Abstract: 

Financial documents such as news, central bank releases, company earnings calls and press releases can significantly alter market movement. We can therefore mine these to extract sentiment to understand and anticipate relevant market movement, strengthen investment theses, and inform trading strategies.

However, general sentiment approaches, which provide a single sentiment for the entire document produce inaccurate results, because it doesn't factor in the entity for which this sentiment is expressed in the document. For example: ""Covid lockdowns force people to work from home, delaying return to physical offices"" - this is positive for a company like Zoom and negative for companies like WeWork and Uber.
Aspect-based Sentiment Analysis is a variety of sentiment analysis that helps in the improvement of the sentiment classification by identifying a specific sentiment for each company that the document is evaluated for. ABSA identifies the aspects (companies) in the given document and also finds if the aspect mentioned in the document belongs to which class of sentiment.

We will provide a hand-on tutorial of ASBA and provide examples where ASBA results in a better sentiment classification and improved trading performance.

Background Knowledge:
Sentiment Analysis, Word Embeddings

Bio: 

Chandini Jain is the CEO/founder of Auquan - a london based fintech using NLP and AI to distill relevant and impactful information from unstructured text. Prior to Auquan, she worked as a derivatives trader at Optiver in Chicago/Amsterdam and Deutsche Bank. At Auquan, she oversee the development of our machine learning strategies.