SBIR Phase I: Correlating Opinions with Outcomes in Business and Industry: Statistical Modelling of Natural Language Data

这个小企业创新研究第一阶段项目将自然语言处理（NLP）方法与统计和机器学习的回归和分类技术相结合，以确定将意见与商业和工业成果相关联的可行性。研究目标如下：1）确定自动提取的意见信息是否与证券价值轨迹、与资产价值轨迹或与一些其他可测量的值（例如，产品或产品线的市场渗透率）相关联，以及2。使用预测模型来调查哪些特定的媒体来源和意见持有者最有影响力，并描述这些对结果的影响。这项研究建立在以前的意见提取研究的基础上，其中来自自然语言处理的信息提取和机器学习技术被用来处理主观语言。该项目侧重于统计建模的研究，其中从自动提取的意见中获得的特征/预测因子将用于增强商业和工业中信息分析师和决策者感兴趣的预测。如果成功，该项目将导致服务的发展，使决策者能够更好地了解谁和什么是影响他们的公司，客户，竞争对手和市场，在一个环境中，趋势设定的内容来自爆炸性数量的信息源。虽然这个SBIR项目的重点是自动化意见分析在商业和金融市场的使用，将开发的技术和服务是独立于领域的：它们可以同样容易地应用于其他行业或政治、监管政策、外交政策、体育和娱乐的意见和结果。这些方法还可以用于跟踪服务用户感兴趣的较窄主题的意见，例如气候变化，城市化，可持续建筑，他们最喜欢的总统候选人。预计未来三年，文本分析的市场机会将从目前的7亿美元增长到20亿美元。

This Small Business Innovation Research Phase I project combines methods from natural language processing (NLP) with regression and classification techniques from statistics and machine learning to determine the feasibility of associating opinions with outcomes in business and industry. The research objectives are the following: 1) Determine whether or not automatically extracted opinion information is associated with security value trajectories, with asset value trajectories, or with some other measurable value (e.g. market penetration of a product or product line) and, 2. Use predictive models to investigate which specific media sources and opinion holders are most influential and describe these influences on the outcome. The research builds on previous opinion-extraction research where information extraction and machine learning techniques from natural language processing were adapted to handle subjective language. This project focuses on research in statistical modeling where features/predictors derived from automatically extracted opinions will be used to augment predictions of interest to information analysts and decision-makers in business and industry. If successful, the project will result in the development of services that allow decision makers to better understand who and what is influencing their company, customers, competitors and marketplace, in an environment where trend-setting content originates from an exploding number of information sources.Although this SBIR project focuses on the uses of automated opinion analysis in business and the financial market, the techniques and services that will be developed are domain independent: they can just as easily be applied to opinions and outcomes in other industries or in politics, regulatory policy, foreign policy, sports and entertainment. The methods might also be used to track opinions on narrower topics of interest for users of the service, e.g. climate change, urbanization, sustainable architecture, their favorite presidential candidate. The market opportunity for text analytics is projected to grow from the current $700 Million to $2 Billion over the next three years.