“When measuring your company’s online sentiment, does your natural language processing (NLP) application know the difference between “I like cake.” And “Cake rocks.”? How about “XYZ company is the bomb.” versus “XYZ company bombed.”?
As illustrated above, traditional NLP algorithms are limited in delivering accurate results because their algorithms operate over strings of words without having information about a language’s grammatical rules or colloquialisms.
An alternate approach is called “compositional semantics”, where algorithms incorporate a language’s specific grammatical rules and sentence structures.
In this session, computational scientist Lee J. O’Riordan discusses one such model—distributional compositional semantics (aka DisCo)—that offers significant improvements to NLP results. However, the main challenge in implementation is its need for large classical computational resources.
But according to O’Riordan the solution is quantum implementation, which lowers storage and compute requirements compared to classic HPC implementation.
Tune in to find out how it works, including:
- The development of a quantum-enabled NLP solution
- A presentation of two quantum algorithms: the “closest vector problem” algorithm and the “CSC sentence similarity” algorithm
- Details of added features to the Intel® Quantum Simulator that address quantum algorithm building blocks
- How using a Python* wrapper allows users to quickly develop solutions and to easily analyze results
- A demo of the algorithm in action, plus sample code and output
DisCo was developed by Intel and the Irish Centre for High-End Computing.
Download the software
Intel® Distribution for Python*
Intel® Quantum Simulator