Sinica BOW (Bilingual Ontological Wordnet): Integration of Bilingual WordNet and SUMO

The Academia Sinica Bilingual Ontological Wordnet (Sinica BOW) integrates three resources: WordNet, English-Chinese Translation Equivalents Database (ECTED), and SUMO (Suggested Upper Merged Ontology). The three resources were originally linked in two pairs: WordNet 1.6 was manually mapped to SUMO (Niles & Pease 2003) and also to ECTED (the English lemmas in WordNet were mapped to their Chinese lexical equivalents). ECTED encodes both equivalent pairs and their semantic relations (Huang et al. 2003). With the integration of these three key resources, Sinica BOW functions both as an English-Chinese bilingual wordnet and a bilingual lexical access to SUMO. Sinica BOW allows versatile access and facilitates a combination of lexical, semantic, and ontological information. Versatility is built in with its bilinguality, and the lemma-based merging of multiple resources. First, either English or Chinese can be used for the query, as well as for presenting the content of the resources. Second, the user can easily access the logical structure of both the WordNet and SUMO ontology using either words or conceptual nodes. Third, multiple linguistic indexing is built in to allow additional versatility. Fourth, domain information allows another dimension of knowledge manipulation. 1 Background and Motivation Conceptual structure and lexical access are two essential elements of human knowledge. Bilingual representation of both conceptual structure and lexical information will enable language independent knowledge processing. In this paper, we introduce a new type of integrated language resources: Bilingual Ontological Wordnet. The Academia Sinica Bilingual Ontological Wordnet (Sinica BOW) was constructed in 2003. We argue that such combination of ontology and wordnet will 1) give each linguistic form a rigorous conceptual location, 2) clarify the relation between the conceptual classification and its linguistic instantiation, and 3) facilitate genuine cross-lingual access of knowledge. 2 Resources and Structure The Academia Sinica Bilingual Ontological Wordnet (Sinica BOW) integrates three resources: WordNet, English-Chinese Translation Equivalents Database (ECTED), and SUMO (Suggested Upper Merged Ontology). WordNet is a lexical knowledgebase for English language that was created at Cognitive Science Laboratory of Princeton University in 1990 (Fellbaum 1998). Its content is divided into four categories based on psycholinguistic principles: nouns, verbs, adjectives and adverbs. WordNet organizes the lexical information according to word meaning and each synset groups together a set of lemmas sharing the same sense. In addition, WordNet is a semantic network linking synsets withvlexical semantic relations. WordNet is widely used in Natural Language Processing applications and linguistic research. The most updated version of WordNet is WordNet 2.0. We adopted WordNet 1.6., the version which is used by most applications so far. ECTED was constructed at Academia Sinica as a crucial step towards bootstrapping a Chinese wordnet with English WordNet (Huang et al. 2002, Huang et al. 2003). The translation equivalence database was hand-crafted by the WordNet team at CKIP, Academia Sinica. First, all possible Chinese translations of an English synset word (from WN 1.6.) are extracted from several available online bilingual (EC or CE) resources. These translation candidates were then checked by a team of translators with near-native bilingual ability. For each of the 99,642 English synsets, the translator selected the three most appropriate translation equivalents whenever possible. The translation equivalences were defaulted to lexicalized words, rather than descriptive phrases, whenever possible. The translation equivalences were then manually verified. Note that after the first round of translation, there were about 5% of the lemmas whose Chinese translation can neither be found in our bilingual resources nor be filled by the translators. We spent another 2 person-year consulting various special dictionaries to fill in the gaps. SUMO is a upper ontology constructed by the IEEE Standard Upper Ontology Working Group and maintained at Teknowledge Corporation. SUMO contains roughly 1,000 conceptual nodes for knowledge representation. It can be applied to automated reasoning, information retrieval and inter-operability in E-commerce, education and NLP tasks. Niles & Pease (2003) mapped synsets of WordNet and concept of SUMO in three relations: synonymy, hypernymy and instantiation. For instance, the synset "animal" (a living organism characterized by voluntary movement) in WordNet is synonymous with the SUMO concept of "Animal". In "bank" (a financial institution that accepts deposits and channels the money into lending activities) this case, bank is a corporation that is a hypernym of the associated synset. President of the United States (the office of the US head of state) is an instantiation of "position" concept. Through the