A visualized framework for representing uncertain and incomplete temporal knowledge

This thesis presents a visualized framework, called Visual Time, for representing and reasoning about incomplete and uncertain temporal information. It is both expressive and versatile, allowing logical conjunctions and disjunctions of both absolute and relative temporal relations, such as “Before”, “Meets”, “Overlaps”, “Starts”, “During”, and “Finishes”, etc. In terms of a visualized framework, Visual Time provides a user-friendly environment for describing scenarios with rich temporal structure in natural language, which can be formatted as structured temporal phrases and modelled in terms of Time Relationship Diagrams (TRD). A TRD can be automatically and visually transformed into a corresponding Time Graph, supported by automatic consistency checker that derives a verdict to confirm if a given scenario is temporally consistent or inconsistent. The thesis provides the following contributions: 1. Extended graphical representation for uncertain and incomplete temporal knowledge: An extended graphical representation for uncertain and incomplete temporal knowledge based on [KM1992] is proposed, supporting both logical connectives ‘∧’ and ‘∨’. In Chapter 3, it is shown all the other logical connectives can be derively defined. 2. Time relation diagram (TRD): A time relation diagram (TRD) is designed for representing temporal relations between time elements which could be both point and interval. Each time element is denoted as a box consisting of three components: Name, Duration and Property. Temporal relations are denoted in terms of directed arcs. TRD allows expressions of both absolute and relative temporal relations, supporting both logical conjunctions and disjunctions. 3. A semi-automatic temporal information extractor: SUTime is a very useful tool for extracting verbs and temporal information [CM2012]. However, the extracted verbs and temporal information may play different roles when modelled by TRD. For example, in "He starts to start the car", "start" is an event while "starts" means the action "start" happens. An improved algorithm called Temporal Extractor algorithm (TE) is introduced in Section 4.2. Based on Stanford SUTime, TE can semi-automatically extract time elements and temporal relations from any arbitrary text to create a TRD. 4. Four algorithms: The first algorithm, Temporal Relation Algorithm (TRM), is designed to extract temporal relations from TRD. The second algorithm, Meets Table Algorithm (MTM) is introduced to convert all the extracted temporal relations into a Meets table. The third algorithm, Time Graph Algorithm (TGM) is described to draw the corresponding time graph of a given TRD. The fourth algorithm, Consistency Checking Algorithm (CCM), is designed to check the consistency of TRD. If the TRD is inconsistent, an audio verdict will alert and the corresponding time element(s) and natural texts will be marked in red colour.

Name dc:type.qualificationname

phd

Level dc:type.qualificationlevel

doctoral

Grantor dc:publisher.institution

University of Greenwich

Year dc:date.issued

2014

Author dc:creator

• Wang, Yue

Advisors dc:contributor.advisor

• Ma, Jixin
• Soper, Alan

• QA Mathematics

Language dc:language

en