Keyword Generator

How it works

1. Get your input

Your input can be in the form of an article or a title. Only text is processed and all HTML elements and images are ignored. This input field is sent over to a script which processes it.

---

2. Get insights from Google Speech API

Pass over your text to the Google Speech API and get the follwing paramters from the response (for every word):

• tag
• label
• headTokenIndex

---

3. Identify Patterns

Find recurring patterns amongst the words in a sentence which also qualify as keywords. For instance in the sentence "5 hurdles faced by data nanlytics programs in healthcare", our desired keywords are data analytics and healthcare

We make a note that the words "data" and "analytics" are both NOUNS, have the label NN and the headtokenIndex of "data" is the same as the index of "analytics". On confirming this pattern in other sentences too, we can conclude that when we encounter two NOUNS who follow the above conditions, the probability of the pair of words being a keyword is high. Therefore we highlight it.

Identify other patterns

---

4. Fine tune your results

Since this tool is only is an beta phase, the keywords generated may or may not be accurate. We therefore encourage you to take this tool as a base and improve on it for your needs.

---

Sources

• Overview: https://cloud.google.com/natural-language/
• Setup: https://cloud.google.com/natural-language/docs/quickstart-client-libraries
• Getting valid information: https://cloud.google.com/natural-language/docs/analyzing-syntax

Use Cases

• Summary Generation: Keywords generated can be monitored for their frequency and associations which in turn can be used to prioritise certain sentences and help trim out other sentences. This will help condense lengthy articles into user friendly reading material without losing out on critical information.
• Suggested content: Content can be classified easily and models can be made to suggest similar categories.
• Content Curation: On classifying external content into categories it can serve as reference content are content writers. This will be automated and will make sure that relevant content is always present for writers to base their facts on.
• User Friendly Research Papers: The main challenge of proper proprgation of complex scientific facts and workings is a lack of deep understanding of the relevant content. Relating complex theories and facts with simpler facts will help generate explanations which can be easily understood. The advantage of automating this will ensure that a subject expert only needs to verify but not write the simple explantion since the syntactical structure of the sentence will remain the same across all topics.
• Revision Notes: Combined with a transcription service, summaries of video tutorials or live classes can be generated. These can be used as handouts and will reduce time spent by students making notes.
• Graphs/Infographics Creation: In financial institutions, templatised graphs and infographics can be generated automatically for all reports to ensure better presentation.