International Journal of Computer Processing of Languages

This paper makes an attempt to segment the handwritten Hindi words. The problem of segmentation is compounded by the possible presence of modifiers known as matras on all sides of the basic characters and due to the uncertainty introduced in the character shapes by way of different writing styles. We have devised a structural approach to capture the similarities and differences between structure classes. The segmentation is performed in hierarchical order: 1) Separating the upper modifiers and header line from the character, 2) Detecting and then segmenting lower modifiers from the characters, 3) Identifying whether a character is conjunct or not, 4) Categorization of top modifiers based on Check_point, Mid_point and Touching_points. The segmentation accuracy has been found to be around 78%. Some general conditions are applied for separating modifiers from the characters. But certain words cannot be segmented because they violate the general conditions. However, specifics are not dealt with in this paper because such an attempt requires an exhaustive study on a large database that is not available presently.

This paper presents the design and construction of an annotated Chinese collocation bank as the resource to support systematic research on Chinese collocations. The definition and properties are first studied. Based on a combination of different properties, a classification scheme is proposed to categorize Chinese collocations into four types. With the help of computational tools, bigram collocations and n-gram collocations of 3,643 headwords are manually identified in a 5-million-word corpus. Furthermore, for each identified bigram collocation, its dependency relation, chunking information and classification are annotated to produce a collocation bank. Currently, the Chinese collocation bank contains 23,581 bigram collocations and 2,752 n-gram collocations. The Chinese collocation bank is a valuable resource for Chinese collocation related research. Through statistical analysis on the collocation bank, some interesting characteristics of Chinese bigram collocations are presented in this paper.

The Arabic letters are used in many written languages. However, little work has been done to analyze and characterize handwritten Arabic letters comprehensively. Such characterization is important for the active research in the computer processing of Arabic written scripts. We extract carefully selected features from a large database of handwritten Arabic letters, from the letter's secondary components, main body, skeleton, and boundary. These features are studied and statistically analyzed to reach the targeted characterization. Observations about the important writing style variations are presented and statistically specified. The Arabic letters have multiple forms depending on the letter's position in the word. Comparisons among the four main letter forms (isolated, initial, medial, and final) are also presented.

In this paper, character-based Japanese dictation method is proposed. This method is based on the kana and kanji string language model proposed by Ito et al. First, sentences in the training corpus are split into character-based units (CBUs). Then strings of CBUs (CBUSes) are chosen from the CBU corpus based on a statistical criterion. We examined three criteria for the CBUS selection. They are the frequency-based selection, the mutual-information based selection and their combination. From the experimental results, it was found that the combined method gave the best result (7.19% and 8.75% CBU error rates for the 20k and the 60k word vocabulary conditions, respectively) which was better than the ordinary word-based method (7.61% and 9.15% CBU error rates for the 20k and the 60k word vocabulary conditions, respectively).

In addition, we carried out a recognition experiment for the Corpus of Spontaneous Japanese to confirm that the proposed method is effective for not only the read speech but also for spontaneous speech. As a result, we obtained the best result (29.82%) using the frequency-based method, which is better than the word-based recognition result (32.80%).

As an important task in Topic Detection and Tracking (TDT), New Event Detection (NED) aims to monitor the stream of news stories and detect new events reported by the first story of a topic. This paper proposes a Temporal Topic Model (TTM) to describe a topic as a series of events corresponding to different time. NED, based on TTM, firstly identifies whether a story includes the same time expressions with old topics, and then it verifies whether the story and the topics include relevant events corresponding to the expressions. Thus, a story will be determined as a new event, if it includes much few simultaneous relevant events of old topics. Additionally, this paper analyzes the distribution of time expressions to identify both seminal and novel events, by which NED modifies the probabilities of stories to be new events based on whether they include seminal or novel events of old topics. We compare our methods with some existing NED systems on TDT4 corpus, which demonstrate our methods substantially improve the efficiency and accuracy of NED.