There’s still a long way to go before complex human traits like humor can be properly emulated by artificial intelligence, but Alphabet Inc. is already starting to inject wit into the research effort. The company last week published a machine learning model called “Parsey McParseface” that can automatically map out the linguist structure of any English-language text.

The algorithm, which is hailed as the most accurate of its kind yet, was created using a neural networking system that became available on GitHub at the same time. Alphabet hopes that its contribution will ease the development of virtual assistants and other modern applications that deal with a lot of human-generated information. Equally importantly for the search giant, the move will also cement its position in the open-source machine learning community, which has emerged as a key focus area for the web-scale crowd.

Amazon Inc. joined the fray too last week by releasing the code for DSSTNE, a homegrown deep learning framework that helps power the product recommendation feature on its website. The company says that its framework can run models twice as fast as Alphabet’s competing software when it’s dealing with partial data, which is the norm rather than the exception in use cases like search.

But of course, even an incomplete record of a user’s online browsing habits or shopping history can amount to a tremendous amount of information. To help organizations deal with the vast volumes of data that are needed for their analytics projects, Syncsort Inc. last week adding support for Apache Kafka to its DMX-h ETL software. As a result, customers can now take advantage of the open-source message broker to automate the distribution of information to their various business intelligence systems and number-crunching applications.

Image via Pixabay