Research on Automatic Writing of Football News based on Deep Learning

KAI ZHANG, JIANSHE ZHOU, WENCHAO WANG, XUEQIANG LV, WENYAN ZHANG, WEILI WANG, NAREN TUYA, JINSHENG SHI

Abstract


This paper enclosed the exploration and analysis on the Sports News Generation from Live Webcast Scripts, which is a sub-project in the competition conducted by Shared Tasks in NLPCC- ICCPOL 2016. The task mainly focuses on evaluating document summarization techniques for producing Chinese sports news articles from live webcast scripts. Due to the analytical characteristics of input data set and output news articles, the team found that it is crucial to precisely classify the type of each sentence of the live webcast. Thus, the Character-Level Convolution Networks for Sentence Classification was developed and applied to distinguish the category of sentences. Comparing with the traditional machine learning, our model offers higher accuracy on results but lacks of further processing after extracting key sentences. The team won the second prize in the evaluation.

Keywords


Automatic Writing, SVM, CNN, Language Intelligence, Semantic Understanding


DOI
10.12783/dtcse/cii2017/17258

Refbacks

  • There are currently no refbacks.