Pub Date : 2015-03-02DOI: 10.1109/ICOSC.2015.7050852
F. Amato, R. Boselli, M. Cesarini, Fabio Mercorio, Mario Mezzanzanica, V. Moscato, Fabio Persia, A. Picariello
Today the Web represents a rich source of labour market data for both public and private operators, as a growing number of job offers are advertised through Web portals and services. In this paper we apply and compare several techniques, namely explicit-rules, machine learning, and LDA-based algorithms to classify a real dataset of Web job offers collected from 12 heterogeneous sources against a standard classification system of occupations.
{"title":"Challenge: Processing web texts for classifying job offers","authors":"F. Amato, R. Boselli, M. Cesarini, Fabio Mercorio, Mario Mezzanzanica, V. Moscato, Fabio Persia, A. Picariello","doi":"10.1109/ICOSC.2015.7050852","DOIUrl":"https://doi.org/10.1109/ICOSC.2015.7050852","url":null,"abstract":"Today the Web represents a rich source of labour market data for both public and private operators, as a growing number of job offers are advertised through Web portals and services. In this paper we apply and compare several techniques, namely explicit-rules, machine learning, and LDA-based algorithms to classify a real dataset of Web job offers collected from 12 heterogeneous sources against a standard classification system of occupations.","PeriodicalId":126701,"journal":{"name":"Proceedings of the 2015 IEEE 9th International Conference on Semantic Computing (IEEE ICSC 2015)","volume":"19 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2015-03-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125633256","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 : 2015-03-02DOI: 10.1109/ICOSC.2015.7050801
M. Kanakaraj, R. R. Guddeti
Mining opinions and analyzing sentiments from social network data help in various fields such as even prediction, analyzing overall mood of public on a particular social issue and so on. This paper involves analyzing the mood of the society on a particular news from Twitter posts. The key idea of the paper is to increase the accuracy of classification by including Natural Language Processing Techniques (NLP) especially semantics and Word Sense Disambiguation. The mined text information is subjected to Ensemble classification to analyze the sentiment. Ensemble classification involves combining the effect of various independent classifiers on a particular classification problem. Experiments conducted demonstrate that ensemble classifier outperforms traditional machine learning classifiers by 3-5%.
{"title":"Performance analysis of Ensemble methods on Twitter sentiment analysis using NLP techniques","authors":"M. Kanakaraj, R. R. Guddeti","doi":"10.1109/ICOSC.2015.7050801","DOIUrl":"https://doi.org/10.1109/ICOSC.2015.7050801","url":null,"abstract":"Mining opinions and analyzing sentiments from social network data help in various fields such as even prediction, analyzing overall mood of public on a particular social issue and so on. This paper involves analyzing the mood of the society on a particular news from Twitter posts. The key idea of the paper is to increase the accuracy of classification by including Natural Language Processing Techniques (NLP) especially semantics and Word Sense Disambiguation. The mined text information is subjected to Ensemble classification to analyze the sentiment. Ensemble classification involves combining the effect of various independent classifiers on a particular classification problem. Experiments conducted demonstrate that ensemble classifier outperforms traditional machine learning classifiers by 3-5%.","PeriodicalId":126701,"journal":{"name":"Proceedings of the 2015 IEEE 9th International Conference on Semantic Computing (IEEE ICSC 2015)","volume":"102 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2015-03-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115631453","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 : 2015-03-02DOI: 10.1109/ICOSC.2015.7050795
Qingliang Miao, Yao Meng, Bo Zhang
Enterprise knowledge graph is crucial for both enterprises and their management agencies. However, enterprise knowledge graph construction faces several challenges such as heterogeneous taxonomies, knowledge inconsistencies or conflicts and lack of semantic links. In this paper, we use Linked Data paradigm to construct enterprise knowledge graph by integrating heterogeneous enterprise data itself as well as link enterprise data with external data. Preliminary experiment on real world dataset shows the proposed approach is effective.
