Based On The Same Field Crfs And Interdisciplinary Under Brand Word Extraction

With the rapid development of Internet, web data is exponentially growing. Most of the data is in the form of text. As an important information carrier, text is a popular subject in related research areas. Brand plays a significant role in business. And we have deeply dived into this issue. With the help of web text, we can systematically analyze a brand. One key step is to extract brand term from text.Brand term extraction is a kind of task pertaining to information extraction. At present, most implementations of information extraction exploit machine learning. Brand term extraction under single and cross domain researched in this paper is based on machine learning. Specifically speaking, by constructing a CRFs model and using related transfer learning algorithms, we realize a system capable of extracting brand terms in single and cross domain.First of all, we make a throughout introduction to two related technologies, CRFs and transfer learning. The general concepts and specific application of both technologies are described. And then this paper illustrates the CRFs pattern and transfer learning pattern in our system, including feature selection, label configuration, and mediate training set establishing. Furthermore, we constructing a corpus fit for brand term extraction and a baseline system for comparison. The details of both the corpus and the baseline system are elaborated. Finally, we conduct a large amount of experiments in single and cross domain and analyze the results of these experiments.In summary, our CRFs-based system, combining with related transfer learning algorithms, can finish the task of extracting brand term in single and cross domain. This paper also discusses several tricky issues regarding CRFs and transfer learning.