Pub Date : 2018-08-01DOI: 10.1109/ICAICTA.2018.8541286
Azka Hanif Imtiyaz, Nur Ulfa Maulidevi
Intelligent agent is specially suited for completing tasks in video games. Intelligent agent in Dota 2 had been proven to be better at professional human with bot that was developed by OpenAI. Other bots haven’t been able to produce good level of performance as generally bots are developed with rule-based approach. This caused the bot to perform as good as the developer’s understanding of the game. One of learning method to be used in this case is opponent modeling; an attempt at modeling opponent based on its behavior. First, bot will play a round of training match to gather environment data. Model is built based on the data that is gathered with opponent’s current action target as the target regression. Prediction from the model is used in bot for consideration in deciding which action is best against opponent’s action. For validation, implemented bot with opponent modeling faced against bot without opponent modeling and also default bot from Dota 2. The results showed that opponent modeling as a component is able to increase the level of performance on the implemented bot. This showed that opponent modeling is able to provide relevant information for the bot to decide the best action.
{"title":"Implementation of Intelligent Agent in Defense of the Ancient 2 through Utilization of Opponent Modeling","authors":"Azka Hanif Imtiyaz, Nur Ulfa Maulidevi","doi":"10.1109/ICAICTA.2018.8541286","DOIUrl":"https://doi.org/10.1109/ICAICTA.2018.8541286","url":null,"abstract":"Intelligent agent is specially suited for completing tasks in video games. Intelligent agent in Dota 2 had been proven to be better at professional human with bot that was developed by OpenAI. Other bots haven’t been able to produce good level of performance as generally bots are developed with rule-based approach. This caused the bot to perform as good as the developer’s understanding of the game. One of learning method to be used in this case is opponent modeling; an attempt at modeling opponent based on its behavior. First, bot will play a round of training match to gather environment data. Model is built based on the data that is gathered with opponent’s current action target as the target regression. Prediction from the model is used in bot for consideration in deciding which action is best against opponent’s action. For validation, implemented bot with opponent modeling faced against bot without opponent modeling and also default bot from Dota 2. The results showed that opponent modeling as a component is able to increase the level of performance on the implemented bot. This showed that opponent modeling is able to provide relevant information for the bot to decide the best action.","PeriodicalId":184882,"journal":{"name":"2018 5th International Conference on Advanced Informatics: Concept Theory and Applications (ICAICTA)","volume":"37 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121631918","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2018-08-01DOI: 10.1109/ICAICTA.2018.8541292
Robert Sebastian Herlim, A. Purwarianti
In natural language processing, the syntactic analysis process (such as constituency parsing) is required to understand word context in the sentence. We propose a modification on using binarization technique alternative and feature multiplication factors for shift-reduce constituency parser using beam search approach and structured learning algorithm. Our modification in binarization technique is inspired from assorted tagging schemes in NER, while the feature multiplication factors is used to scale up our scoring system for beam search algorithm. For evaluation, we mainly used the new INACL Treebank (consisting 11,356 and 4,457 instances for training and test set), resulted 50.3% in f1-score. Our parser also compared with previous work by using the same training and test set for IDN-Treebank, resulted 74.0% in f1-score.
