• Digital Humanities and Cultural Heritage
• Publications close
• Cluster Computing close

In this paper, we propose to adapt the F-measure to evaluate an automatic summaries of texts; we the main key to our proposal is to prove that the automatic summary task can be modeled in supervised classification. First, we will start the research by to make a comparison between the automatic summary task and the supervised classification. After that, we are going to define how to draw a confusion matrix which classification evaluation base, and from it we calculate the F-Measure. In this vein, we must prove that the new measure is valid and trustworthy, And for this we will calculates the correlation with ROUGE Evaluation. Before ending we analyze and interpret the main results in order to answer the questions put forward in this research. At the end, we are going to conclude our study with a set of facts based on the collected data.

English grammar teaching is an indispensable part of language teaching, and learners’ mastery and use of grammatical knowledge have a positive and important role in obtaining speech ability. Aiming at the problem of characteristic processing in business English parallel corpora in business contracts, this paper proposes a similar-Apriori mining semantic processing algorithm under two-stage high expectation weight business English parallel corpus item set. First of all, aiming at the problem that in the traditional hierarchical mining algorithm, only business English parallel corpus item set upper bound is used, which has not ideal effect for the high expectation weight problem, the downward closure characteristics of hierarchy mining algorithm is designed, and the proof procedure is given, this characteristic can effectively reduce the processing quantity of the candidate business English parallel corpus while ensuring the accuracy. Secondly, the two-stage data mining process is constructed, in the first stage, a high expectation weight business English parallel corpus item set is obtained for a set of candidate business English parallel corpus item set based on a hierarchical search method, in the second stage, the database is scanned again to obtain high expectation weight of business English parallel corpus item set, so as to complete the data mining process. Finally, the effectiveness of the algorithm is verified through simulation experiment.

At present, the intelligent translation of computer is developing rapidly. In this paper, the overall design of interactive intelligent auxiliary translation platform based on multi-strategy is studied. In this paper, the algorithm of similarity of sentences is deeply studied. According to the needs of English–Chinese translation memory system, the similarity between sentences is calculated from the sentence syntax and semantic based on syntactic and semantic English sentence similarity calculation method. Finally, the author uses the English–Chinese translation of Navigator 7.0 manual to examine the effect of the algorithm platform developed in this paper. The experimental results show that the proposed algorithm improves the speed of translation by about 35%.

In this paper, the cross-language retrieval model based on statistical language model, cross-lingual text categorization method and cross-lingual text clustering method are studied systematically and deeply. Without any help of cross-lingual resources such as machine translation and bilingual dictionaries, this paper can solve the many-to-many problem of word translation in CLIR and solve the problem of unregistered words partially. Under a unified framework, a series of topics are extracted from bilingual parallel corpora to form the thematic space for each language. Thematic space of each language exists independently, and the bilingual subject space is established through the bilingual semantic correspondence. The bilingual subject space reflects the semantic correspondence between documents and documents, words and words. It reveals the inherent structure and internal relations among languages and languages.

Even though various advances have been made in recent years, the recognition of handwritten characters is still an open challenge in the Pattern Recognition field. Different approaches are invented for the recognition of printed characters of Indian languages. However, few attempts are done for the recognition of handwritten characters. A high degree of recognition accuracy for the handwritten characters is yet to be achieved. In this paper, a new approach based on deep belief network with the distributed average of gradients feature is presented for the recognition of isolated handwritten characters of Kannada, which is the official language of Karnataka state in India. In the proposed methods, a better accuracy is achieved.

Student answer script contains different type of answers to be evaluated. The answers may be of objective type answers, subjective type answers, mathematical answers, diagrammatic answers, classification type answers, each of which may require unique approaches for automated evaluation. Classification type answer makes a student to map an answer/object to a particular class/type, like types of grammars, types of normal forms and types of functions. This paper proposes an approach to automate the classification type answers using discrete model which classifies the types of relations in discrete mathematics domain. The proposed approach achieves 100% classification accuracy when compared with other types of classification since the pellet reasoner is used as a classifier which uses predefined classification rules to classify the types of relations.

Writing essays and technical documents can be a challenging task for many people, especially for non-native speakers. Good content and ideas are both important in writing, but clear and effective expressions that can accurately convey the meaning of these ideas to the readers are essential for good writing. Many writers often face difficulty in selecting the proper words that would fit into their sentences. Proper words may be widely used words that appear in similar contexts. These can be identified by a statistical analysis of a corpus, which is a collection of a large number of sentences. This paper propses a method that can recommend suitable words based on word pattern queries, which are expressed as a combination of words, part-of-speech (POS) tags, and wild card words, such as ‘ {1:2} idea.’ The proposed method enables to recommend some words for the POS tags of a word pattern query, along with their popularity and example sentences in a corpus. To facilitate such query processing, the method first conducts the POS tagging for all the sentences in a corpus. From the tagged sentences, it generates the 2-grams up to 5-grams, which consist of words, POS tags, and the special wild card word symbol ‘*’. It then builds an inverted file-like data structure which keeps the relevant information for each potential word pattern from the n-grams. Due to the large number of word patterns and sentences, the MapReduce algorithms are developed to realize the proposed method and HBase are deployed to manage the inverted file-like data structure. Some experiment results are presented to show the characteristics of the proposed method.

This paper presents a comprehensive survey on knowledge retrieval of historical concepts using semantic web technologies. The scope of this discussion encompasses research material in the field of computing and humanities including but not limited to Cultural Heritage, Museum and Digital Library research. Semantic web techniques used to describe, store, search, query and retrieve historic concepts such as RDF, RDFs, OWL, and Ontologies are discussed along with other extensively researched methods from domains like Natural Language Processing and Computer Vision. The discussion elaborates on Ontological-enabled information retrieval and the concept of new knowledge discovery. This work presents a comprehensive set of problems identified in the field of information retrieval and their solutions through the semantic web along with an insight in the future directions and the open-ended questions in the domain of retrieval of historic concept.

The performance of knowledge-based word sense disambiguation (WSD) is confused with the acquisition of knowledge base and the selection of related feature words. It is difficult to automatically construct high-quality knowledge base and precisely select related words of ambiguous word. Aiming at the problems, the paper proposes word sense disambiguation based on dependency constraint knowledge. The method fully utilizes the advantage of dependency parsing to solve the confusions of knowledge-based WSD. A large-scale corpus is parsed to construct a high-quality dependency constraint knowledge base. The sentence is parsed to obtain the precise dependency constraint cells for each ambiguous word. Based on dependency constraint knowledge base, posterior probability of each sense of ambiguous word on dependency constraint cells is computed. The sense with max posterior probability is selected as right sense. The method has achieved the best performance among unsupervised and knowledge-based methods in SemEval dataset.