During today’s e-learning era, students often need to opt for self-study sessions to solve word problems. Word problems are generally solved by transforming them into corresponding algebraic equations. This paper presents an approach for automatically generating such algebraic representations for any word problems in the physics domain. The current system first suggests a suitable template to generate an abstract version of the given word problem. These templates are formed with one or more equations, replacing specific values with generic variables. The motivation for choosing such templates was limiting the number of possible templates to the minimum to reduce further ambiguities. Once the templates are predicted, the values for the variables are tagged by parsing the word problems using named entity recognition (NER) based and rule-based approaches. While developing and assessing our system, we collected a database of Physics word problems and asked experts to determine their algebraic representations manually. Inconsistencies are filtered out using proper preprocessing and normalization measures. Although our system is a preliminary step towards automatically solving physics word problems, the template generation and tagging modules performed their corresponding tasks with around 70% accuracy.