The development of a topic in a set of topic documents constitutes a series of person interactions at a specific time and place. Knowing the interactions of the persons involved in these documents is helpful for readers to better comprehend the documents and their topics. To discover those interactions, we need a detection method that can identify text segments containing information about the interactions. Information extraction algorithms can then be utilized to analyze the segments in order to extract interaction tuples and construct an interaction network of topic persons. Moreover, we base on the recognition of reader’s emotion of topic documents to further predict publicity of public figure in the social interaction network. In addition, we plan to develop a flexible approach for topic classification that simulates such process in human perception. We attempt to integrate a variety of knowledge to generate discriminative linguistic patterns for representing essential information in the topic. These learn patterns can be acknowledged as the fundamental knowledge for each topic, and are notably comprehensible for humans. We foresee it to be able to embrace the advantages of both rule-based and machine learning-based approaches that free us from the burden of human labor and can be easily up-scaled to larger corpora. Besides, the generated knowledge can be accumulated and adopted to other tasks without re-tuning the features as required for most machine-learning systems. We believe that it is a promising direction for not only text categorization but also other natural language applications.
|Effective start/end date||11/1/17 → 10/31/18|
- Text Mining
- Topic Summarization
- Interaction Detection
- Interaction Extraction
- Topic Person Interaction Network
Explore the research topics touched on by this project. These labels are generated based on the underlying awards/grants. Together they form a unique fingerprint.