{"title":"Chinese enterprise knowledge graph construction based on Linked Data","authors":"Qingliang Miao, Yao Meng, Bo Zhang","doi":"10.1109/ICOSC.2015.7050795","DOIUrl":"https://doi.org/10.1109/ICOSC.2015.7050795","url":null,"abstract":"Enterprise knowledge graph is crucial for both enterprises and their management agencies. However, enterprise knowledge graph construction faces several challenges such as heterogeneous taxonomies, knowledge inconsistencies or conflicts and lack of semantic links. In this paper, we use Linked Data paradigm to construct enterprise knowledge graph by integrating heterogeneous enterprise data itself as well as link enterprise data with external data. Preliminary experiment on real world dataset shows the proposed approach is effective.","PeriodicalId":126701,"journal":{"name":"Proceedings of the 2015 IEEE 9th International Conference on Semantic Computing (IEEE ICSC 2015)","volume":"18 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2015-03-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"117128015","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 : 2015-03-02DOI: 10.1109/ICOSC.2015.7050820
Hsin-Yu Ha, Shu‐Ching Chen, Min Chen
Feature selection is an actively researched topic in varies domains, mainly owing to its ability in greatly reducing feature space and associated computational time. Given the explosive growth of high-dimensional multimedia data, a well-designed feature selection method can be leveraged in classifying multimedia contents into high-level semantic concepts. In this paper we present a multi-phase feature selection method using maximum spanning tree built from feature correlation among multiple modalities (FC-MST). The method aims to first thoroughly explore not only the correlation between features within and across modalities, but also the association of features towards semantic concepts. Secondly, with the correlations, we identify important features and exclude redundant or irrelevant ones. The proposed method is tested on a well-known benchmark multimedia data set called NUS-WIDE and the experimental results show that it outperforms four well-known feature selection methods in all three important measurement metrics.
{"title":"FC-MST: Feature correlation maximum spanning tree for multimedia concept classification","authors":"Hsin-Yu Ha, Shu‐Ching Chen, Min Chen","doi":"10.1109/ICOSC.2015.7050820","DOIUrl":"https://doi.org/10.1109/ICOSC.2015.7050820","url":null,"abstract":"Feature selection is an actively researched topic in varies domains, mainly owing to its ability in greatly reducing feature space and associated computational time. Given the explosive growth of high-dimensional multimedia data, a well-designed feature selection method can be leveraged in classifying multimedia contents into high-level semantic concepts. In this paper we present a multi-phase feature selection method using maximum spanning tree built from feature correlation among multiple modalities (FC-MST). The method aims to first thoroughly explore not only the correlation between features within and across modalities, but also the association of features towards semantic concepts. Secondly, with the correlations, we identify important features and exclude redundant or irrelevant ones. The proposed method is tested on a well-known benchmark multimedia data set called NUS-WIDE and the experimental results show that it outperforms four well-known feature selection methods in all three important measurement metrics.","PeriodicalId":126701,"journal":{"name":"Proceedings of the 2015 IEEE 9th International Conference on Semantic Computing (IEEE ICSC 2015)","volume":"24 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2015-03-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"117188327","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 : 2015-03-02DOI: 10.1109/ICOSC.2015.7050849
Zhengping Wu
In today's cloud service market, different providers have very different low-level mechanisms to accommodate various types of policies from their users. Enforcement of policies over multiple cloud provider domains is an intrinsically complex problem for both sides. In reality, cloud providers have to either manually update enforcement mechanisms or negotiate adjusted policies with their users for enforcement. To save these high-cost and error-prone manual updates or adjustments, an automatic and flexible solution is desired. This paper proposes a semantic modeling and mapping based approach to help enforce high-level user policies across cloud domain boundaries when applications or IT operations have to span over multiple cloud domains. This approach creates policy models and maps these models across cloud domain boundaries automatically or semi-automatically. Policy rules following these mappings can be tied to multiple enforcement mechanisms in different cloud domains. If a rule cannot be mapped, a manual adjustment solution will be suggested. A case study is also included to demonstrate the efficiency and accuracy of this approach.
