Atousa Duprat

Paul Anderson photo

Research Software Engineer

Dr. Atousa Duprat graduated from the department of computer science at the University of Victoria with a Ph.D. degree in Software Engineering. She was a part of the Rigi group, under the supervision of Dr. Hausi A. Müller. She got her Master of Mathematics from the University of Waterloo, at the Cheriton School of Computer Science. Her main area of research is software engineering. Her research topics include web service search and discovery, self-adaptive system, software architecture, and security in SOA. Currently, She is working at Uber internal search engine. You can find the list of her publications here;

Atousa Duprat is speaking at the following session/s

Supporting Query Tagging/Suggestion in Fusion 4.2

Wednesday | 3:20PM - 4:00PM |

The basic text processing steps include stemming, removing stop words, expanding words via lemmatization, implementing synonyms, and tokenization. These have all been used in the search. However, these processes do not provide insights into the user query. Matching query terms with patterns captured from user’s activities is a useful strategy for query tagging – the idea is to give users access to relevant queries in the corpus that match their interest. Several ML and NLP mechanisms exist at Uber to facilitate query tagging in different domains. We collaborate between teams to integrate the existing tagging architecture with the Fusion 4.2 platform. After integrating the existing models into Fusion we evaluate auto-suggest and query completion since these usually require careful consideration to do well.

Audience Takeaway
If you thought of implementing ML models with Fusion, but always wondered how to start or what is involved in the process. We will cover the entire journey of integration between Fusion and another ML/NLP architecture to fulfill the specific requirements of query classification and suggestion.

Intended Audience
Attendees should be familiar with Fusion concepts. A basic understanding of machine learning or natural language processing is recommended.


Additional speakers