Measuring Semantic Similarity between sentences in Python

Nikaash Puri (~nikaash)




The talk will cover a set of approaches to measure semantic similarity between phrases. Concretely, consider the following sentences. "Harry is running fast" and "Harry is Sprinting". Traditional word-matching algorithms fail to capture the semantic overlap between such phrases. Such similarity scores form essential parts of systems for Information Retrieval (IR), Question Answering, Automatic Grading and several other Natural Language Processing applications. The talk will also briefly discuss a Propositional Logic Generation System (PLGS) that converts sentences in natural language to their propositional logic representations. For instance, the sentence "Harry is running fast" would be represented as:- - obj(Harry, o1) - run(o1, e1) - prop(e1, fast) in propositional logic. Such propositions form the bedrock of powerful reasoning systems for question answering and Information Retrieval.


The audience is expected to have a brief idea about the meaning and applications of natural language processing. Some familiarity with the nltk package in Python would be useful but not essential.

Speaker Info:

Nikaash Puri is the developer and founder of the Reverse Dictionary app ( He is currently working in Adobe systems and has several years of experience in Python programming specifically in the fields of machine learning and Natural Language Processing. He studied at Delhi Technological University (formerly Delhi College of Engineering) and while there developed a Propositional Logic Generation System (PLGS) to transform sentences in natural language into their propositional forms. Such propositional logic representation are then further utilized for question answering and Information Retrieval.

Speaker Links:

Section: Scientific Computing
Type: Talks
Target Audience: Intermediate
Last Updated: