چارچوب یادگیری چند مرحله ای برای سیستم هوشمند
کد مقاله | سال انتشار | تعداد صفحات مقاله انگلیسی |
---|---|---|
5647 | 2013 | 11 صفحه PDF |
Publisher : Elsevier - Science Direct (الزویر - ساینس دایرکت)
Journal : Expert Systems with Applications, Volume 40, Issue 9, July 2013, Pages 3378–3388
چکیده انگلیسی
As information technologies advance and user-friendly interfaces develop, the interaction between humans and computers, information devices, and new consumer electronics is increasingly gaining attention. One example that most people can relate to is Apple’s innovation in human–computer interaction which has been used on many products such as iPod and iPhone. Siri, the intelligent personal assistant, is a typical application of machine-learning human–computer interaction. Algorithms in machine learning have been employed in many disciplines, including gesture recognition, speaker recognition, and product recommendation systems. While the existing learning algorithms compute and learn from a large quantity of data, this study proposes an improved learning to rank algorithm named MultiStageBoost. In addition to ranking data through multiple stages, the MultiStageBoost algorithm significantly improves the existing algorithms in two ways. Firstly, it classifies and filters data to small quantities and applies the Boosting algorithm to achieve faster ranking performance. Secondly, it enhances the original binary classification by using the reciprocal of fuzzily weighted membership as the ranking distance. The importance of data is revealed in their ranked positions. Usually data ranked in the front are given more attention than those ranked in the middle. For example, after ranking 10,000 pieces of data, the top 10, or at most 100, are the most important and relevant. Whether the data after the top ones are ranked precisely does not really matter. Due to this reason, this study has made improvement on the conventional methods of the pair-wise ranking approach. Not only are data classified and ranked binarily, they are also given different weights depending on whether they are concordant or discordant. Incorporating the concept of weighting into the ranking distance allows us to increase the precision of ranking. Results from experiments demonstrate that our proposed algorithm outperforms the conventional methods in three evaluation measures: P@n, MAP, and NDCG. MultiStageBoost was then applied to speech recognition. However, we do not aim to improve the technology of speech recognition, but simply hope to provide evidences that MultiStageBoost can be used in the classification and ranking in speech recognition. Experiments show that the recognition optimization procedures established by this study are able to increase the recognition rate to over 95% in the personal computing device and industrial personal computer. It is expected that in the future this voice management system will accurately and effectively identify speakers answering the voice response questionnaire and will successfully carry out the functions in the choice of answers, paying the way for the formation of a virtual customer service person.
مقدمه انگلیسی
As more data are computed and stored in the clouds, the data we collect are getting larger in quantity, more complex in feature, and more diverse in type. Being overwhelmed by flood of information, a common goal of many researchers is to provide methods and strategies to simplify data. If the most important data can be extracted from the boundary less data universe, decision makers will be able to exploit the hidden knowledge and wisdom for their references. One of the common strategies in data simplification is to choose features that can be used to classify data, especially critical when data amount is enormous. After the features are determined, data can then be selected, classified, and computed. Machine learning is a discipline which focuses on processing raw data in order to retrieve useful information. It has been applied to human–computer interaction and information systems, making computers, information devices, and new consumer electronics smarter and people’s lives more convenient. Learning to rank is a research area which combines information retrieval and machine learning. The purpose of learning to rank is to let the computer automatically generate a ranking function based on the training data which consists of the perfect ranking lists of documents for each query. The performance of the ranking function is evaluated by comparing it to the correct ranking results. However, the evaluation method used here is somewhat different from regular methods. Since the ranking function compares data in two lists, not only the data at the matched position in each list are compared (for example, pair-wise comparison of data at position n in list A and position n in list B), the relationships of similarity of the two lists must also be taken into consideration while the evaluation is performed. The learning to rank algorithm proposed in this paper is developed from AdaBoost. AdaBoost pools together a series of rough weighting rules to create highly precise rules. Its advantages include easy execution and highly precise classification. However, it also has drawbacks. For instance, AdaBoost is sensitive to noise, which is a disadvantage shared by most ranking and classification algorithms. Further, its execution performance relies on the weak classifiers used. AdaBoost boasts high efficiency and detection rate, hence is suitable for handling high-dimensional feature data. Such Boosting-based ranking algorithm is very effective in human–computer interaction and is even applicable to recognition technologies.
نتیجه گیری انگلیسی
This study proposes an improved learning to rank algorithm, combining the merits of Boosting and fuzzy classification. In addition to ranking data through multiple stages, MultiStageBoost significantly improves the existing algorithms in two ways. Firstly, it classifies and filters data to small quantities and applies the Boosting algorithm to achieve faster ranking performance. Secondly, it enhances the original binary classification by using the reciprocal of fuzzily weighted membership as the ranking distance. Results from experiments demonstrate that our proposed algorithm outperforms the conventional methods in three evaluation measures: P@n, MAP, and NDCG. Data ranked in the front are usually given more attention than those ranked in the middle. For example, after ranking 10,000 pieces of data, the top 10, or at most 100, are the most important and relevant. Whether the data after the top ones are ranked precisely does not really matter. Due to this reason, this study has made improvement on the conventional methods of the pair-wise ranking approach. Not only are data classified and ranked binarily, they are also given different weights depending on whether they are concordant or discordant. Incorporating the concept of weighting into the ranking distance allows us to increase the precision of ranking. We have also proved that MultiStageBoost can be of valuable use in applications in three different areas. Our study does not aim to analyze and recognize voices but simply rank and recognize features extracted by using other voice technologies. Although our method is able to reach 94.5% speaker identification rate, such performance is not satisfactory for the application of guard and security systems, which require 100% identification rate. Our future goal would be to enhance its performance for security system applications. In comparison, since voice questionnaire does not require a precision rate as high as security systems, our method in this HCI application is promising for commercialization.