Query-based Text Summarization using Averaged Query

Automatic text summarization is one of the most common problems in natural language proc e ssin g a nd machine learning. Text summarization usually works by shortening a given passage and conveying the general meaning of the passage. There are two approaches to this: extraction-based summarization and abstraction based summarization. Extraction - b a s e d summarization, while easier to implement, is usually grammatically incorrect. Abstraction based summarization overcomes this by framing its own sentences using grammatical knowledge of the language and is therefore much harder to implement. There is an ever-increasing amount of unstructured data in the world, so situations can arise where it is only necessary to extract the summary of a part of the given data. This is the case when a person wants to learn something about a certain topic but often has to skim through a lot of unrelated or unnecessary information. While this is not a problem for small amounts o f d a t a , it can quickly turn into a burden as the size of the data increases. This can lead to a loss of focus or interest in the topic. In this work, the proposed technique allows the user to enter a keyword and get a summary related to say query using the word frequencies to find the most relevant words. We also use cosine similarities to remove redundant sentences from t h e s u m m a r y. This query based text summarization technique produces a unique summary for every unique keyword. Better readability can be achieved by using abstractive text summarization.