‘The amount of savings in time and effort [the search optimization] can deliver for our home offices, for our customers, is incredible.’
--Mayank Gupta, SVP for data, LPL Financial
View the slides from Dr. Alessandro Negro’s presentation at GraphTour DC on how to convert unstructured data silos into powerful knowledge graphs.
View Christophe’s slides from the GraphTour Meetup that took place March 1, 2018.
In 2016, 25% of web searches on Android were made by voice and this percentage is predicted to double by 2018. From Amazon Alexa to Google Home, smartwatches and in-car systems, touch is no longer the primary user interface. In this talk, Alessandro and Christophe will demonstrate how graphs and machine learning are used to create an extracted and enriched graph representation of knowledge from text corpus and other data sources. This representation will then be used to map user intents made by voice to an entry point in this Neo4j backed knowledge graph. Every user interaction will then have to be taken into account at any further steps and we will highlight why graphs are an ideal data structure for keeping an accurate representation of a user context in order to avoid what is called machine or bot amnesia. The speakers will then conclude the session by explaining about how recommendations algorithms are used to predict next steps of the user’s journey.
Neo4j as a viable tool in a relevant search ecosystem demonstrating that it offers not only a suitable model for representing several complex data, like text, user models, business goal, and context information but also providing efficient ways for navigating this data in real time. Moreover at an early stage in the “search improvement process” Neo4j can help relevance engineers to identify salient features describing the content, the user or the search query, later will be helpful to find a way to instruct the search engine about those features through extraction and enrichment.
Moreover, the talk demonstrates how the graph model can provide the right support for all the components of the relevant search and concludes with the presentation of a complete end-to-end infrastructure for providing relevant search in a real use case. It will show how it is integrated with other tools like Elasticsearch, Apache Kafka, Stanford NLP, OpenNLP, Apache Spark.