{"title":"Indonesian Shift-Reduce Constituency Parser Using Feature Templates & Beam Search Strategy","authors":"Robert Sebastian Herlim, A. Purwarianti","doi":"10.1109/ICAICTA.2018.8541292","DOIUrl":"https://doi.org/10.1109/ICAICTA.2018.8541292","url":null,"abstract":"In natural language processing, the syntactic analysis process (such as constituency parsing) is required to understand word context in the sentence. We propose a modification on using binarization technique alternative and feature multiplication factors for shift-reduce constituency parser using beam search approach and structured learning algorithm. Our modification in binarization technique is inspired from assorted tagging schemes in NER, while the feature multiplication factors is used to scale up our scoring system for beam search algorithm. For evaluation, we mainly used the new INACL Treebank (consisting 11,356 and 4,457 instances for training and test set), resulted 50.3% in f1-score. Our parser also compared with previous work by using the same training and test set for IDN-Treebank, resulted 74.0% in f1-score.","PeriodicalId":184882,"journal":{"name":"2018 5th International Conference on Advanced Informatics: Concept Theory and Applications (ICAICTA)","volume":"117 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134645130","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2018-08-01DOI: 10.1109/icaicta.2018.8541327
{"title":"ICAICTA 2018 Committees","authors":"","doi":"10.1109/icaicta.2018.8541327","DOIUrl":"https://doi.org/10.1109/icaicta.2018.8541327","url":null,"abstract":"","PeriodicalId":184882,"journal":{"name":"2018 5th International Conference on Advanced Informatics: Concept Theory and Applications (ICAICTA)","volume":"57 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127024052","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2018-08-01DOI: 10.1109/ICAICTA.2018.8541324
I. Suwardi, Achmad Imam Kistijantoro, Tjokorda Agung Budi Wirayuda, Ginar Santika Niwanputri
Facial images are one of the raw data that can be processed to produce various information/representations, especially for computer vision, pattern recognition, and biometrics. Moreover, identity recognition, expression recognition and visitor demographic calculations are applications that can be generated through the processing of facial images. In order to perform face image processing, a face detection mechanism is needed to isolate the face area (region-of-interest-ROI). Previous research generally views facial images as unity for further processing with feature extraction techniques and recognition. This paper proposes post-processing from face detection (Viola-Jones based) to produce a bag of facial components as a data representation for the next processes which are feature extraction and recognition. The post-processing is done based on the geometric rules of the face and golden ratio to produce more accurate detection. From the experiment, the proposed method achieves 96.88% of accuracy on the development part whilst the accuracy of testing part reaches 92.52% (with precision 95.32% and recall 96.62%).
{"title":"Bag of Facial Components: A Data Enrichment Approach for Face Image Processing","authors":"I. Suwardi, Achmad Imam Kistijantoro, Tjokorda Agung Budi Wirayuda, Ginar Santika Niwanputri","doi":"10.1109/ICAICTA.2018.8541324","DOIUrl":"https://doi.org/10.1109/ICAICTA.2018.8541324","url":null,"abstract":"Facial images are one of the raw data that can be processed to produce various information/representations, especially for computer vision, pattern recognition, and biometrics. Moreover, identity recognition, expression recognition and visitor demographic calculations are applications that can be generated through the processing of facial images. In order to perform face image processing, a face detection mechanism is needed to isolate the face area (region-of-interest-ROI). Previous research generally views facial images as unity for further processing with feature extraction techniques and recognition. This paper proposes post-processing from face detection (Viola-Jones based) to produce a bag of facial components as a data representation for the next processes which are feature extraction and recognition. The post-processing is done based on the geometric rules of the face and golden ratio to produce more accurate detection. From the experiment, the proposed method achieves 96.88% of accuracy on the development part whilst the accuracy of testing part reaches 92.52% (with precision 95.32% and recall 96.62%).","PeriodicalId":184882,"journal":{"name":"2018 5th International Conference on Advanced Informatics: Concept Theory and Applications (ICAICTA)","volume":"9 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125122472","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2018-08-01DOI: 10.1109/ICAICTA.2018.8541289
A. Romadhony, A. Purwarianti, D. H. Widyantoro, Alfan Farizki Wicaksono
Open Information Extraction (Open IE), which has been extensively studied as a new paradigm on unrestricted information extraction, produces relation tuples (results) which serve as intermediate structures in several natural language processing tasks, one of which is question answering system. In this paper, we investigate ways to learn the vector representation of Open IE relation tuples using various approaches, ranging from simple vector composition to more advanced methods, such as recursive autoencoder (RAE). The quality of vector representation was evaluated by conducting experiments on the relation tuple similarity task. While the results show that simple linear combination (i.e., averaging the vectors of the words participating in the tuple) outperforms any other methods, including RAE, RAE itself has its own advantage in dealing with a case, in which the similarity criterion is characterized by each element in the tuple, in cases where the simple linear combination is unable to identify them.
