language[2]. The following is the syntax of LE: - the set of syntactic categories (Syn): two basic categories are used, i.e., 0 of sentence and 1 of name. Given categories T,O~, ... ,O,, then <T,Ol, ... ,On> is the category of a mapping that makes an expression of category T out of expressions of category ol, ... ,~, respectively. - the set of symbols (F): F = #~y~FO, where Fo is a finite set of symbols, and if Ol#o 2 then FO,~F~=~. - the set of variables (X): X= ~5~Xo, where Xo is a set of variables such that if ~i#o~ then X~znXa~=~ , and that intersection of F and X is empty. - the set of expressions (E): E = ~ ~E~, where E~ fills the following properties: (i) Xo GEo, …

Content truncated to 500 words