File Information
File: 05-lr/acl_arc_1_sum/cleansed_text/xml_by_section/concl/92/p92-1003_concl.xml
Size: 2,303 bytes
Last Modified: 2025-10-06 13:56:57
<?xml version="1.0" standalone="yes"?> <Paper uid="P92-1003"> <Title>A SIMPLE BUT USEFUL APPROACH TO CONJUNCT IDENTIFICATION 1</Title> <Section position="8" start_page="19" end_page="20" type="concl"> <SectionTitle> FUTURE RESEARCH </SectionTitle> <Paragraph position="0"> In addition to the above, the most important step to be taken at this point is to build the comma specialist and clause recognition specialist. Another problem that needs to be addressed involves deciding priorities when one or more prepositional phrases are attached to oneof the conjuncts of a coordinate conjunction. For example, we need to decide between the structures \[\[A and B\] in dogs\] and \[A and \[B in dogs\]\], where A and B are typically large structures themselves, A and B should be conjoined, and 'in dogs' may appropriately be attached to B. It is not clear whether the production of the appropriate structure in such cases rightfully belongs to the knowledge analysis portion of our system, or whether most such questions can be answered by the NLP portion of our system with the means at its disposal. Further, the basic organization of the NLP component, with the tagger and the semi-parser generating the flat structure and then the various specialist programs working on the sentence structure to improve it, looks a lot like a blackboard system architecture.</Paragraph> <Paragraph position="1"> Therefore, one of the future ventures could be to try to look into some blackboard architecture and assess its applicability in this system.</Paragraph> <Paragraph position="2"> Finally, there are ambiguities inherently associated with coordinate conjunctions, including the problem of differentiating between &quot;segregatory&quot; and &quot;combinatory&quot; use of conjunctions \[Quirk et al., 1982\] (e.g. &quot;fly and mosquito repellants&quot; could refer to 'fly' and 'mosquito repellants' or to 'fly repellants' and 'mosquito repcllants'), and the determination of whether the 'or' in a sentence is really used as an 'and' (e.g. &quot;dogs with glaucoma or keratoconjunctivitis will recover&quot; implies that dogs with glaucoma and dogs with keratoconjunctivitis will recover). The current algorithm does not address these issues.</Paragraph> </Section> class="xml-element"></Paper>