基于进化模糊机制的Web新闻挖掘关键技术

The low efficiency in mining has become the main obstruction that the web news mining technologies are applied. The main problems of the existing web news mining methods are the low accuracies of the key mining technologies for such as identification and extraction of web news, classification and summarization of news contents. In response to these issues, this project plans to construct high-performance models of web news mining based on analyzing, borrowing and excavating the traditional technologies and methods. And for constructing the models, not only the systematic exploration of the fuzzy features and the dynamic evolving features of the web news but also the systematic study of the optimization methods of the mining models need to be performed. Firstly, the evolving fuzzy system is imported into the webpage classification model. Then all the webpages can be categorized by using the model. The web news can be identified and extracted from the webpages by using the selected method corresponding to the webpage category. This method can help to improve the accuracy of the web news identification and extraction. Secondly, the evolving fuzzy system is imported into the web news classification model. Then all the web news can be categorized by using the web news classification model. This method can help to improve the accuracy of the web news classification, by means of solving the problem that the traditional approach categorize the web news under the predefined categories and can’t add a new category in an evolving manner. Thirdly, the evolving fuzzy system is imported into the web news summarization model. In the generation of summarization, this model can solve the problem that the key sentence selection rule can’t be updated according to the dynamic evolution of the web news. So, by solving the problem, this model can contribute to the improvement of accuracy of the web news summarization. Lastly, the models and methods can be validated on a simulation and experiment platform.

Web新闻挖掘技术应用的主要障碍是挖掘效率低，已有web新闻挖掘方法在web新闻识别提取、web新闻内容分类、web新闻内容摘要等关键应用挖掘技术方面存在准确度低的问题。项目围绕上述问题，为构建高性能的web新闻挖掘模型，在分析、借鉴和发掘传统技术和方法的基础上，系统地探索web新闻的模糊性和动态演变性特点及挖掘模型优化的方法。首先，将进化模糊机制引入网页分类模型，根据新闻所处网页的类型，分别选择采用相应的新闻识别提取方法，提高识别提取的准确度。其次，将进化模糊机制引入新闻分类模型，解决传统方法存在的按照预先定义的类别分类而不能动态地增加新类别的问题，提高新闻分类的准确度。然后，又将进化模糊机制引入新闻摘要模型，解决传统方法在一个确定的模型中摘要短文的候选关键句选择规则不能根据新闻的动态演变情况而改变的问题，提高新闻摘要的准确度。最后，利用仿真试验平台对模型和方法进行验证。

Web新闻挖掘技术应用的主要障碍是挖掘效率低，已有web新闻挖掘方法在web新闻识别提取、web新闻内容分类、web新闻内容摘要等关键应用挖掘技术方面存在准确度低的问题。围绕上述问题，为构建高性能的web新闻挖掘模型，在分析、借鉴和发掘传统技术和方法的基础上，系统地探索了web新闻的模糊性和动态演变性特点及挖掘模型优化的理论，研究了基于进化模糊机制的Web新闻挖掘相关技术和方法。提出了基于webpage特征提取和进化模糊规则的web新闻网页分类方法、基于角点检测的视频文字区域定位方法、基于BP神经网络和频谱分析的新闻视频文字区域定位方法、基于进化模糊规则的web新闻内容分类方法及基于遗传算法与模糊规则的新闻摘要生成模型和方法等。测试结果表明，研究提出的关键技术和方法大大提高了Web新闻挖掘的准确度和效率。基于研究成果开发的具有进化机能的web新闻挖掘系统，能更好地适用于目前及将来website类型较多、webpage变化多样、web新闻类型和内容实时更新的复杂情况，具有广阔的应用前景。