DE eng

Search in the Catalogues and Directories

Hits 1 – 8 of 8

1
Application of artificial intelligence in gastrointestinal disease: a narrative review
In: Ann Transl Med (2021)
BASE
Show details
2
What makes an effective translation teacher?: a qualitative exploration of effective translation teaching and teachers in the university classroom
Huang, Zhi. - : Sydney, Australia : Macquarie University, 2019
BASE
Show details
3
Highly elastic and conductive graphene [<Journal>]
Huang, Zhi-Ming [Verfasser]; Liu, Xiao-Yu [Sonstige]; Wu, Wen-Gang [Sonstige].
DNB Subject Category Language
Show details
4
Theme choices in translation and target readers' reactions to different theme choices
Kim, Mira; Huang, Zhi. - : Ewha Research Institute for Translation Studies, 2012
BASE
Show details
5
Linear programming method for multiattribute group decision making using IF sets
In: Information sciences. - New York, NY : Elsevier Science Inc. 180 (2010) 9, 1591-1609
OLC Linguistik
Show details
6
What makes a successful EFL teacher in China? A case study of an English language teacher at Nanjing University of Chinese Medicine
Huang, Zhi. - : Canadian Center of Science and Education, 2010
BASE
Show details
7
What Makes a Successful EFL Teacher in China? A Case Study of an English Language Teacher at Nanjing University of Chinese Medicine
In: English Language Teaching; Vol 3, No 3 (2010); P20 (2010)
BASE
Show details
8
高精準度線上手寫中文文句注音辨識 ; High-accuracy on-line handwritten Chinese text phonetic sequence recognition
Abstract: 本論文提出高精準度線上手寫中文文句注音辨識系統,其中包含三個子系統,一個是手寫中文注音符號辨識子系統、手寫中文有效注音序列(effective phonetic sequence,簡稱EPS)辨識子系統和Web-based language model子系統。第一,本論文提出一高精準度的線上手寫中文注音符號辨識子系統,該系統在訓練階段,首先建立含37個注音符號參考字形之注音符號參考資料庫,並針對特定注音字形分別作結構性和限制性合併的標記。在辨識階段,使用者之手寫注音字形經線段抽取及大小正規化後,即與各參考注音符號字形作使用結構性特徵之動態規劃比對,並能得知手寫字形與參考字形間的線段對應關係,進一步提供特定參考注音與手寫字形作限制性合併之動態規劃比對,經實驗證明本系統辨識單一線上手寫注音之辨識率已可達99.95%。第二,本論文提出一高精準度線上手寫EPS辨識子系統,全部中文字都可以由1,211個EPS合成,但必須產生3,168個EPS參考字形(含13個注音符號之25種變異筆順寫法)之EPS參考資料庫,在訓練階段,每個參考EPS是由多個單一的注音符號所組成的,在辨識階段,先將手寫EPS字形進行聲調辨識,以決定是否去除本身聲調符號,然後,將手寫EPS字形與所有同聲調EPS參考字形進行動態規劃比對,並得到n名候選EPS,依該手寫EPS字形與n個候選EPS的分割暗筆線段之對應,將每個拆開部份均對應到候選EPS的一個特定注音符號,並重新正規化與對應的參考注音符號字形進行動態規劃比對,即可重新計算手寫EPS字形與每個候選EPS的精確相異值,以重新排序候選EPS。實驗結果證明,該子系統的辨識準確率是99.97%。第三,本論文提出一個以Web-based language model子系統以供線上手寫EPS辨識系統作後處理的動作,該子系統不僅涉及詞典裡的詞,而且還使用在Internet的文句。除了搜尋詞典裡的詞,該子系統更進一步搜尋在Internet上通過搜索引擎找出的文句資訊,以供輸入手寫中文文句有效注音序列作後處理的動作。根據實驗結果,此Web-based language model子系統可以大大提高辨識同音詞能力。 ; This thesis presents a high-accuracy online handwritten Chinese text phonetic sequence recognition system, which includes three subsystems, a handwritten Chinese phonetic symbol recognition subsystem, a Chinese handwritten effective phonetic sequence(EPS) recognition subsystem and a Web-based language model subsystem. First, this thesis proposes a high-accuracy on-line handwritten Chinese phonetic symbol recognition subsystem. In the training stage, the system creates the database which contains the reference handwritten patterns of 37 Chinese phonetic symbols. The reference patterns of some specific phonetic symbols contain the tags for structure features and limited merging operation. In the recognition stage, the stroke segments of the input handwritten phonetic symbol are first extracted and are then normalized. Next, dynamic programming is performed to match the input stroke segments with those of a reference phonetic symbol. Based on the stroke-matching result, the limited merging operation can be further performed to adjust the result of matching the input handwritten symbol with a specific reference phonetic symbol having the tag of limited merging operation. The experimental result shows that the subsystem’s recognition accuracy is 99.95%. Second, this thesis proposes a high-accuracy on-line handwritten Chinese EPS recognition subsystem. There are totally 1,211 EPS in Chinese. But there are up to 3,168 EPS handwritten references in the database because there are 25 more alternative writing styles for 13 phonetic symbols. In the training stage, each EPS handwritten reference is generated by combining the reference patterns of its component phonetic symbols. In the recognition stage, the tone of an input handwritten EPS is decided and is removed first. Then, each EPS reference with the same tone is matched with the input handwritten EPS. The first n best EPS references are selected as the candidates. According to the component phonetic symbols of each candidate EPS reference, the input handwritten EPS is separated into corresponding parts. Each separated part is then normalized and is matched with the reference pattern of the corresponding phonetic symbol by dynamic programming. We thus can recalculate the total difference between the input handwritten EPS and a candidate EPS. The order of candidate EPSs is rescheduled according the newly calculated total differences. The experimental result shows that the subsystem’s recognition accuracy is 99.97%. Third, this thesis proposes a Web-based language model subsystem to perform postprocessing after the input handwritten EPS recognition. The subsystem involves not only the words in a dictionary but also the contexts in the Internet. Following the search of the words in dictionary, the subsystem further searches the Internet through the search engine to propose the context information of the input handwritten Chinese text phonetic sequence. According to the experimental results, the new proposed Web-based language model subsystem can greatly improve the ability of discriminating the words with the same pronunciation.
Keyword: effective phonetic sequence;Web-based language model;handwritten phonetic symbol recognition;dynamic programming;handwritten EPS recognition;structure feature; 結構性特徵;動態規劃;手寫注音符號辨識;有效注音序列;手寫EPS辨識;Web-based language model
URL: http://140.127.82.166/bitstream/987654321/17005/-1/096NPC05396012-001.pdf
http://140.127.82.166/handle/987654321/17005
BASE
Hide details

Catalogues
0
0
1
0
1
0
0
Bibliographies
0
0
0
0
0
0
0
0
0
Linked Open Data catalogues
0
Online resources
0
0
0
0
Open access documents
6
0
0
0
0
© 2013 - 2024 Lin|gu|is|tik | Imprint | Privacy Policy | Datenschutzeinstellungen ändern