Distilling useful insights from a high-volume of text requires scalable analytical techniques employing AI, machine learning, and statistical methods, as well as the robust hardware needed to run these data-intensive projects.
Cornerstone Research uses the most modern and defensible natural language processing methods to enhance the discovery process and support expert testimony. Our capabilities allow us to sift through vast quantities of text, delivering efficient and accurate retrieval of critical information.
- Developed a pipeline for querying documents based on their similarity to documents with known case relevance. Traditional keyword and pattern-matching searches rely on manual review to codify a given document’s most distinctive features. Our automated pipeline considers both the content of individual documents and the entire corpus to generate reliable document queries in a fraction of the time.
- Developed case-specific neural network models to visually identify key documents within a massive document corpus, even when reliable OCR (optical character recognition) is unattainable or inapplicable.
- Developed a tool for querying SEC filings at scale, facilitating text search and other analyses spanning across the SEC’s EDGAR filings database.
Content analysis is the systematic, objective, quantitative assessment of message characteristics. Messages can be text-based (such as news articles, commentary on websites, social media posts), visual (photos, video), or aural (radio programming, speeches). Using specific techniques, researchers can analyze these messages in a manner that is both replicable and consistent with the scientific method.
- Performed content analysis of academic articles using natural language processing and supervised machine learning.
- Automobile Product Liability and Safety Defect Cases Assessed the volume and content of news coverage related to the alleged defect. Our analysis supported the conclusion that the defect was a major news story, and that consumers’ understanding of the defect varied substantially because the media reported the story in a wide range of contexts.
Sentiment analysis is the interpretation and classification of emotions (positive, negative and neutral) within text data using text analysis techniques. Sentiment analysis allows businesses to identify customer sentiment toward products, brands, or services in online conversations and feedback.
- Leveraged existing sentiment models to efficiently generate large-scale, methodologically consistent sentiment scores associated with online reviews.
- Developed custom sentiment models to generate sentiment scores more closely associated with specific products and features, rather than the broad assessment resulting from stock models.