开放信息抽取(Open Information Extraction, Open IE)作为一种新的无限制信息抽取范式得到了广泛的研究，它产生的关系元组(结果)在许多自然语言处理任务中充当中间结构，问答系统就是其中之一。在本文中，我们研究了使用各种方法来学习Open IE关系元组的向量表示的方法，从简单的向量组合到更高级的方法，如递归自编码器(RAE)。通过对关系元组相似性任务的实验，评价了向量表示的质量。虽然结果表明，简单线性组合(即对元组中参与的单词的向量取平均值)优于包括RAE在内的任何其他方法，但RAE本身在处理简单线性组合无法识别元组中的每个元素的相似性标准的情况时具有自身的优势。
{"title":"A Dense Vector Representation for Relation Tuple Similarity","authors":"A. Romadhony, A. Purwarianti, D. H. Widyantoro, Alfan Farizki Wicaksono","doi":"10.1109/ICAICTA.2018.8541289","DOIUrl":"https://doi.org/10.1109/ICAICTA.2018.8541289","url":null,"abstract":"Open Information Extraction (Open IE), which has been extensively studied as a new paradigm on unrestricted information extraction, produces relation tuples (results) which serve as intermediate structures in several natural language processing tasks, one of which is question answering system. In this paper, we investigate ways to learn the vector representation of Open IE relation tuples using various approaches, ranging from simple vector composition to more advanced methods, such as recursive autoencoder (RAE). The quality of vector representation was evaluated by conducting experiments on the relation tuple similarity task. While the results show that simple linear combination (i.e., averaging the vectors of the words participating in the tuple) outperforms any other methods, including RAE, RAE itself has its own advantage in dealing with a case, in which the similarity criterion is characterized by each element in the tuple, in cases where the simple linear combination is unable to identify them.","PeriodicalId":184882,"journal":{"name":"2018 5th International Conference on Advanced Informatics: Concept Theory and Applications (ICAICTA)","volume":"42 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127457156","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2018-08-01DOI: 10.1109/ICAICTA.2018.8541326
Bandit Tagmatcha, Komate Amphawan
In the past decade, frequent-regular itemset mining (FRIM) has been proposed and applied in a wide range of applications. It aims to discover interesting itemsets frequently and regularly occurring in a static database. However, in real-world applications, the occurrence behavior of items/itemsets may change whenever the database is updated and there may be the situation of overwhelming or none of results generated if the user set inappropriate support threshold. Thus, we here introduce a new approach to mine top-k frequent-regular itemsets from incremental transactional database for mining results which allows users to control the number of results. In this approach, a set of k itemsets having highest frequency of occurrence and regularity occurring in a incremental database is generated. To mine such itemsets, an efficient single-pass algorithm called IMTFRI (Incremental Miner of Top-k Frequent-Regular Itemset) is proposed. The partitioned dynamic bit-vector is utilized to maintain occurrence information of each item/itemsets while mining. In addition, to avoid mining on each incremental database from scratch, the mining with baseline frequency setting technique is designed. Last, experimental studies have been conducted to investigate efficiency of IMTFRI algorithm in the terms of computational time and memory usage.
