Books
in black and white
Main menu
Home About us Share a book
Books
Biology Business Chemistry Computers Culture Economics Fiction Games Guide History Management Mathematical Medicine Mental Fitnes Physics Psychology Scince Sport Technics
Ads

Communicating with Databases in Natural Language - Wallace M.

Wallace M. Communicating with Databases in Natural Language - Ellis Horwood Limited, 1985. - 170 p.
Download (direct link): comumunicatingwthisdatabase1985.djvu
Previous << 1 .. 13 14 15 16 17 18 < 19 > 20 21 22 23 24 25 .. 59 >> Next

In Chaptei 5 the parser is discussed with close attention to the applicability restrictions imposed by the D&Qs and the application to which the system is attached. Examples are taken from an application on elections and an application called ‘COPSE’ about customers, orders, products and stocks (see Appendix 4) Finally the QPROC implementation in PROLOG is examined with some discussion of the limitations of QPROC and further illustrations of PROLOG as a language for writing parsers.
3
Formalising natural language
1 THE DIFFERENCE BETWEEN NATURAL LANGUAGE AND FORMAL LANGUAGES
Natural language is very diffeient from any formal language. In the first case the meaning of natural language words cannot be formally defined, secondly natural language is ambiguous, and thirdly the interpretation of natural language is pragmatic. We shall review these three differences individually,
1 1 Word meanings
‘CUSTOMER’ records in a database refer specifically to the customers for a certain business, and a fixed set of relevant facts about them. In natural language, however, “customer” might be used in a variety of senses (eg referring to a person “He’s a nasty customer”). Natural language includes many vague words, such as “many” in this very sentence (“Many” contrasts with “few”, see Chapter
2, section 4.2 Incidentally the natural language facility to self-refer as exampled above is also impossible to formalise since it results in sentences which cannot be interpreted such as “This sentence is false” )
The first step in formalising natural language is to select (one or more) specific interpretations for each word For a given application only certain words can be formalised; the others must therefore be excluded from the natural language subset understood by the system,
1.2 Ambiguity
Even when the words have been formalised they may still have more than one meaning. Moreover natural language syntax is also ambiguous. This classic
[Ch, 3]
FORMALISING NATURAL LANGUAGE 51
example is a sentence with no ambiguous words, but at least three possible interpretations:
“I saw a man on a hill with a telescope ”
(I, the man or the hill could have the telescope.)
Another type of ambiguity arises in connection with ‘quantifier scope’ “Who placed orders in June and July?”, for example, has two possible interpretations:
— “Who placed orders in both June and July?” (where the ‘scope’ of “Who” is ‘outside’ the scope of “June and July”),
— “Who placed orders in June or July or both?” (which could be expressed as “In June and July, who placed orders?”, with the ‘scope’ of “June and July” ‘outside’ the scope of “Who”),
13 Pragmatics
The third difference between formal and natural language is the pragmatic aspect of understanding natural language which was discussed in Chapter 2, section 11, To understand and answer a natural language question we often need to use common sense. For these occasions there are no rules — that is why it is called common sense. It is very hard to write a natural language system so complex that it does not appear to be following any fixed rules! Thus it is hard to build common sense into such a system,
2. QPROC OBJECTIVES
The job of natural language understanding can be broken into two parts,
— Formalising natural language,
— Executing formal language statements
Many recognised techniques have been developed for parsing and executing formal language statements and questions: thus the execution part is better understood. The formalisation part should therefore be kept as small as possible by developing a formal language as close as possible to the structure of natural language,
2.1 Deep structure
The formal language must have an unambiguous formal vocabulary and a unique parsing for each statement. One possible candidate for such a language is a parse tree of the original sentence with the formal interpretation of the natural language words as its ‘leaves’,
A refined version of this idea — the “deep structure” of a sentence — was introduced by Chomsky [4], However, it is now argued by many linguists that Chomsky’s attempt to separate the syntax of natural language from its semantics has been a failure (King in [55]),
52 FORMALISING NATURAL LANGUAGE
[Ch
In the context of natural language front ends there are several arguments why parse trees are not a very satisfactory formalisation. Firstly, of course, a sentence may have many parse trees which have nothing to do with its meaning. For example the sentence,
“I saw a man on a hill with a dark brown suit”
must have three parse trees corresponding to the three interpretations of “I saw a man on a hill with a telescope” However, two of these parse trees could not represent possible formalisations of the new sentence, since this only has one meaning.
Secondly it is necessary to map sentences such as “Pencils cost ?5” and “It costs ?5 for pencils”
onto similar formal structures. Such a requirement has driven linguists to propose various ‘case’ grammars which parse a sentence, not into subject, verb and object, but into a verb and its ‘cases’.
Previous << 1 .. 13 14 15 16 17 18 < 19 > 20 21 22 23 24 25 .. 59 >> Next