Thanks to Key Point Summarization, you can extract detailed and actionable insights from a large collection of texts that represent people’s opinions.
When to use Key Point Summarization
When you have a large collection of texts that represent people’s opinions (such as product reviews, survey answers, or comments on social media), it is difficult to understand the key issues that come up in the data. Going over thousands of comments is prohibitively expensive. Automated approaches are often limited to identifying recurring phrases or concepts and the overall sentiment toward them, but do not provide detailed or actionable insights. Thanks to Key Point Summarization (KPS), you can extract these detailed and actionable insights in an organized, hierarchical way.
How Key Point Summarization works
KPS splits the data into sentences and detects the stance of each one (positive, negative, neutral). It then identifies key points from these sentences and matches the remaining sentences to these key points. Any duplications are removed, and related key points are organized in an hierarchical manner. The prevalence of each key point is quantified by the number of matching sentences.
How to use Key Point Summarization
KPS is available only in these notebook environments:
NLP + DO Runtime 24.1 on Python 3.11GPU V100 Runtime 24.1 on Python 3.11
The analysis uses large language models and therefore needs a GPU.
- Run and test KPS with a small subset of the data using a CPU only environment:
NLP + DO Runtime 24.1 on Python 3.11 - Run the KPS at scale on hundreds or thousands of comments using an environment with a GPU:
GPU V100 Runtime 24.1 on Python 3.11
These two environments include the required Python packages for running KPS.
For an example of how to use KPS, see the example notebook.
Parent topic: Notebooks and scripts