MEANING, WHAT IS IT?

Meanings, in contrast to texts, cannot be observed directly. As we mentioned above, we consider the Meaning to be the structures in the human brain which people experience as ideas and thoughts. Since we do not know and cannot precisely represent those brain processes, for practical purposes we must use a representation of Meaning, which is more suitable for manipulation in a computer. Thus, for our purposes, Meaning will be identified with that representation.

In the future, neurophysiological research will eventually discover what signals really correspond to meanings in our brain, and what structure these signals have, but for now those signals remain only vaguely understood. Hence we take the pragmatic viewpoint that, if a representation we use allows a computer to manipulate and respond to texts with an ability close to that of a human, then this representation is rather good for the real Meaning and fits our purposes.

As it was described earlier, the task of language is to transform information from one representation, the Text, into another, the Meaning, and vice versa. A computer program that models the function of language must perform the same task. In any application, there should be some other system or device that consumes the results of the transformation of texts, and produces the information that is to be transformed into a text. The operations of such a device or a system is beyond the scope of computational linguistics itself. Rather, such a system uses the linguistic module as its interface with the outer world (see Figure IV.7).

For a linguistic module of a system, the Meaning is a formal language or a format of information representation immediately understandable for, or executable on, the consumer of the information: the underlying expert or reasoning system, database, robot control system, etc. It is supposed that this underlying system produces its responses just in the same format. Thus, in practice, the format of Meaning is already given to the developers of the linguistic module for any specific application system.

Usually, such systems are aware on the entities mentioned in the text, their states, properties, processes, actions, and relationships.

Besides, there exists other type of information in a text, such as beliefs, estimations, and intentions of its author. For example, in the Spanish sentence Creo que su esposa está aquí, the part reflecting the author’s belief is given in bold face. The author usually flavors any text with these elements. Words reflecting basic information, through some stylistic coloring, can additionally express author’s attitude. For example, the Spanish sentence Este hombrón no está trabajando ahora has the meaning ‘this man is not working now and I consider him big and coarse’.

FIGURE IV.7. Structure of an application system with a natural language interface.

Perhaps only very formal texts like legal documents do not contain subjective attitude of the author(s) to the reflected issues. The advanced application system should distinguish the basic information delivered in texts from author’s beliefs, estimations, and intentions.

Additionally, even a very formal text contains many explicit references and explanations of links between parts of a text, and these elements serve as a content table or a protocol of the author’s information about text structuring. This is not the information about the relevant matters as such. Instead, this is some kind of meta-information about how the parts of the text are combined together, i.e., a “text about text.” We will not discuss such insertions in this book. Since properties, processes, and actions can be represented as relationships between entities touched upon in Text, just these features are used to represent Meaning.

The entities, states, properties, processes, and actions are usually denoted by some names in semantic representation. These names can be compared with the names of variables and functions in a computer program. Such names have no parts of speech, so that a process or an action of, say, ‘development’ can be equally called in Spanish desarrollar or desarrollo, while the property of ‘small’ can be equally called pequeño or ser pequeño. Usually only one wordform is used for the name, so that the property itself is called, for instance, pequeño (neither pequeña nor pequeñas). Concerning the value plural of grammatical category of number, on the semantic level it is transformed to the notion of multiplicity and is represented by a separate element.