{"title":"Mining Top-k Frequent-regular Itemsets from Incremental Transactional Database","authors":"Bandit Tagmatcha, Komate Amphawan","doi":"10.1109/ICAICTA.2018.8541326","DOIUrl":"https://doi.org/10.1109/ICAICTA.2018.8541326","url":null,"abstract":"In the past decade, frequent-regular itemset mining (FRIM) has been proposed and applied in a wide range of applications. It aims to discover interesting itemsets frequently and regularly occurring in a static database. However, in real-world applications, the occurrence behavior of items/itemsets may change whenever the database is updated and there may be the situation of overwhelming or none of results generated if the user set inappropriate support threshold. Thus, we here introduce a new approach to mine top-k frequent-regular itemsets from incremental transactional database for mining results which allows users to control the number of results. In this approach, a set of k itemsets having highest frequency of occurrence and regularity occurring in a incremental database is generated. To mine such itemsets, an efficient single-pass algorithm called IMTFRI (Incremental Miner of Top-k Frequent-Regular Itemset) is proposed. The partitioned dynamic bit-vector is utilized to maintain occurrence information of each item/itemsets while mining. In addition, to avoid mining on each incremental database from scratch, the mining with baseline frequency setting technique is designed. Last, experimental studies have been conducted to investigate efficiency of IMTFRI algorithm in the terms of computational time and memory usage.","PeriodicalId":184882,"journal":{"name":"2018 5th International Conference on Advanced Informatics: Concept Theory and Applications (ICAICTA)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129990737","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2018-08-01DOI: 10.1109/ICAICTA.2018.8541317
Hokuto Tateyama, Shigeru Kuriyama
According to the spread of digitally-controllable decorative illuminations, dimming mechanisms using digital images have been developed for efficiently controlling various pattern of full colors. The color of lights, however, cannot be correctly converted due to the large difference in color gamut between those of LED illuminants and image pixels. This article, therefore, introduces a perception-based color enhancement technology. We can enrich the color of illuminations while preserving hue component of input images and avoiding color saturation, by which the color quality of illuminations can be improved through a simple calibration process.
{"title":"Perceptual Color Enhancement for LED Illuminations","authors":"Hokuto Tateyama, Shigeru Kuriyama","doi":"10.1109/ICAICTA.2018.8541317","DOIUrl":"https://doi.org/10.1109/ICAICTA.2018.8541317","url":null,"abstract":"According to the spread of digitally-controllable decorative illuminations, dimming mechanisms using digital images have been developed for efficiently controlling various pattern of full colors. The color of lights, however, cannot be correctly converted due to the large difference in color gamut between those of LED illuminants and image pixels. This article, therefore, introduces a perception-based color enhancement technology. We can enrich the color of illuminations while preserving hue component of input images and avoiding color saturation, by which the color quality of illuminations can be improved through a simple calibration process.","PeriodicalId":184882,"journal":{"name":"2018 5th International Conference on Advanced Informatics: Concept Theory and Applications (ICAICTA)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130772268","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2018-08-01DOI: 10.1109/ICAICTA.2018.8541344
Kitiya Suriyachay, Virach Sornlertlamvanich
In the Thai language, named entity can be used with or without a prefix or an indication of word. This may cause confusion between named entity and other types of noun. However, a named entity is likely to be used in adjacent to verbs or prepositions. This means that the adjacent verbs or prepositions to a noun can be as a good feature to determine the type of named entity. There are some studies on named entity recognition (NER) task in other languages such as Indonesian showing that combination of word embedding and part-of-speech (POS) tag can improve the performance of the NER model. In this paper, we investigate the Thai Named Entity Recognition task using Bi-LSTM model with word embedding and POS embedding for dealing with the relatively small and disjointedly labeled corpus. We compare our model with the one without POS tag, and the baseline model of CRF with the similar set of feature. The experiment results show that our proposed model outperforms the other two in all F1-score measures. Especially, in the case of location file, the F1-score is increased by 14 percent.
{"title":"Named Entity Recognition Modeling for the Thai Language from a Disjointedly Labeled Corpus","authors":"Kitiya Suriyachay, Virach Sornlertlamvanich","doi":"10.1109/ICAICTA.2018.8541344","DOIUrl":"https://doi.org/10.1109/ICAICTA.2018.8541344","url":null,"abstract":"In the Thai language, named entity can be used with or without a prefix or an indication of word. This may cause confusion between named entity and other types of noun. However, a named entity is likely to be used in adjacent to verbs or prepositions. This means that the adjacent verbs or prepositions to a noun can be as a good feature to determine the type of named entity. There are some studies on named entity recognition (NER) task in other languages such as Indonesian showing that combination of word embedding and part-of-speech (POS) tag can improve the performance of the NER model. In this paper, we investigate the Thai Named Entity Recognition task using Bi-LSTM model with word embedding and POS embedding for dealing with the relatively small and disjointedly labeled corpus. We compare our model with the one without POS tag, and the baseline model of CRF with the similar set of feature. The experiment results show that our proposed model outperforms the other two in all F1-score measures. Especially, in the case of location file, the F1-score is increased by 14 percent.","PeriodicalId":184882,"journal":{"name":"2018 5th International Conference on Advanced Informatics: Concept Theory and Applications (ICAICTA)","volume":"41 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124289492","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2018-08-01DOI: 10.1109/ICAICTA.2018.8541296
T. Thwin, S. Vasupongayya
The blockchain systems are analyzed under the context of the personal health record system (PHRs) requirements. The transparent property of blockchain may cause the privacy and confidentiality concerns for PHRs. The append-only storage of blockchain can be a barrier for implementing the revocability of consent in PHRs. Moreover, the health care data can be very large exceeding the practical storage capabilities of the current blockchain usages. The most important issues of blockchain include the limited storage, privacy, consent revocation, performance, energy consumption and scalability. A blockchain based secret-data sharing model is proposed by using a proxy re-encryption technique to support the PHRs in this work. Some potential attacks which can attempt on the proposed model and how the model can handle such attempts is also discussed.
