Rachada Kongkachandra

Abductive Reasoning for Keyword Recovering in Semantic-Based Keyword Extraction

This paper proposes semantic based keyphrase recovery for domain-independent keyphrase extraction. In this method, we add a keyphrase recovery function as a post- process of the conventional keyphrase extractors in order to reconsider the failed keyphrases by semantic matching based on sentence meaning. We also add the Domain Identification Function to determine the related domain of the keyphrases by using keyphrases extracted from the conventional systems in order to make the system as domain-independent. The semantic matching is performed to compare the similar meanings between ones of failed keyphrases and ones in the knowledge base. Therefore, the failed keyphrases that are matched by semantic matching are recused as keyphrases. The experiments with the summary sentences in 60 articles of IEICE Transactions on Information and Systems and glossaries from four resources are performed in initializing Domain Knowledge Base. Other summary sentences in 100 articles of IEICE Transactions on Information and Systems and in 15 chapters in a Computer Information System textbook are experimented in recovering the failed keyphrases. The results reveal that the proposed method increases the average performance of conventional EXTRACTOR and KEA approximately by 33.16 and 41.30% of precision, and 36.10 and 39.17% of recall, respectively.

Improving Thai word segmentation with Named Entity Recognition

Segmenting words in Thai language is a very difficult task since there is no distinguished clue such as blank, period and other punctuations as in English. Several previous researches employed dictionary as the main resource for consideration. However there still exist two problems including ambiguous words and unknown words. These unknown words can be categorized into two groups, -i.e., newly defined words and named entities. This paper presents an approach for improving the performance of Thai word segmentation by merging Named Entity Recognition (NER) to the Thai word segmentation. The Conditional Random Fields (CRFs) algorithm is applied for training and recognizing Thai named entities. The prefixes and suffixes of Thai named entities are selected as main features for learning the models. The performance evaluations are experimented by using the Thai standard word segmentation corpus, namely BEST20091, which consists of 5 million words. Various word-level grams (i.e., three, five and seven) are also employed to construct the Thai NER models. The experimental results show that the 7-gram NER model provides the best performance. Merging the proposed NER model to the Thai word segmentation called TLex (Thai Lexeme Analyzer) can improve the performance measured by F1-measure from 92.39% to 93.96%.

Improvement of word alignment in thai-english statistical machine translation by grammatical attributes identification

This paper presents a method to handle difference of Thai and English language in an alignment process for statistical machine translation (SMT). By identification of grammar notations within both texts, the method can analyze a type of the grammatical attribute and differently handle both Thai and English words accordingly based on linguistic knowledge. This method works as a pre-process of a standard co-occurrence alignment, GIZA. An experimental result showed that this method gained 48% higher accuracy result than the widely used conventional alignment. We can conclude that a different grammatical attribute should be pre-process handled since this issue greatly affects the result of bilingual alignment and SMT.

Clustered microcalcification classification using CC-MLO-View corresponding shape and distribution features

Shape of single microcalcifications (muCa++s) and distribution of them in a cluster are two key features for a radiologist to diagnose this abnormality appearing on mammograms into benign type or malignant type of breast cancer. These two features from two-dimensional (2-D) mammogram image from two mammographic views, cranio-caudad view (CC) and medio-lateral oblique view (MLO), are inevitable conflicted because of lack of depth information. It makes a large contradictory information of the same microcalcification cluster in different view. This paper proposes to use three-dimensional (3-D) shape and distribution features exacted from the view correspondence. To identify a 3-D position of microcalcifications, the candidate pairs in CC view and MLO view are stereo-matched based on their relative intensity and size. Occluded microcalcifications are separated by x-ray absorption property. The 3-D shape features are represented by their structural outline, spherical measurement, and thickness which are computed from Fourier descriptor of surface outline, compactness and its intensity, respectively. The distribution feature is represented by 3-D cluster size, average distance between each microcalcifications, and cluster density. There are 12 features used as input features for three-layer feed-forward backpropagation neural network classifier which is constructed dynamically and weighted be training with forty benign and forty malignant microcalcifications. The evaluated performance of the proposed method is 96 percent sensitivity and 91 percent specificity.

Improving Thai-English word alignment for interrogative sentences in SMT by grammatical knowledge

This paper presents a method to improve Thai-English word alignment in statistical machine translation (SMT) for interrogative sentences in a parallel corpus. We utilize the Thai and English grammatical knowledge i.e. tense, part of speech (POS), and question inversion pattern. The proposed method handles the difference of Thai and English interrogative sentences using sentence transformation, interrogative grammatical attribute extraction, and interrogative grammatical attribute annotation. This method works as a pre-processing of GIZA, a standard word co-occurrence alignment tool in SMT. We hypothesize that using grammatical knowledge as a pre-processing of GIZA can provide higher accuracy. We experiment by using 43,500 interrogative sentences to compare alignment result between interrogative sentences attached an interrogative grammatical label and interrogative sentences unattached interrogative grammatical label. The experimental results yield 95% of accuracy with significant improvement than the conventional one. With the increasing accuracy of word alignment, the translation accuracy is consequently improved.

Passive monitoring method for analysis Quantum Key Distribution performance statistics

This paper emphasizes the development of network monitoring appliance for monitoring network health, network resources and for further analysis of Quantum Key Distribution (QKD) performance where QKD devices located in different locations. It provides tools and services for end users who enable to view real time QKD performance statistics on the network. As it is very important to keep track of each key parameter involved with device status and QKD temporal evolution to perform traffic analysis and to determine the error probability during the phase of key generation. The system has been developed from a Network Health Analysis and Monitoring system (NetHAM) [1] implemented under an open source platform, offered more features both graphical statistics display and user friendly interface. In addition, this paper proposed a conformance passive monitoring method to analysis the QKD performance. The log from QKD layer has been converted to well organized database format before storing it in the database server. The relevant information will be invoked for network topology display, link and device status that would be helpful for end user can easily view the QKD performance with frequently updated through the dynamic graphical display and for further historical data analysis.

Automatic Ripeness Grading of Durian Pulp Using Color Histograms and Density

Abstract Grading of durian is an important preprocess in durian production line. In this process, since destructive ways such as manual way, etc lose apart of pulp. processing time and quality, nondestructive way is required. This paper proposes an automatic ripeness grading of durian pulp by using color histograms and density. In this method, both color and density of durian pulp are utilized to classify the pulp ripeness. Since in case of ripe pulp, the color changes from white to yellow and density increase, we employ color obtained from video cameras. density from video cameras and weight measurement unit. Ripe pulp is graded by determining appropriate threshold values of color and density. In the experiments, we can classify the ripeness of durian pulp with 87.5% of accuracy