Abstract This article discusses the parsing of multi-verb constructions in the Indic language Odia, in a Deterministic Finite State Automaton (DFA). It deals with two verb (V-v) sequences, and in the case of passivization, 3 verb sequences. These V-v sequences are formed by combining a main verb and a fully or partially bleached ‘light’ verb; where the main verb carries the lexical semantic information;... This article discusses the parsing of multi-verb constructions in the Indic language Odia, in a Deterministic Finite State Automaton (DFA). It deals with two verb (V-v) sequences, and in the case of passivization, 3 verb sequences. These V-v sequences are formed by combining a main verb and a fully or partially bleached ‘light’ verb; where the main verb carries the lexical semantic information; the Tense, Aspect, Mode marking occurs on the light verb; and the sequence as a whole determines the argument structure. Although both the verbs in the V-v sequences are form-identical with a main verb in the language, the second one is a light verb because of its semantic and grammatical bleaching. The verbs cannot change their position, and are always spelled as a single word. Such double positional slots for the verbs pose constraint for the processing of the string by DFA, as it would be different from a standardly accepted single verb string. Moreover, both the verbs vary in their grammatical and semantic functions, and hence, choose a particular type of verb to co-occur with. This paper proposes the parsing of such multi-verb constructions in a DFA. Such morphological parsing is new in Odia and can be used in various applications like morphological analyzer, spell-checker, machine translation, information retrieval, etc. Read More Read Less

Abstract Sentiment classification of posts on Weibo is a key to analyse people’s opinions and attitudes toward products, services, and social events. It is also at the core of many other natural language processing tasks. In this paper, we apply a new feature representation technique to building a sentiment classifier and classifying the posts. Firstly, we crawled a set of posts from Weibo and... Sentiment classification of posts on Weibo is a key to analyse people’s opinions and attitudes toward products, services, and social events. It is also at the core of many other natural language processing tasks. In this paper, we apply a new feature representation technique to building a sentiment classifier and classifying the posts. Firstly, we crawled a set of posts from Weibo and labelled them. Then, we cleaned the data to remove the noisy information and transformed the varying-length posts into fixed-sized input vectors based on five different feature engineering techniques. Finally, we evaluated the performances of the five different feature representation techniques on a same data set with the use of support vector machines, naive Bayes and classification and regression tree. Experimental results demonstrate that our new method is efficient and outperforms the other ones. Read More Read Less

Abstract In this paper, we proposed new model for digital library which is an extension for the 5S Model to include the semantic web layer in the structure of digital library to be 6S model for digital library. We also discuss the important role of semantic web in digital library and how are semantic web technologies affect the information retrieval accuracy. We represent the... In this paper, we proposed new model for digital library which is an extension for the 5S Model to include the semantic web layer in the structure of digital library to be 6S model for digital library. We also discuss the important role of semantic web in digital library and how are semantic web technologies affect the information retrieval accuracy. We represent the semantic layer in 6S Model by adding ontology to a digital library that satisfies the 5S model and enhance ontology by updating it automatically. This ontology is used in books classification and retrieval according to concepts in ontology. Read More Read Less