Shun Ishizaki

Disambiguating Japanese compound verbs

The purpose of this study is to disambiguate Japanese compound verbs (JCVs) using two methods: (1) a statistical sense discrimination method based on verb-combinatoric information, which feeds into a first-sense statistical sense disambiguation method and (2) a manual rule-based sense disambiguation method which draws on argument structure and verb semantics. In evaluation, we found that the rule-based method outperformed the statistical method at 94.6% token-level accuracy, suggesting that fine-grained semantic analysis is an important component of JCV disambiguation. At the same time, the performance of the fully automated statistical method was found to be surprisingly good at 82.6%, without making use of syntactic or lexical semantic knowledge. e.

Sequence disambiguation and pattern completion by cooperation between autoassociative and heteroassociative memories of functionally divided hippocampal CA3

The hippocampus memorizes event sequences and some events are shared by a few sequences. When a sequence is retrieved from the linked sequences, ambiguity of sequences becomes a serious problem because the shared events have some possible ways to retrieve other events. We have focused on location dependency elucidated from the hippocampus and suggested that CA3 is functionally divided into autoassociative and heteroassociative memories. Computer simulation results show that the functionally divided CA3 concurrently enables both sequence disambiguation and pattern completion. Consequently, it demonstrates that cooperation between both memories of CA3 brings out the abilities of sequence disambiguation and pattern completion in the hippocampus.

A Connectionist Approach to the Generation of Abstracts

This chapter discusses a method for extracting significant portions out of what we call contextual representation structure (CRS). The method is based on a connectionist paradigm in which information processing in the human brain is accounted for in terms of signal propagation in a network which reflects the topology of neural connections. The key idea is to regard the degree of importance of each node in a CRS network as the activation intensity of that node after a saturation of signal propagation across the relevant part of the network. Both top-down and bottom-up information are naturally accommodated in the abstract by this single principle.

A Disambiguation Method for Japanese Compound Verbs

The purpose of this study is to construct a semantic analysis method for disambiguating Japanese compound verbs. Japanese speakers produce a rich variety of compound verbs, making it difficult to process them by computer. We construct a method employing 110 disambiguation rules based on the semantic features of the first verb of a compound and syntactic patterns consisting of co-occurrence between verbs and nouns. The disambiguation rules are evaluated by applying them to compound verbs in the dictionary. The obtained accuracy is 87.19% for our rules. This result shows the advantage of our method.

Network features and pathway analyses of a signal transduction cascade

The scale-free and small-world network models reflect the functional units of networks. However, when we investigated the network properties of a signaling pathway using these models, no significant differences were found between the original undirected graphs and the graphs in which inactive proteins were eliminated from the gene expression data. We analyzed signaling networks by focusing on those pathways that best reflected cellular function. Therefore, our analysis of pathways started from the ligands and progressed to transcription factors and cytoskeletal proteins. We employed the Python module to assess the target network. This involved comparing the original and restricted signaling cascades as a directed graph using microarray gene expression profiles of late onset Alzheimers disease. The most commonly used method of shortest-path analysis neglects to consider the influences of alternative pathways that can affect the activation of transcription factors or cytoskeletal proteins. We therefore introduced included k-shortest paths and k-cycles in our network analysis using the Python modules, which allowed us to attain a reasonable computational time and identify k-shortest paths. This technique reflected results found in vivo and identified pathways not found when shortest path or degree analysis was applied. Our module enabled us to comprehensively analyse the characteristics of biomolecular networks and also enabled analysis of the effects of diseases considering the feedback loop and feedforward loop control structures as an alternative path.

Sinhalese morphological analysis: a step towards machine processing of Sinhalese

Initial steps toward the machine processing of Sinhalese, including automatic translation, are discussed. Script input representation methods, morphological analysis, and some characteristics of Sinhalese are considered.<<ETX>>

Machine processing of a natural language with interchangeable phrases

Abstract Machine processing of natural languages has been a difficult problem in the intelligent systems research area for the last four decades. Early approaches to machine processing of natural languages failed, partly because of interaction effects of complex phenomena that made the machine process appear to be unmanageable. Recent approaches to the problem have been more successful but are limited on many language-specific rules related to rigid word order of the languages. Many machine processing systems depend heavily on the language-specific rigid word order. The problem addressed in this paper is the construction of a machine processing system relying on language-specific imprecise computations. This paper presents an alternative approach to machine processing of natural languages. The system relies on interchangeable phrases and flexible word order. The application of imprecise computation for machine translation problems is approached by examining Sinhalese. The design and implementation of the system are described and illustrated with detailed examples of overall system operation.

Context analysis system for Japanese text

A natural language understanding system is described which extracts contextual information from Japanese texts. It integrates syntactic, semantic and contextual processing serially. The syntactic analyzer obtains rough syntactic structures from the text. The semantic analyzer treats modifying relations inside noun phrases and case relations among verbs and noun phrases. Then, the contextual analyzer obtains contextual information from the semantic structure extracted by the semantic analyzer. Our system understands the context using precoded contextual knowledge on terrorism and plugs the event information in input sentences into the contextual structure.

Analysis system for Sinhalese unit structure

Abstract Sinhalese is the major language in Sri Lanka, spoken by 15 million people. It has never been analysed by a computer. The paper describes an original computational linguistic analysis for Sinhalese. A machine representation of Sinhalese script is developed. It provides an easy means of input and facilitates the formalization of the linking phenomena. An analysis sytem for Sinhalese morphology is developed. A Sinhalese sentence consists of a few units separated by spaces. The unit structure is formalized as a root and suffixes. Connection rules and linking rules are developed. The grammatical features of a unit are characterized by a set of attributes, and a mechanism to compute these attributes is developed from the features of the root and suffixes. The unit structure involves much important grammatical information such as case and attributes. The analyser can handle any kind of Sinhalese unit efficiently. The system will be used as the base for machine processing of Sinhalese, its syntactic and ...

Generating Japanese text from conceptual representation

Most methods for machine generation of Japanese sentences work from intermediary representations in syntactic structures, or tree structures, especially in machine translation systems. As the first Japanese text generation program to work directly from conceptual representations, our system represents a major departure from the approach. These conceptual representations include no explicit syntactic information. Rather, the representation contains scenes that correspond to events, including input on actors, objects, time, location and relations to other events. The generator builds complete Japanese sentence structure from the representation, and then refines it.