<?xml version="1.0" standalone="yes"?>
<Paper uid="P84-1014">
  <Title>Interaction of Knowledge Sources in a Portable Natural Language Interface</Title>
  <Section position="3" start_page="0" end_page="0" type="metho">
    <SectionTitle>
2. Overview of DATALOG Architecture
</SectionTitle>
    <Paragraph position="0"> The architecture of DATALOG is based on Cascaded ATN grammar, a general approach to the design of language processors which is an extension of Augmented Transition Network grammar \[13\]. The Cascaded ATN approach to NL processing was first developed in the RUS parser \[2\] and was formally characterized by Woods \[14\]. Figure 1 shows the architecture of a Cascaded ATN for NL processing: the syntactic-and semantic components are implemented as separate processes which operate in parallel, communicating information back and forth. This communication (represented by the &amp;quot;interface&amp;quot; portions of the diagram) allows a linguistic ATN grammar to interact with a semantic processor, creating a conceptual representation of the input in a step-by-step manner and rejecting semantically incorrect analyses at an early stage.</Paragraph>
    <Paragraph position="1"> DATALOG extends the architecture shown in Figure 1 in the direction of increased portability, by dividing semantics into two parts (see Figure 2). A &amp;quot;general&amp;quot; semantic processor based on the relational model of data \[5\] interprets a wide variety of information requests applied to  abstract database objects. This level of knowledge is equivalent to what Hendrix has labelled &amp;quot;pragmatic grammar&amp;quot; \[9\]. Domain knowledge is represented in a semantic network, which encodes the conceptual structure of the user's database.</Paragraph>
    <Paragraph position="2"> These two levels of knowledge representation ar~ linked together, as described in Section 4 below.</Paragraph>
    <Paragraph position="3"> The output of the cascaded ATN grammar is a combined linguistic and conceptual representation of the query (see Figure 3), which includes a &amp;quot;SEMANTICS&amp;quot; component along with the usual linguistic constituents in the interpretation of each phrase.</Paragraph>
    <Paragraph position="4"> 3. Interaction of Syntax and Semantics The DATALOG interface between syntax and semantics is a simplification of the RUS approach, which has been described in detail elsewhere \[ii\].</Paragraph>
    <Paragraph position="5"> The linguistic portion of the interface is imple- null mented by adding a new arc action called &amp;quot;ASSIGN&amp;quot; to the ATN model of grammar. ASSIGN communicates partial linguistic analyses to a semantic interpreter, which incrementally creates a conceptual representation of the input. If an assignment is nonsensical or incompatible with previous assignments, the semantic interpreter can reject the assignment, causing the parser to back up and try another path through the grammar.</Paragraph>
    <Paragraph position="6"> In DATALOG, ASSIGN is a function of three arguments: the BEAD of the current clause or phrase, the CONSTITUENT which is being added to the interpretation of the phrase, and the SYNTACTIC SLOT which the constituent occupies. As a simplified example, an ATN gram, mr might process noun phrases by &amp;quot;collecting&amp;quot; determiners, numbers, superlatives and other pre-modifiers in registers until the head noun is found. Then the head is assigned to the NPHEAD slot; the pre-modifiers are assigned (in reverse order) to the NPPREMOD slot; superlatives are assigned to the SUPER slot; and numbers are assigned to the NUMBER slot. Finally, the determiners are assigned to the DETERMINER slot.</Paragraph>
    <Paragraph position="7"> If all of these assignments are acceptable to the s~m~ntic interpreter, an interpretation is constructed for the &amp;quot;base noun phrase&amp;quot;, and the parser can then begin to process the noun phrase post-modifiers. Figure 3 illustrates the interpretation of &amp;quot;the tallest female employee&amp;quot;, according to this scheme. A more detailed description of how DATALOG constructs interpretations is contained in another report \[8\]. During parsing, semantic information is collected in &amp;quot;semantic&amp;quot; registers, which are inaccessible (by convention) to the grammar. This convention ensures the generality of the grammar; although the linguistic component (through the assignment mechanism) controls the information that is passed to the semantic interpreter, the only information that flows back to the grazm~ar is</Paragraph>
  </Section>
  <Section position="4" start_page="0" end_page="58" type="metho">
    <SectionTitle>
CONSTITUENT SYNTACTIC SLOT
</SectionTitle>
    <Paragraph position="0"> the acceptance or rejection of each assignment.</Paragraph>
    <Paragraph position="1"> When the grammar builds a constituent structure for a phrase or clause, it includes an extra constituent called &amp;quot;SEMANTICS&amp;quot;, which it takes from a semantic register. However, generality of the grammar is maintained by forbidding the gra~mmar to examine the contents of the SEMANTICS constituent.</Paragraph>
  </Section>
  <Section position="5" start_page="58" end_page="58" type="metho">
    <SectionTitle>
4. Interaction of General and Application
</SectionTitle>
    <Paragraph position="0"/>
    <Section position="1" start_page="58" end_page="58" type="sub_section">
      <SectionTitle>
Semantics
</SectionTitle>
      <Paragraph position="0"> The semantic interpreter is divided into two levels: a &amp;quot;lower-level&amp;quot; semantic network representing the objects and relationships in the application domain; and a &amp;quot;higher-level&amp;quot; network representing general knowledge about database structures, data analysis, and information requests. Each node of the domain network, in addition to its links with other domain concepts, has a &amp;quot;hook&amp;quot; attaching it to the higher-level concept of which it is an instance. Semantic procedures are also attached to the higher-level concepts; in this way, domain concepts are indirectly linked to the semantic procedores that are used to interpret them.</Paragraph>
      <Paragraph position="1"> Figure C/ illustrates the relationship between the general concepts of DATALOG and the domain semantic network of a personnel application.</Paragraph>
      <Paragraph position="2"> Domain concepts such as &amp;quot;female&amp;quot; and &amp;quot;dollar&amp;quot; are attached to general concepts such as /SUBCLASS/ and /UNIT/. (The higher-level concepts are delimited by slash &amp;quot;/&amp;quot; characters.) When a phrase such as &amp;quot;40000 dollars&amp;quot; is analyzed, the semantic procedures for the general concept ,'b~::T/ are invoked to interpret it.</Paragraph>
      <Paragraph position="3"> The general concepts also organized ~nto a network, which supports inheritance of s~msntic procedures. For example, two of the general concepts in DATALOG are /ATTR/, which can represent any attribute in the database, and /NUMATTR/, which represents numeric attributes such as &amp;quot;salary&amp;quot; and &amp;quot;age&amp;quot;. Since /ATTR/ is the parent of /NUMATTR/ in the general concept network, its semantic procedures are automatically invoked when required during interpretation of a phrase whose head is a numeric attribute. This occurs whenever no /NUMATTR/ procedure exists for a given syntactic slot; thus, sub-concepts can be defined by specifying only those cases where their interpretations differ from the parent.</Paragraph>
      <Paragraph position="4"> Figure 5 shows the same diagram as Figure 4, with concepts from the computer hardware database substituted for personnel concepts. This illustrates how the semantic procedures that interpreted personnel queries can be easily transported to a different domain.</Paragraph>
    </Section>
  </Section>
class="xml-element"></Paper>