Author: Edward Ma Natural Language Process Abstractive Summarization NLP offers two methods to summarize text. Extractive is the first approach. It is simply a way to extract keywords and sentences from articles. The performance of this approach isn’t very high and there are limitations. The approach is prone to irrelevance, and even redundancy. This second approach is the abstractive one. It generates new sentences based on an article. This requires more sophisticated techniques, but produces better results. Kelly Sikkema, Unsplash. This was used mainly to text. The abstractive method creates an internal semantic representation from the source content and uses this representation to produce a summary closer to what a person might say. To condense the text, extraction may be more effective than extracting it. This transformation is computationally more difficult than extracting. It involves both natural language processing, and sometimes deep knowledge of the subject matter of the original text, in the case where it relates to an exclusive field of knowledge. Daily Use There are many ways to leverage text summarization. News summarization is one of these uses. Detail news can include many paragraphs, and more than 1000 words. The entire news can take several minutes to go through. People find it difficult to absorb a lot of international and local news, which covers many topics, such as sports, financials, and so on. News summarization is a quick way to get a good understanding of the news quickly. We don’t have to spend 5 minutes reading news articles that are not pertinent to us. Instead, we can take seconds and get the basic idea of the news summary. Obi Onyeador, Unsplash. Another use is to find relevant research papers. This section gives us a general idea about the problem and solutions that practitioners are looking for. To determine if this paper is appropriate, we might need to go through 10 pages. Is it possible to summarize news using a machine-learning model? New technology in NLP makes it possible to summarize news. The state-of the-art NLP architecture, such as sequence to-sequence or transformer can be used. You will also need to have a data set that includes both abstract and detail news articles. To train a news summary model, you will need to have a strong machine. You can also leverage APIs to obtain the summarized news. All you need is the news content. You can also get an abstraction of it without using machine learning codes. The pandemic: Working hours. This is the summary generated from my API. It shows that working hours have increased in Britain, Canada, Austria and the United States since last week’s pandemic. According to recent research, home-working workers are more likely to work longer hours than ever before. Growth of Latino-owned companies Here’s another summary generated from my API. This news API shows the difficulties faced by Latinos in obtaining capital from their national banks. Latino-owned companies are growing at a faster rate than other industries. They have grown 34 percent in the past 10 year, compared with just 1% for small businesses. Take Away: I created a website server and trained a deep learning news summarization model. If you’d like to test this API service, drop me a message or email. Would you like to know more? Data Scientist in the Bay Area. I am a Data Scientist in the Bay Area. Connect with me via LinkedIn and Github. Extended Reading Summarize the document using a combination of abstractive and extractive steps. Summarizing News with Abstractive Approach originally appeared in . People are responding and highlighting this story on Medium. Published via