{"title":"Blockchain Based Secret-Data Sharing Model for Personal Health Record System","authors":"T. Thwin, S. Vasupongayya","doi":"10.1109/ICAICTA.2018.8541296","DOIUrl":"https://doi.org/10.1109/ICAICTA.2018.8541296","url":null,"abstract":"The blockchain systems are analyzed under the context of the personal health record system (PHRs) requirements. The transparent property of blockchain may cause the privacy and confidentiality concerns for PHRs. The append-only storage of blockchain can be a barrier for implementing the revocability of consent in PHRs. Moreover, the health care data can be very large exceeding the practical storage capabilities of the current blockchain usages. The most important issues of blockchain include the limited storage, privacy, consent revocation, performance, energy consumption and scalability. A blockchain based secret-data sharing model is proposed by using a proxy re-encryption technique to support the PHRs in this work. Some potential attacks which can attempt on the proposed model and how the model can handle such attempts is also discussed.","PeriodicalId":184882,"journal":{"name":"2018 5th International Conference on Advanced Informatics: Concept Theory and Applications (ICAICTA)","volume":"68 3 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121411576","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2018-08-01DOI: 10.1109/ICAICTA.2018.8541300
Alson Cahyadi, M. L. Khodra
In order to improve performance of previous aspect-based sentiment analysis (ABSA) on restaurant reviews in Indonesian language, this paper adapts the research achieving the highest F1 at SemEval 2016. We use feedforward neural network with one-vs-all strategy for aspect category classification (Slot 1), Conditional Random Field (CRF) for opinion target expression extraction (Slot 2), and Convolutional Neural Network (CNN) for sentiment polarity classification (Slot 3). Aside from lexical features we also use additional features learned from neural networks. We train our model on 992 sentences and evaluate them on 382 sentences. Higher performances are achieved for Slot 1 (F1 0.870) and Slot 3 (F1 0.764) but lower on Slot 2 (F1 0.787).
{"title":"Aspect-Based Sentiment Analysis Using Convolutional Neural Network and Bidirectional Long Short-Term Memory","authors":"Alson Cahyadi, M. L. Khodra","doi":"10.1109/ICAICTA.2018.8541300","DOIUrl":"https://doi.org/10.1109/ICAICTA.2018.8541300","url":null,"abstract":"In order to improve performance of previous aspect-based sentiment analysis (ABSA) on restaurant reviews in Indonesian language, this paper adapts the research achieving the highest F1 at SemEval 2016. We use feedforward neural network with one-vs-all strategy for aspect category classification (Slot 1), Conditional Random Field (CRF) for opinion target expression extraction (Slot 2), and Convolutional Neural Network (CNN) for sentiment polarity classification (Slot 3). Aside from lexical features we also use additional features learned from neural networks. We train our model on 992 sentences and evaluate them on 382 sentences. Higher performances are achieved for Slot 1 (F1 0.870) and Slot 3 (F1 0.764) but lower on Slot 2 (F1 0.787).","PeriodicalId":184882,"journal":{"name":"2018 5th International Conference on Advanced Informatics: Concept Theory and Applications (ICAICTA)","volume":"35 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132699171","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：