Legal document recommendation is a requisite for specialized domains like socio, law and legal studies is a mandatory requirement as there are no existing specialized search engines which are semantically inclined. In this paper, alegal recommendationsystem Bi-LDR framework has been proposed which is a semantically driven model and based on biclassification technique. Bi-LDR framework which is a framework for legal document recommendations using knowledge synthesis driven by both the user preferences and the existential knowledge. Datasets are subjected to LogisticRegression and Long Short-Term Memory (LSTM) classifiers. By using logistic regression, we can find the best fit model to describe the relationship between independent and dependent variables of a dataset. The dataset is also classified using LSTM, which is a better recurring neural network over the existing traditional neural network accounting for memory efficiency. In order to enrich the relevance computation, semantic similarities are computed using NormalizedCompression Distance (NCD). Entity similarity computations are done with that of the obtained individual terms from the user preferences by using Normalized Compression Distance, Shannon’s Entropy and K-L divergence. An overall Accuracy and F-Measure of 97.62 and 97.61 with the lowest FDR of 0.04 has been achieved by the proposed framework.

Knowledge Synthesis, Legal Document Recommendation, LSTM, Semantic Similarity, Web 3.0