Extracting Semantic Frames from Thai Medical-Symptom Unstructured Text with Unknown Target-Phrase Boundaries

Abstract

Peerasak INTARAPAIBOON†a) , Student Member, Ekawit NANTAJEEWARAWAT †b) , and Thanaruk THEERAMUNKONG †c) , Members SUMMARY Due to the limitations of language-processing tools for the Thai language, pattern-based information extraction from Thai documents requires supplementary techniques. Based on sliding-window rule application and extraction filtering, we present a framework for extracting semantic information from medical-symptom phrases with unknown boundaries in Thai unstructured-text information entries. A supervised rule learning algorithm is employed for automatic construction of information extraction rules from hand-tagged training symptom phrases. Two filtering components are introduced: one uses a classification model to predict rule application across a symptom-phrase boundary based on instantiation features of rule internal wildcards, the other uses weighted classification confidence to resolve conflicts arising from overlapping extractions. In our experimental study, we focus our attention on two basic types of symptom phrasal descriptions: one is concerned with abnormal characteristics of some observable entities and the other with human-body locations at which primitive symptoms appear. The experimental results show that the filtering components improve precision while preserving recall satisfactorily.

A Survey of Thai Knowledge Extraction for the Semantic Web Research and Tools

“…As the accuracy for word segmentation is only around 90%, some work on high-level tasks (such as relation extraction) omits it altogether. Other basic NLP tools such as chunkers (shallow parsing) or full parsers are currently not available for Thai [8].…”

“…In the area of pattern-based systems, Imsombut and Sirikayon [28] extract NEs which are instances of tourism attractions or tourism activities with Conditional Random Fields (CRF) and a hand-crafted tourism ontology for categorization -so the approach is rather a method for domainspecific ontology population. As a preprocessing step before doing rule extraction, Intarapaiboon et al [34] extract and annotate specific entities, namely chemical reaction names and chemical substances in domain text, or medical entities, respectively [8]. Similarly, Sitthisarn and Bahoh [32] apply NER during preprocessing for their ontology population and relation extraction task.…”

“…Finally to provide background knowledge, the concept entities are linked to existing ontologies about chemical reactions. In a similar approach, Intarapaiboon et al [8] extract semantic frames using rules learned with WHISK from hand-tagged training data. Again, the rules are applied using a sliding window on unsegmented text, and the method is evaluated in multiple domains.…”

“…Event extraction is similar to (binary) relation extraction, but focusing on specific event types and allowing n-ary relations. Due to the situation of missing chunk parsers for Thai language, Intarapaiboon et al [8], [34], [43] modify the WHISK information extraction technique using sliding windows to operate on Thai text. Those supervised approaches for pattern learning in multiple domains are related to event or frame extraction, they aim at detecting predefined semantic frames from text.…”

An Application of Intuitionistic Fuzzy Sets to Improve Information Extraction from Thai Unstructured Text

“…When the slot type of a slot filler assigned by the classifier is inconsistent with that by the IE pattern the extracted event is discarded. Similarly, in [5], an pattern-based IE framework to extract multi-slot frames was proposed. To improve precision by removing false extraction, two extraction filtering modules were proposed.…”