{"title":"Multi-cloud policy enforcement through semantic modeling and mapping","authors":"Zhengping Wu","doi":"10.1109/ICOSC.2015.7050849","DOIUrl":"https://doi.org/10.1109/ICOSC.2015.7050849","url":null,"abstract":"In today's cloud service market, different providers have very different low-level mechanisms to accommodate various types of policies from their users. Enforcement of policies over multiple cloud provider domains is an intrinsically complex problem for both sides. In reality, cloud providers have to either manually update enforcement mechanisms or negotiate adjusted policies with their users for enforcement. To save these high-cost and error-prone manual updates or adjustments, an automatic and flexible solution is desired. This paper proposes a semantic modeling and mapping based approach to help enforce high-level user policies across cloud domain boundaries when applications or IT operations have to span over multiple cloud domains. This approach creates policy models and maps these models across cloud domain boundaries automatically or semi-automatically. Policy rules following these mappings can be tied to multiple enforcement mechanisms in different cloud domains. If a rule cannot be mapped, a manual adjustment solution will be suggested. A case study is also included to demonstrate the efficiency and accuracy of this approach.","PeriodicalId":126701,"journal":{"name":"Proceedings of the 2015 IEEE 9th International Conference on Semantic Computing (IEEE ICSC 2015)","volume":"37 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2015-03-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124707988","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 : 2015-03-02DOI: 10.1109/ICOSC.2015.7050774
T. C. Chinsha, Shibily Joseph
Opinion mining or sentiment analysis is the process of analysing the text about a topic written in a natural language and classify them as positive negative or neutral based on the humans sentiments, emotions, opinions expressed in it. Nowadays, the opinions expressed through reviews are increasing day by day on the web. It is practically impossible to analyse and extract opinions from such huge number of reviews manually. To solve this problem an automated opinion mining approach is needed. This task of automatic opinion mining can be done mainly at three different levels, which are document level, sentence level and aspect level. Most of the previous work is in the field of document or sentence level opinion mining. This paper focus on aspect level opinion mining and propose a new syntactic based approach for it, which uses syntactic dependency, aggregate score of opinion words, SentiWordNet and aspect table together for opinion mining. The experimental work was done on restaurant reviews. The dataset of restaurant reviews was collected from web and tagged manually. The proposed method achieved total accuracy of 78.04% on the annotated test set. The method was also compared with the method, which uses Part-Of-Speech tagger for feature extraction; the obtained results show that the proposed method gives 6% more accuracy than previous one on the annotated test set.
{"title":"A syntactic approach for aspect based opinion mining","authors":"T. C. Chinsha, Shibily Joseph","doi":"10.1109/ICOSC.2015.7050774","DOIUrl":"https://doi.org/10.1109/ICOSC.2015.7050774","url":null,"abstract":"Opinion mining or sentiment analysis is the process of analysing the text about a topic written in a natural language and classify them as positive negative or neutral based on the humans sentiments, emotions, opinions expressed in it. Nowadays, the opinions expressed through reviews are increasing day by day on the web. It is practically impossible to analyse and extract opinions from such huge number of reviews manually. To solve this problem an automated opinion mining approach is needed. This task of automatic opinion mining can be done mainly at three different levels, which are document level, sentence level and aspect level. Most of the previous work is in the field of document or sentence level opinion mining. This paper focus on aspect level opinion mining and propose a new syntactic based approach for it, which uses syntactic dependency, aggregate score of opinion words, SentiWordNet and aspect table together for opinion mining. The experimental work was done on restaurant reviews. The dataset of restaurant reviews was collected from web and tagged manually. The proposed method achieved total accuracy of 78.04% on the annotated test set. The method was also compared with the method, which uses Part-Of-Speech tagger for feature extraction; the obtained results show that the proposed method gives 6% more accuracy than previous one on the annotated test set.","PeriodicalId":126701,"journal":{"name":"Proceedings of the 2015 IEEE 9th International Conference on Semantic Computing (IEEE ICSC 2015)","volume":"23 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2015-03-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121648619","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 : 2015-03-02DOI: 10.1109/ICOSC.2015.7050816
S. Gaurav, Y. Jithendranath, Aruna Adil, Sudhakar Yadav, B. Kasturi
Today, search engines play a vital role in accessing the online content. However, the data in the webpages are not clearly perceived by search engines. As a result, it provides a lot of irrelevant data with little desired information. In addition, it takes a lot of time in searching the appropriate result. By studying the online educational needs of Indian school children, we aim to retrieve appropriate educational information in less time through effective search. Schema.org [5] is a collection of markups which helps webmasters to mark up the webpages for retrieval of relevant information. But, properties related to education are not covered completely. Learning Resource Metadata Initiative (LRMI) [9] has created few properties for education and added in schema.org. We map our study with LRMI's work, and propose some new properties as an extension to the schema, which can be useful for students and teachers.
