Inference has been a central topic in artificial intelligence from the start, but comparatively little progress has been made on the problem natural language inference (NLI), that is, determining whether a text can be inferred from another text. Language inference is at core of multiple NLP tasks such as information retrieval (IR), question answers (Q&A), machine translation and text summarization by improving the quality of machine’s language understanding. For stronger reasoning and developing machine’s understanding of natural language, we establish text entailment.Stanford Natural Language Inference (SNLI) [Bowman et al. (2015)] and Multi-Natural Language Inference (MNLI) [Bowman et al. (2017)] are two of the standard datasets used for training the models for establishing text entailment. The major limitation of these two datasets is, when machine learning classifier or neural network models are trained on them, they cannot generalize very well on specialized tasks such as, legal data, dialogue systems and larger documents etc. In this thesis, we focused on establishing entailment for legal domain using topological representation of data. To proceed with our aim, we participated in a Competition of Legal Information Extraction and Entailment (COLIEE 2018), organized by University of Alberta, Canada. We also verified our models on SNLI and MNLI datasets to discuss the effectiveness and limitations of our approach.