{"title":"A study to assess and enhance educational specific search on web for school children","authors":"S. Gaurav, Y. Jithendranath, Aruna Adil, Sudhakar Yadav, B. Kasturi","doi":"10.1109/ICOSC.2015.7050816","DOIUrl":"https://doi.org/10.1109/ICOSC.2015.7050816","url":null,"abstract":"Today, search engines play a vital role in accessing the online content. However, the data in the webpages are not clearly perceived by search engines. As a result, it provides a lot of irrelevant data with little desired information. In addition, it takes a lot of time in searching the appropriate result. By studying the online educational needs of Indian school children, we aim to retrieve appropriate educational information in less time through effective search. Schema.org [5] is a collection of markups which helps webmasters to mark up the webpages for retrieval of relevant information. But, properties related to education are not covered completely. Learning Resource Metadata Initiative (LRMI) [9] has created few properties for education and added in schema.org. We map our study with LRMI's work, and propose some new properties as an extension to the schema, which can be useful for students and teachers.","PeriodicalId":126701,"journal":{"name":"Proceedings of the 2015 IEEE 9th International Conference on Semantic Computing (IEEE ICSC 2015)","volume":"6 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2015-03-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132196765","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 : 2015-03-02DOI: 10.1109/ICOSC.2015.7050825
Sunila Gollapudi
We are seeing a sea change down the pike in terms of financial information aggregation and consumption; this could potentially be a game changer in financial services space with focus on ability to commoditize data. Financial Services Industry deals with a tremendous amount of data that varies in its structure, volume and purpose. The data is generated in the ecosystem (its customers, its own accounts, partner trades, securities transactions etc.), is handled by many systems - each having its own perspective. Front-office systems handle transactional behavior of the data, middle office systems which typically work with a drop-copy of the data subject it to intense processing, business logic, computations (such as inventory positions, fee calculations, commissions) and the back office systems deal with reconciliation, cleansing, exception management etc. Then there are the analytic systems which are concerned with auditing, compliance reporting as well as business analytics. Data that flows through this ecosystem gets aggregated, transformed, and transported time and again. Traditional approaches to managing such data leverage Extract-Transform-Load (ETL) technologies to set up data marts where each data mart serves a specific purpose (such as reconciliation or analytics). The result is proliferation of transformations and marts in the Organization. The need is to have architectures and IT systems that can aggregate data from many such sources without making any assumptions on HOW, WHERE or WHEN this data will be used. The incoming data is semantically annotated and stored in the triple store within storage tier and offers the ability to store, query and draw inferences using the ontology. There is a probable need for a Big Data Solution here that helps ease data liberation and co-location. This paper is a summary of one such business case of the Financial Services Industry where traditional ETL silos was broken to support the structurally dynamic, ever expanding and changing data usage needs employing Ontology and Semantic techniques like RDF/RDFS, SPARQL, OWL and related stack.
{"title":"Aggregating financial services data without assumptions: A semantic data reference architecture","authors":"Sunila Gollapudi","doi":"10.1109/ICOSC.2015.7050825","DOIUrl":"https://doi.org/10.1109/ICOSC.2015.7050825","url":null,"abstract":"We are seeing a sea change down the pike in terms of financial information aggregation and consumption; this could potentially be a game changer in financial services space with focus on ability to commoditize data. Financial Services Industry deals with a tremendous amount of data that varies in its structure, volume and purpose. The data is generated in the ecosystem (its customers, its own accounts, partner trades, securities transactions etc.), is handled by many systems - each having its own perspective. Front-office systems handle transactional behavior of the data, middle office systems which typically work with a drop-copy of the data subject it to intense processing, business logic, computations (such as inventory positions, fee calculations, commissions) and the back office systems deal with reconciliation, cleansing, exception management etc. Then there are the analytic systems which are concerned with auditing, compliance reporting as well as business analytics. Data that flows through this ecosystem gets aggregated, transformed, and transported time and again. Traditional approaches to managing such data leverage Extract-Transform-Load (ETL) technologies to set up data marts where each data mart serves a specific purpose (such as reconciliation or analytics). The result is proliferation of transformations and marts in the Organization. The need is to have architectures and IT systems that can aggregate data from many such sources without making any assumptions on HOW, WHERE or WHEN this data will be used. The incoming data is semantically annotated and stored in the triple store within storage tier and offers the ability to store, query and draw inferences using the ontology. There is a probable need for a Big Data Solution here that helps ease data liberation and co-location. This paper is a summary of one such business case of the Financial Services Industry where traditional ETL silos was broken to support the structurally dynamic, ever expanding and changing data usage needs employing Ontology and Semantic techniques like RDF/RDFS, SPARQL, OWL and related stack.","PeriodicalId":126701,"journal":{"name":"Proceedings of the 2015 IEEE 9th International Conference on Semantic Computing (IEEE ICSC 2015)","volume":"33 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2015-03-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114791171","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 : 2015-03-02DOI: 10.1109/ICOSC.2015.7050850
Minjae Song, Sungkwang Eom, Sangjin Shin, Kyong-Ho Lee
With the increasing number of mobile devices, there have been many researches on searching and managing a large volume of mobile data. Most of the mobile platforms today provide users with keyword-based full text search (FTS) in order to search for mobile data. Recently, voice search interfaces have been deployed. These search methods, however, query only the keywords given as an input to local databases in mobile devices. Therefore, it is quite difficult to figure out and to provide what a user really wants. To overcome this limitation, we propose a semantic search method for mobile platforms. The proposed method augments the results of semantic search on local databases with their related useful Web information according to the intention and context information of a user. Although there are various semantic search techniques, it is hard to apply the existing methods to mobile devices due to the characteristics of mobile devices such as isolated database structures and limited computing resources. To enable semantic search on mobile devices, we also propose a lightweight mobile ontology. The proposed mobile ontology is also aligned with related Web information to enrich search results. Experimental results from prototype implementation of the proposed method verify that our approach provides more accurate results than the conventional FTS does. In addition, the proposed method shows an acceptable amount of response time and battery consumption.
{"title":"Enriching mobile semantic search with web services","authors":"Minjae Song, Sungkwang Eom, Sangjin Shin, Kyong-Ho Lee","doi":"10.1109/ICOSC.2015.7050850","DOIUrl":"https://doi.org/10.1109/ICOSC.2015.7050850","url":null,"abstract":"With the increasing number of mobile devices, there have been many researches on searching and managing a large volume of mobile data. Most of the mobile platforms today provide users with keyword-based full text search (FTS) in order to search for mobile data. Recently, voice search interfaces have been deployed. These search methods, however, query only the keywords given as an input to local databases in mobile devices. Therefore, it is quite difficult to figure out and to provide what a user really wants. To overcome this limitation, we propose a semantic search method for mobile platforms. The proposed method augments the results of semantic search on local databases with their related useful Web information according to the intention and context information of a user. Although there are various semantic search techniques, it is hard to apply the existing methods to mobile devices due to the characteristics of mobile devices such as isolated database structures and limited computing resources. To enable semantic search on mobile devices, we also propose a lightweight mobile ontology. The proposed mobile ontology is also aligned with related Web information to enrich search results. Experimental results from prototype implementation of the proposed method verify that our approach provides more accurate results than the conventional FTS does. In addition, the proposed method shows an acceptable amount of response time and battery consumption.","PeriodicalId":126701,"journal":{"name":"Proceedings of the 2015 IEEE 9th International Conference on Semantic Computing (IEEE ICSC 2015)","volume":"68 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2015-03-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122157489","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 : 2015-03-02DOI: 10.1109/ICOSC.2015.7050778
David Etter, C. Domeniconi
Multimedia retrieval suffers from the lack of common feature representation between a text based query and the visual content of a video repository. One approach to bridging this representation gap is known as query-by-concept, where a query and video are mapped into a common semantic feature space. One of the challenges with using semantic concepts for multimedia retrieval, is that the available vocabulary size is generally not sufficient for representing the content of the query and video. In addition, the lack of training data and visual feature representation often leads to low precision models. In this work, we explore the use of a query-by-concept approach for the multimedia Known Item Search (KIS) problem. We propose a semantic rank learning model, called SemRank, to overcome the challenges of the vocabulary size and lack of training data. First, we construct a semantic fusion model to combine the output from many noisy classifiers. Next, we train a gradient boosted regression tree model, using a semantic feature space derived from the query, video, and query-video similarity. Our approach is evaluated over a large internet video repository, and the results show that query-by-concept can be an effective model for multimedia KIS.
{"title":"SemRank: Semantic rank learning for multimedia retrieval","authors":"David Etter, C. Domeniconi","doi":"10.1109/ICOSC.2015.7050778","DOIUrl":"https://doi.org/10.1109/ICOSC.2015.7050778","url":null,"abstract":"Multimedia retrieval suffers from the lack of common feature representation between a text based query and the visual content of a video repository. One approach to bridging this representation gap is known as query-by-concept, where a query and video are mapped into a common semantic feature space. One of the challenges with using semantic concepts for multimedia retrieval, is that the available vocabulary size is generally not sufficient for representing the content of the query and video. In addition, the lack of training data and visual feature representation often leads to low precision models. In this work, we explore the use of a query-by-concept approach for the multimedia Known Item Search (KIS) problem. We propose a semantic rank learning model, called SemRank, to overcome the challenges of the vocabulary size and lack of training data. First, we construct a semantic fusion model to combine the output from many noisy classifiers. Next, we train a gradient boosted regression tree model, using a semantic feature space derived from the query, video, and query-video similarity. Our approach is evaluated over a large internet video repository, and the results show that query-by-concept can be an effective model for multimedia KIS.","PeriodicalId":126701,"journal":{"name":"Proceedings of the 2015 IEEE 9th International Conference on Semantic Computing (IEEE ICSC 2015)","volume":"22 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2015-03-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127805088","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：